diff --git "a/test-dataset.json" "b/test-dataset.json"
deleted file mode 100644--- "a/test-dataset.json"
+++ /dev/null
@@ -1,108 +0,0 @@
-[
-  {
-    "repo": "celery/celery",
-    "pull_number": 2840,
-    "instance_id": "celery__celery-2840",
-    "issue_numbers": [
-      "1628"
-    ],
-    "base_commit": "045b52f1450d6d5cc500e0057a4b498250dc5692",
-    "patch": "diff --git a/celery/app/defaults.py b/celery/app/defaults.py\n--- a/celery/app/defaults.py\n+++ b/celery/app/defaults.py\n@@ -132,6 +132,7 @@ def __repr__(self):\n         'REDIS_DB': Option(type='int', **_REDIS_OLD),\n         'REDIS_PASSWORD': Option(type='string', **_REDIS_OLD),\n         'REDIS_MAX_CONNECTIONS': Option(type='int'),\n+        'REJECT_ON_WORKER_LOST': Option(type='bool'),\n         'RESULT_BACKEND': Option(type='string'),\n         'RESULT_DB_SHORT_LIVED_SESSIONS': Option(False, type='bool'),\n         'RESULT_DB_TABLENAMES': Option(type='dict'),\ndiff --git a/celery/app/task.py b/celery/app/task.py\n--- a/celery/app/task.py\n+++ b/celery/app/task.py\n@@ -220,6 +220,12 @@ class Task(object):\n     #: :setting:`CELERY_ACKS_LATE` setting.\n     acks_late = None\n \n+    #: When CELERY_ACKS_LATE is set to True, the default behavior to\n+    #: handle worker crash is to acknowledge the message. Setting\n+    #: this to true allows the message to be rejected and requeued so\n+    #: it will be executed again by another worker.\n+    reject_on_worker_lost = None\n+\n     #: Tuple of expected exceptions.\n     #:\n     #: These are errors that are expected in normal operation\n@@ -248,6 +254,7 @@ class Task(object):\n         ('rate_limit', 'CELERY_DEFAULT_RATE_LIMIT'),\n         ('track_started', 'CELERY_TRACK_STARTED'),\n         ('acks_late', 'CELERY_ACKS_LATE'),\n+        ('reject_on_worker_lost', 'CELERY_REJECT_ON_WORKER_LOST'),\n         ('ignore_result', 'CELERY_IGNORE_RESULT'),\n         ('store_errors_even_if_ignored',\n             'CELERY_STORE_ERRORS_EVEN_IF_IGNORED'),\ndiff --git a/celery/worker/request.py b/celery/worker/request.py\n--- a/celery/worker/request.py\n+++ b/celery/worker/request.py\n@@ -326,7 +326,6 @@ def on_retry(self, exc_info):\n     def on_failure(self, exc_info, send_failed_event=True, return_ok=False):\n         \"\"\"Handler called if the task raised an exception.\"\"\"\n         task_ready(self)\n-\n         if isinstance(exc_info.exception, MemoryError):\n             raise MemoryError('Process got: %s' % (exc_info.exception,))\n         elif isinstance(exc_info.exception, Reject):\n@@ -352,7 +351,13 @@ def on_failure(self, exc_info, send_failed_event=True, return_ok=False):\n                 )\n         # (acks_late) acknowledge after result stored.\n         if self.task.acks_late:\n-            self.acknowledge()\n+            reject_and_requeue = (self.task.reject_on_worker_lost and\n+                isinstance(exc, WorkerLostError) and\n+                self.delivery_info.get('redelivered', False) is False)\n+            if reject_and_requeue:\n+                self.reject(requeue=True)\n+            else:\n+                self.acknowledge()\n \n         if send_failed_event:\n             self.send_event(\n",
-    "test_patch": "diff --git a/celery/tests/worker/test_request.py b/celery/tests/worker/test_request.py\n--- a/celery/tests/worker/test_request.py\n+++ b/celery/tests/worker/test_request.py\n@@ -325,6 +325,20 @@ def test_on_failure_Reject_rejects_with_requeue(self):\n             req_logger, req.connection_errors, True,\n         )\n \n+    def test_on_failure_WrokerLostError_rejects_with_requeue(self):\n+        einfo = None\n+        try:\n+            raise WorkerLostError()\n+        except:\n+            einfo = ExceptionInfo(internal=True)\n+        req = self.get_request(self.add.s(2, 2))\n+        req.task.acks_late = True\n+        req.task.reject_on_worker_lost = True\n+        req.delivery_info['redelivered'] = False\n+        req.on_failure(einfo)\n+        req.on_reject.assert_called_with(req_logger,\n+            req.connection_errors, True)\n+\n     def test_tzlocal_is_cached(self):\n         req = self.get_request(self.add.s(2, 2))\n         req._tzlocal = 'foo'\n",
-    "problem_statement": "Message being acknowledged on WorkerLostError when CELERY_ACKS_LATE=True\nWhen using celery v3.0.24, with `CELERY_ACKS_LATE = True` , if the OOM killer kills the celery worker, then the worker acknowledges the message.\nAs per [this](https://github.com/celery/celery/commit/e810420c) commit. The `exc_info.internal` comes in as `false`, which means it is not a internal error, due to which the message is acknowledged.\nThe desirable behaviour, in such a case would be to not acknowledge the message (and be able to know, whether its a OOM error), so that some other worker can pick it up. \nAs a workaround, I've commented out the [code](https://github.com/siddharth96/celery/commit/427695d1b23034dadda85fd7a48f7367831be4fa), where celery acknowledges the message, because in such a case, message will be lost.\n\n",
-    "hints_text": "This is deliberate as if a task is killed it may mean that the next invocation will also cause the same to happen.  If the task is redelivered it may cause a loop where the same conditions occur again and again.  Also, sadly you cannot distinguish processes killed by OOM from processes killed by other means, and if an administrator kills -9 a task going amok, you usually don't want that task to be called again.\n\nThere could be a configuration option for not acking terminated tasks, but I'm not sure how useful that would be.\nA better solution could be to use `basic_reject(requeue=False)` instead of `basic_ack`, that way you can configure\na dead letter queue so that the killed tasks will be sent to a queue for manual inspection.\n\nI must say, regardless of the status of this feature request, the documentation is misleading. Specifically, [this FAQ makes it seem that process failures would NOT acknowledge messages](http://celery.readthedocs.org/en/latest/faq.html#faq-acks-late-vs-retry). And [this FAQ boldface states](http://celery.readthedocs.org/en/latest/faq.html#id54) that in the event of a kill signal (9), that acks_late will allow the task to re-run (which again, is patently wrong based on this poorly documented behavior). Nowhere in the docs have I found that if the process _dies_, the message will be acknowledged, regardless of acks_late or not. (for instance, I have a set of 10k+ tasks, and some 1% of tasks wind up acknowledged but incomplete when a WorkerLostError is thrown in connection with the worker, although there are no other errors of any kind in any of my logs related to that task).\n\nTL;DR at the least, appropriately document the current state when describing the functionality and limitations of acks_late. A work-around would be helpful -- I'm not sure I understand the solution of using `basic_reject`, although I'll keep looking into it.\n\nThe docs are referring to killing the worker process with KILL, not the child processes.  The term worker will always refer to the worker instance, not the pool processes.  The section within about acks_late is probably not very helpful and should be removed\n",
-    "created_at": "2015-10-06T05:34:34Z",
-    "version": "1.0"
-  },
-  {
-    "repo": "NVIDIA/NeMo",
-    "pull_number": 473,
-    "instance_id": "NVIDIA__NeMo-473",
-    "issue_numbers": [
-      "406"
-    ],
-    "base_commit": "ba4616f1f011d599de87f0cb3315605e715d402a",
-    "patch": "diff --git a/nemo/backends/pytorch/actions.py b/nemo/backends/pytorch/actions.py\n--- a/nemo/backends/pytorch/actions.py\n+++ b/nemo/backends/pytorch/actions.py\n@@ -937,26 +937,16 @@ def __extract_dynamic_axes(port_name: str, ntype: NeuralType, dynamic_axes: defa\n                     if axis.kind == AxisKind.Batch or axis.kind == AxisKind.Time:\n                         dynamic_axes[port_name].append(ind)\n \n-        # This is a hack for Jasper to Jarvis export -- need re-design for this\n-        inputs_to_drop = set()\n-        outputs_to_drop = set()\n-        if type(module).__name__ == \"JasperEncoder\":\n-            logging.info(\n-                \"Module is JasperEncoder. We are removing input and output length ports since they are not needed for \"\n-                \"deployment\"\n-            )\n-            inputs_to_drop.add(\"length\")\n-            outputs_to_drop.add(\"encoded_lengths\")\n-\n+        # extract dynamic axes and remove unnecessary inputs/outputs\n         # for input_ports\n         for port_name, ntype in module.input_ports.items():\n-            if port_name in inputs_to_drop:\n+            if port_name in module._disabled_deployment_input_ports:\n                 input_names.remove(port_name)\n                 continue\n             __extract_dynamic_axes(port_name, ntype, dynamic_axes)\n         # for output_ports\n         for port_name, ntype in module.output_ports.items():\n-            if port_name in outputs_to_drop:\n+            if port_name in module._disabled_deployment_output_ports:\n                 output_names.remove(port_name)\n                 continue\n             __extract_dynamic_axes(port_name, ntype, dynamic_axes)\ndiff --git a/nemo/collections/asr/jasper.py b/nemo/collections/asr/jasper.py\n--- a/nemo/collections/asr/jasper.py\n+++ b/nemo/collections/asr/jasper.py\n@@ -118,14 +118,14 @@ def output_ports(self):\n         }\n \n     @property\n-    def disabled_deployment_input_ports(self):\n+    def _disabled_deployment_input_ports(self):\n         return set([\"length\"])\n \n     @property\n-    def disabled_deployment_output_ports(self):\n+    def _disabled_deployment_output_ports(self):\n         return set([\"encoded_lengths\"])\n \n-    def prepare_for_deployment(self):\n+    def _prepare_for_deployment(self):\n         m_count = 0\n         for m in self.modules():\n             if type(m).__name__ == \"MaskedConv1d\":\ndiff --git a/nemo/core/neural_factory.py b/nemo/core/neural_factory.py\n--- a/nemo/core/neural_factory.py\n+++ b/nemo/core/neural_factory.py\n@@ -610,7 +610,7 @@ def deployment_export(\n             input_example: sometimes tracing will require input examples\n             output_example: Should match inference on input_example\n         \"\"\"\n-        module.prepare_for_deployment()\n+        module._prepare_for_deployment()\n \n         return self._trainer.deployment_export(\n             module=module,\ndiff --git a/nemo/core/neural_modules.py b/nemo/core/neural_modules.py\n--- a/nemo/core/neural_modules.py\n+++ b/nemo/core/neural_modules.py\n@@ -393,7 +393,7 @@ def output_ports(self) -> Optional[Dict[str, NeuralType]]:\n         \"\"\"\n \n     @property\n-    def disabled_deployment_input_ports(self) -> Optional[Set[str]]:\n+    def _disabled_deployment_input_ports(self) -> Optional[Set[str]]:\n         \"\"\"Returns names of input ports that will not be included in an export\n \n         Returns:\n@@ -402,7 +402,7 @@ def disabled_deployment_input_ports(self) -> Optional[Set[str]]:\n         return set([])\n \n     @property\n-    def disabled_deployment_output_ports(self) -> Optional[Set[str]]:\n+    def _disabled_deployment_output_ports(self) -> Optional[Set[str]]:\n         \"\"\"Returns names of output ports that will not be included in an export\n \n         Returns:\n@@ -410,7 +410,7 @@ def disabled_deployment_output_ports(self) -> Optional[Set[str]]:\n         \"\"\"\n         return set([])\n \n-    def prepare_for_deployment(self) -> None:\n+    def _prepare_for_deployment(self) -> None:\n         \"\"\"Patch the module if required to prepare for deployment\n \n         \"\"\"\n",
-    "test_patch": "diff --git a/tests/unit/core/test_deploy_export.py b/tests/unit/core/test_deploy_export.py\n--- a/tests/unit/core/test_deploy_export.py\n+++ b/tests/unit/core/test_deploy_export.py\n@@ -46,9 +46,11 @@\n import nemo.collections.nlp.nm.trainables.common.token_classification_nm\n from nemo import logging\n \n+TRT_ONNX_DISABLED = False\n+\n # Check if the required libraries and runtimes are installed.\n+# Only initialize GPU after this runner is activated.\n try:\n-    # Only initialize GPU after this runner is activated.\n     import pycuda.autoinit\n \n     # This import causes pycuda to automatically manage CUDA context creation and cleanup.\n@@ -63,16 +65,17 @@\n     )\n     from .tensorrt_runner import TensorRTRunnerV2\n except:\n-    # Skip tests.\n-    pytestmark = pytest.mark.skip\n+    TRT_ONNX_DISABLED = True\n \n \n @pytest.mark.usefixtures(\"neural_factory\")\n class TestDeployExport(TestCase):\n-    def setUp(self):\n-        logging.setLevel(logging.WARNING)\n-        device = nemo.core.DeviceType.GPU\n-        self.nf = nemo.core.NeuralModuleFactory(backend=nemo.core.Backend.PyTorch, placement=device)\n+    # def setUp(self):\n+    #    super().setUp()\n+\n+    #    logging.setLevel(logging.WARNING)\n+    #    device = nemo.core.DeviceType.GPU\n+    #    self.nf = nemo.core.NeuralModuleFactory(backend=nemo.core.Backend.PyTorch, placement=device)\n \n     def __test_export_route(self, module, out_name, mode, input_example=None):\n         out = Path(out_name)\n@@ -112,7 +115,13 @@ def __test_export_route(self, module, out_name, mode, input_example=None):\n             loader_cache = DataLoaderCache(data_loader)\n             profile_shapes = OrderedDict()\n             names = list(module.input_ports) + list(module.output_ports)\n-\n+            names = list(\n+                filter(\n+                    lambda x: x\n+                    not in (module._disabled_deployment_input_ports | module._disabled_deployment_output_ports),\n+                    names,\n+                )\n+            )\n             if isinstance(input_example, tuple):\n                 si = [tuple(input_example[i].shape) for i in range(len(input_example))]\n             elif isinstance(input_example, OrderedDict):\n@@ -152,7 +161,7 @@ def __test_export_route(self, module, out_name, mode, input_example=None):\n                 input_names = list(input_metadata.keys())\n                 for i in range(len(input_names)):\n                     input_name = input_names[i]\n-                    if input_name in module.disabled_deployment_input_ports:\n+                    if input_name in module._disabled_deployment_input_ports:\n                         continue\n                     inputs[input_name] = (\n                         input_example[input_name].cpu().numpy()\n@@ -209,8 +218,8 @@ def __test_export_route(self, module, out_name, mode, input_example=None):\n             ort_inputs = ort_session.get_inputs()\n             for i in range(len(input_names)):\n                 input_name = input_names[i]\n-                if input_name in module.disabled_deployment_input_ports:\n-                    input_name = ort_inputs[i].name\n+                if input_name in module._disabled_deployment_input_ports:\n+                    continue\n                 inputs[input_name] = (\n                     input_example[input_name].cpu().numpy()\n                     if isinstance(input_example, OrderedDict)\n@@ -263,9 +272,10 @@ def __test_export_route(self, module, out_name, mode, input_example=None):\n \n     def __test_export_route_all(self, module, out_name, input_example=None):\n         if input_example is not None:\n-            self.__test_export_route(\n-                module, out_name + '.trt.onnx', nemo.core.DeploymentFormat.TRTONNX, input_example=input_example\n-            )\n+            if not TRT_ONNX_DISABLED:\n+                self.__test_export_route(\n+                    module, out_name + '.trt.onnx', nemo.core.DeploymentFormat.TRTONNX, input_example=input_example\n+                )\n             self.__test_export_route(module, out_name + '.onnx', nemo.core.DeploymentFormat.ONNX, input_example)\n             self.__test_export_route(module, out_name + '.pt', nemo.core.DeploymentFormat.PYTORCH, input_example)\n         self.__test_export_route(module, out_name + '.ts', nemo.core.DeploymentFormat.TORCHSCRIPT, input_example)\n@@ -323,9 +333,7 @@ def test_jasper_encoder(self):\n         )\n \n         self.__test_export_route_all(\n-            module=jasper_encoder,\n-            out_name=\"jasper_encoder\",\n-            input_example=(torch.randn(16, 64, 256).cuda(), torch.randn(256).cuda()),\n+            module=jasper_encoder, out_name=\"jasper_encoder\", input_example=torch.randn(16, 64, 256).cuda(),\n         )\n \n     @pytest.mark.unit\n@@ -343,7 +351,5 @@ def test_quartz_encoder(self):\n         )\n \n         self.__test_export_route_all(\n-            module=jasper_encoder,\n-            out_name=\"quartz_encoder\",\n-            input_example=(torch.randn(16, 64, 256).cuda(), torch.randint(20, (16,)).cuda()),\n+            module=jasper_encoder, out_name=\"quartz_encoder\", input_example=torch.randn(16, 64, 256).cuda(),\n         )\n",
-    "problem_statement": "Jasper Encoder Export failed\nThe export of Jasper Encoder is failing. I am using the core API [deployment_export](https://nvidia.github.io/NeMo/api-docs/nemo.html#nemo.core.neural_factory.NeuralModuleFactory.deployment_export) like in the script:  https://github.com/NVIDIA/NeMo/blob/403238f82d26879ba5fca53fbf75b3cdc70fb49b/scripts/export_jasper_to_onnx.py#L92\r\n\r\nI believe the issue (as shown below) is that the` input_example` provided does not match the `output_example`. \r\n\r\n```\r\n/opt/conda/lib/python3.6/site-packages/torch/jit/__init__.py:1023: TracerWarning: Output nr 1. of the traced function does not match the corresponding output of the Python function. Detailed error:\r\nNot within tolerance rtol=1e-05 atol=1e-05 at input[0, 870, 67] (0.6547648906707764 vs. 0.6546438932418823) and 812 other locations (0.00%)\r\n  check_tolerance, _force_outplace, True, _module_class)\r\n\r\n[NeMo E 2020-02-23 19:10:07 actions:1023] module export failed for JasperEncoder with exception number of output names provided (2) exceeded number of outputs (1)\r\n```\r\n\r\n**What is the correct `input_example` and `output_example` to export JasperEncoder?** \r\n\r\n\r\nThe full output can be seen here:\r\n```\r\nadrianaf@2a520c7abb1e:/tmp/NeMo$ ! python /tmp/NeMo/scripts/export_jasper_to_onnx.py --config /raid/datasets/asr/data/config_files/WSJ-test_acoustic_quartznet15x5.yaml --nn_encoder /home/adrianaf/projects/nemo_asr_app/models/quartznet15x5/JasperEncoder-STEP-247400.pt --nn_decoder /home/adrianaf/projects/nemo_asr_app/models/quartznet15x5/JasperDecoderForCTC-STEP-247400.pt --onnx_encoder /raid/datasets/asr/data/models/ONNX/pre-trained_encoder.onnx --onnx_decoder /raid/datasets/asr/data/models/ONNX/pre-trained_decoder.onnx\r\n/opt/conda/lib/python3.6/site-packages/torchvision/io/_video_opt.py:17: UserWarning: video reader based on ffmpeg c++ ops not available\r\n  warnings.warn(\"video reader based on ffmpeg c++ ops not available\")\r\n/tmp/NeMo/nemo/collections/asr/audio_preprocessing.py:48: UserWarning: Could not import torchaudio. Some features might not work.\r\n  warnings.warn('Could not import torchaudio. Some features might not work.')\r\n[NeMo I 2020-02-23 19:09:42 export_jasper_to_onnx:48] Loading config file...\r\n[NeMo I 2020-02-23 19:09:42 export_jasper_to_onnx:52] Determining model shape...\r\n[NeMo I 2020-02-23 19:09:42 export_jasper_to_onnx:60]   Num encoder input features: 64\r\n[NeMo I 2020-02-23 19:09:42 export_jasper_to_onnx:61]   Num decoder input features: 1024\r\n[NeMo W 2020-02-23 19:09:42 deprecated:68] Function ``_get_trainer`` is deprecated. It is going to be removed in the future version.\r\n[NeMo I 2020-02-23 19:09:42 export_jasper_to_onnx:65] Initializing models...\r\n[NeMo I 2020-02-23 19:09:45 export_jasper_to_onnx:76] Loading checkpoints...\r\n[NeMo I 2020-02-23 19:09:45 export_jasper_to_onnx:91] Exporting encoder...\r\n[NeMo W 2020-02-23 19:09:45 neural_factory:627] Turned off 170 masked convolutions\r\n[NeMo I 2020-02-23 19:09:45 actions:937] Module is JasperEncoder. We are removing input and output length ports since they are not needed for deployment\r\n[NeMo W 2020-02-23 19:09:46 deprecated:68] Function ``local_parameters`` is deprecated. It is going to be removed in the 0.11 version.\r\n/opt/conda/lib/python3.6/site-packages/torch/jit/__init__.py:1023: TracerWarning: Output nr 1. of the traced function does not match the corresponding output of the Python function. Detailed error:\r\nNot within tolerance rtol=1e-05 atol=1e-05 at input[0, 870, 67] (0.6547648906707764 vs. 0.6546438932418823) and 812 other locations (0.00%)\r\n  check_tolerance, _force_outplace, True, _module_class)\r\n[NeMo E 2020-02-23 19:10:07 actions:1023] module export failed for JasperEncoder with exception number of output names provided (2) exceeded number of outputs (1)\r\n[NeMo I 2020-02-23 19:10:07 export_jasper_to_onnx:98] Exporting decoder...\r\ngraph(%encoder_output : Float(1, 1024, 128),\r\n      %1 : Float(29, 1024, 1),\r\n      %2 : Float(29)):\r\n  %3 : Float(1, 29, 128) = onnx::Conv[dilations=[1], group=1, kernel_shape=[1], pads=[0, 0], strides=[1]](%encoder_output, %1, %2), scope: JasperDecoderForCTC/Sequential[decoder_layers]/Conv1d[0] # /opt/conda/lib/python3.6/site-packages/torch/nn/modules/conv.py:202:0\r\n  %4 : Float(1, 128, 29) = onnx::Transpose[perm=[0, 2, 1]](%3), scope: JasperDecoderForCTC # /tmp/NeMo/nemo/collections/asr/jasper.py:235:0\r\n  %output : Float(1, 128, 29) = onnx::LogSoftmax[axis=2](%4), scope: JasperDecoderForCTC # /opt/conda/lib/python3.6/site-packages/torch/nn/functional.py:1317:0\r\n  return (%output)\r\n\r\n/opt/conda/lib/python3.6/site-packages/torch/onnx/utils.py:774: UserWarning: No names were found for specified dynamic axes of provided input.Automatically generated names will be applied to each dynamic axes of input encoder_output\r\n  'Automatically generated names will be applied to each dynamic axes of input {}'.format(key))\r\n/opt/conda/lib/python3.6/site-packages/torch/onnx/utils.py:774: UserWarning: No names were found for specified dynamic axes of provided input.Automatically generated names will be applied to each dynamic axes of input output\r\n  'Automatically generated names will be applied to each dynamic axes of input {}'.format(key))\r\n[NeMo I 2020-02-23 19:10:07 export_jasper_to_onnx:105] Export completed successfully.\r\n```\r\n\n",
-    "hints_text": "",
-    "created_at": "2020-03-10T03:03:23Z",
-    "version": "1.0"
-  },
-  {
-    "repo": "NVIDIA/NeMo",
-    "pull_number": 3632,
-    "instance_id": "NVIDIA__NeMo-3632",
-    "issue_numbers": [
-      "3613"
-    ],
-    "base_commit": "022f0292aecbc98d591d49423d5045235394f793",
-    "patch": "diff --git a/nemo_text_processing/text_normalization/__init__.py b/nemo_text_processing/text_normalization/__init__.py\n--- a/nemo_text_processing/text_normalization/__init__.py\n+++ b/nemo_text_processing/text_normalization/__init__.py\n@@ -21,7 +21,7 @@\n except (ModuleNotFoundError, ImportError):\n     logging.warning(\n         \"`pynini` is not installed ! \\n\"\n-        \"Please run the `nemo_text_processing/setup.sh` script\"\n+        \"Please run the `nemo_text_processing/setup.sh` script \"\n         \"prior to usage of this toolkit.\"\n     )\n \ndiff --git a/nemo_text_processing/text_normalization/en/graph_utils.py b/nemo_text_processing/text_normalization/en/graph_utils.py\n--- a/nemo_text_processing/text_normalization/en/graph_utils.py\n+++ b/nemo_text_processing/text_normalization/en/graph_utils.py\n@@ -159,7 +159,7 @@ def convert_space(fst) -> 'pynini.FstLike':\n     \"\"\"\n     Converts space to nonbreaking space.\n     Used only in tagger grammars for transducing token values within quotes, e.g. name: \"hello kitty\"\n-    This is making transducer significantly slower, so only use when there could be potential spaces within quotes, otherwise leave it. \n+    This is making transducer significantly slower, so only use when there could be potential spaces within quotes, otherwise leave it.\n \n     Args:\n         fst: input fst\n@@ -208,9 +208,9 @@ def add_tokens(self, fst) -> 'pynini.FstLike':\n         \"\"\"\n         Wraps class name around to given fst\n \n-        Args: \n+        Args:\n             fst: input fst\n-        \n+\n         Returns:\n             Fst: fst\n         \"\"\"\ndiff --git a/nemo_text_processing/text_normalization/en/taggers/punctuation.py b/nemo_text_processing/text_normalization/en/taggers/punctuation.py\n--- a/nemo_text_processing/text_normalization/en/taggers/punctuation.py\n+++ b/nemo_text_processing/text_normalization/en/taggers/punctuation.py\n@@ -22,7 +22,7 @@\n     import pynini\n     from pynini.lib import pynutil\n \n-    PYNINI_AVAILABLE = False\n+    PYNINI_AVAILABLE = True\n except (ModuleNotFoundError, ImportError):\n     PYNINI_AVAILABLE = False\n \ndiff --git a/nemo_text_processing/text_normalization/en/verbalizers/whitelist.py b/nemo_text_processing/text_normalization/en/verbalizers/whitelist.py\n--- a/nemo_text_processing/text_normalization/en/verbalizers/whitelist.py\n+++ b/nemo_text_processing/text_normalization/en/verbalizers/whitelist.py\n@@ -12,8 +12,6 @@\n # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n # See the License for the specific language governing permissions and\n # limitations under the License.\n-\n-\n from nemo_text_processing.text_normalization.en.graph_utils import NEMO_CHAR, NEMO_SIGMA, GraphFst, delete_space\n \n try:\n@@ -21,6 +19,7 @@\n     from pynini.lib import pynutil\n \n     PYNINI_AVAILABLE = True\n+\n except (ModuleNotFoundError, ImportError):\n     PYNINI_AVAILABLE = False\n \ndiff --git a/nemo_text_processing/text_normalization/en/verbalizers/word.py b/nemo_text_processing/text_normalization/en/verbalizers/word.py\n--- a/nemo_text_processing/text_normalization/en/verbalizers/word.py\n+++ b/nemo_text_processing/text_normalization/en/verbalizers/word.py\n@@ -12,7 +12,6 @@\n # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n # See the License for the specific language governing permissions and\n # limitations under the License.\n-\n from nemo_text_processing.text_normalization.en.graph_utils import NEMO_CHAR, NEMO_SIGMA, GraphFst, delete_space\n \n try:\n@@ -20,6 +19,7 @@\n     from pynini.lib import pynutil\n \n     PYNINI_AVAILABLE = True\n+\n except (ModuleNotFoundError, ImportError):\n     PYNINI_AVAILABLE = False\n \ndiff --git a/nemo_text_processing/text_normalization/es/__init__.py b/nemo_text_processing/text_normalization/es/__init__.py\nnew file mode 100644\n--- /dev/null\n+++ b/nemo_text_processing/text_normalization/es/__init__.py\n@@ -0,0 +1,15 @@\n+# Copyright (c) 2022, NVIDIA CORPORATION & AFFILIATES.  All rights reserved.\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+#     http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+\n+LOCALIZATION = \"eu\"  # Set to am for alternate formatting\ndiff --git a/nemo_text_processing/text_normalization/es/data/__init__.py b/nemo_text_processing/text_normalization/es/data/__init__.py\nnew file mode 100644\n--- /dev/null\n+++ b/nemo_text_processing/text_normalization/es/data/__init__.py\n@@ -0,0 +1,13 @@\n+# Copyright (c) 2022, NVIDIA CORPORATION & AFFILIATES.  All rights reserved.\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+#     http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\ndiff --git a/nemo_text_processing/text_normalization/es/data/dates/__init__.py b/nemo_text_processing/text_normalization/es/data/dates/__init__.py\nnew file mode 100644\n--- /dev/null\n+++ b/nemo_text_processing/text_normalization/es/data/dates/__init__.py\n@@ -0,0 +1,13 @@\n+# Copyright (c) 2022, NVIDIA CORPORATION & AFFILIATES.  All rights reserved.\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+#     http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\ndiff --git a/nemo_text_processing/text_normalization/es/data/electronic/__init__.py b/nemo_text_processing/text_normalization/es/data/electronic/__init__.py\nnew file mode 100644\n--- /dev/null\n+++ b/nemo_text_processing/text_normalization/es/data/electronic/__init__.py\n@@ -0,0 +1,13 @@\n+# Copyright (c) 2022, NVIDIA CORPORATION.  All rights reserved.\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+#     http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\ndiff --git a/nemo_text_processing/text_normalization/es/data/fractions/__init__.py b/nemo_text_processing/text_normalization/es/data/fractions/__init__.py\nnew file mode 100644\n--- /dev/null\n+++ b/nemo_text_processing/text_normalization/es/data/fractions/__init__.py\n@@ -0,0 +1,13 @@\n+# Copyright (c) 2022, NVIDIA CORPORATION & AFFILIATES.  All rights reserved.\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+#     http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\ndiff --git a/nemo_text_processing/text_normalization/es/data/measures/__init__.py b/nemo_text_processing/text_normalization/es/data/measures/__init__.py\nnew file mode 100644\n--- /dev/null\n+++ b/nemo_text_processing/text_normalization/es/data/measures/__init__.py\n@@ -0,0 +1,13 @@\n+# Copyright (c) 2022, NVIDIA CORPORATION & AFFILIATES.  All rights reserved.\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+#     http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\ndiff --git a/nemo_text_processing/text_normalization/es/data/money/__init__.py b/nemo_text_processing/text_normalization/es/data/money/__init__.py\nnew file mode 100644\n--- /dev/null\n+++ b/nemo_text_processing/text_normalization/es/data/money/__init__.py\n@@ -0,0 +1,13 @@\n+# Copyright (c) 2022, NVIDIA CORPORATION & AFFILIATES.  All rights reserved.\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+#     http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\ndiff --git a/nemo_text_processing/text_normalization/es/data/numbers/__init__.py b/nemo_text_processing/text_normalization/es/data/numbers/__init__.py\nnew file mode 100644\n--- /dev/null\n+++ b/nemo_text_processing/text_normalization/es/data/numbers/__init__.py\n@@ -0,0 +1,13 @@\n+# Copyright (c) 2022, NVIDIA CORPORATION & AFFILIATES.  All rights reserved.\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+#     http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\ndiff --git a/nemo_text_processing/text_normalization/es/data/ordinals/__init__.py b/nemo_text_processing/text_normalization/es/data/ordinals/__init__.py\nnew file mode 100644\n--- /dev/null\n+++ b/nemo_text_processing/text_normalization/es/data/ordinals/__init__.py\n@@ -0,0 +1,13 @@\n+# Copyright (c) 2022, NVIDIA CORPORATION & AFFILIATES.  All rights reserved.\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+#     http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\ndiff --git a/nemo_text_processing/text_normalization/es/data/roman/__init__.py b/nemo_text_processing/text_normalization/es/data/roman/__init__.py\nnew file mode 100644\n--- /dev/null\n+++ b/nemo_text_processing/text_normalization/es/data/roman/__init__.py\n@@ -0,0 +1,13 @@\n+# Copyright (c) 2022, NVIDIA CORPORATION.  All rights reserved.\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+#     http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\ndiff --git a/nemo_text_processing/text_normalization/es/data/time/__init__.py b/nemo_text_processing/text_normalization/es/data/time/__init__.py\nnew file mode 100644\n--- /dev/null\n+++ b/nemo_text_processing/text_normalization/es/data/time/__init__.py\n@@ -0,0 +1,13 @@\n+# Copyright (c) 2022, NVIDIA CORPORATION & AFFILIATES.  All rights reserved.\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+#     http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\ndiff --git a/nemo_text_processing/text_normalization/es/graph_utils.py b/nemo_text_processing/text_normalization/es/graph_utils.py\nnew file mode 100644\n--- /dev/null\n+++ b/nemo_text_processing/text_normalization/es/graph_utils.py\n@@ -0,0 +1,179 @@\n+# Copyright (c) 2022, NVIDIA CORPORATION & AFFILIATES.  All rights reserved.\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+#     http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+from nemo_text_processing.text_normalization.en.graph_utils import NEMO_SIGMA, NEMO_SPACE\n+from nemo_text_processing.text_normalization.es import LOCALIZATION\n+from nemo_text_processing.text_normalization.es.utils import get_abs_path, load_labels\n+\n+try:\n+    import pynini\n+    from pynini.lib import pynutil\n+\n+    digits = pynini.project(pynini.string_file(get_abs_path(\"data/numbers/digit.tsv\")), \"input\")\n+    tens = pynini.project(pynini.string_file(get_abs_path(\"data/numbers/ties.tsv\")), \"input\")\n+    teens = pynini.project(pynini.string_file(get_abs_path(\"data/numbers/teen.tsv\")), \"input\")\n+    twenties = pynini.project(pynini.string_file(get_abs_path(\"data/numbers/twenties.tsv\")), \"input\")\n+    hundreds = pynini.project(pynini.string_file(get_abs_path(\"data/numbers/hundreds.tsv\")), \"input\")\n+\n+    accents = pynini.string_map([(\"\u00e1\", \"a\"), (\"\u00e9\", \"e\"), (\"\u00ed\", \"i\"), (\"\u00f3\", \"o\"), (\"\u00fa\", \"u\")])\n+\n+    if LOCALIZATION == \"am\":  # Setting localization for central and northern america formatting\n+        cardinal_separator = pynini.string_map([\",\", NEMO_SPACE])\n+        decimal_separator = pynini.accep(\".\")\n+    else:\n+        cardinal_separator = pynini.string_map([\".\", NEMO_SPACE])\n+        decimal_separator = pynini.accep(\",\")\n+\n+    ones = pynini.union(\"un\", \"\u00fan\")\n+    fem_ones = pynini.union(pynini.cross(\"un\", \"una\"), pynini.cross(\"\u00fan\", \"una\"), pynini.cross(\"uno\", \"una\"))\n+    one_to_one_hundred = pynini.union(digits, tens, teens, twenties, tens + pynini.accep(\" y \") + digits)\n+    fem_hundreds = hundreds @ pynini.cdrewrite(pynini.cross(\"ientos\", \"ientas\"), \"\", \"\", NEMO_SIGMA)\n+\n+    PYNINI_AVAILABLE = True\n+\n+except (ModuleNotFoundError, ImportError):\n+    digits = None\n+    tens = None\n+    teens = None\n+    twenties = None\n+    hundreds = None\n+\n+    accents = None\n+\n+    cardinal_separator = None\n+    decimal_separator = None\n+\n+    ones = None\n+    fem_ones = None\n+    one_to_one_hundred = None\n+    fem_hundreds = None\n+\n+    PYNINI_AVAILABLE = False\n+\n+\n+def strip_accent(fst: 'pynini.FstLike') -> 'pynini.FstLike':\n+    \"\"\"\n+    Converts all accented vowels to non-accented equivalents\n+\n+    Args:\n+        fst: Any fst. Composes vowel conversion onto fst's output strings\n+    \"\"\"\n+    return fst @ pynini.cdrewrite(accents, \"\", \"\", NEMO_SIGMA)\n+\n+\n+def shift_cardinal_gender(fst: 'pynini.FstLike') -> 'pynini.FstLike':\n+    \"\"\"\n+    Applies gender conversion rules to a cardinal string. These include: rendering all masculine forms of \"uno\" (including apocopated forms) as \"una\" and\n+    Converting all gendered numbers in the hundreds series (200,300,400...) to feminine equivalent (e.g. \"doscientos\" -> \"doscientas\"). Converssion only applies\n+    to value place for <1000 and multiple of 1000. (e.g. \"doscientos mil doscientos\" -> \"doscientas mil doscientas\".) For place values greater than the thousands, there\n+    is no gender shift as the higher powers of ten (\"millones\", \"billones\") are masculine nouns and any conversion would be formally\n+    ungrammatical.\n+    e.g.\n+        \"doscientos\" -> \"doscientas\"\n+        \"doscientos mil\" -> \"doscientas mil\"\n+        \"doscientos millones\" -> \"doscientos millones\"\n+        \"doscientos mil millones\" -> \"doscientos mil millones\"\n+        \"doscientos millones doscientos mil doscientos\" -> \"doscientos millones doscientas mil doscientas\"\n+\n+    Args:\n+        fst: Any fst. Composes conversion onto fst's output strings\n+    \"\"\"\n+    before_mil = (\n+        NEMO_SPACE\n+        + (pynini.accep(\"mil\") | pynini.accep(\"mil\u00e9simo\"))\n+        + pynini.closure(NEMO_SPACE + hundreds, 0, 1)\n+        + pynini.closure(NEMO_SPACE + one_to_one_hundred, 0, 1)\n+        + pynini.union(pynini.accep(\"[EOS]\"), pynini.accep(\"\\\"\"), decimal_separator)\n+    )\n+    before_double_digits = pynini.closure(NEMO_SPACE + one_to_one_hundred, 0, 1) + pynini.union(\n+        pynini.accep(\"[EOS]\"), pynini.accep(\"\\\"\")\n+    )\n+\n+    fem_allign = pynini.cdrewrite(fem_hundreds, \"\", before_mil, NEMO_SIGMA)  # doscientas mil dosciento\n+    fem_allign @= pynini.cdrewrite(fem_hundreds, \"\", before_double_digits, NEMO_SIGMA)  # doscientas mil doscienta\n+    fem_allign @= pynini.cdrewrite(\n+        fem_ones, \"\", pynini.union(\"[EOS]\", \"\\\"\", decimal_separator), NEMO_SIGMA\n+    )  # If before a quote or EOS, we know it's the end of a string\n+\n+    return fst @ fem_allign\n+\n+\n+def shift_number_gender(fst: 'pynini.FstLike') -> 'pynini.FstLike':\n+    \"\"\"\n+    Performs gender conversion on all verbalized numbers in output. All values in the hundreds series (200,300,400) are changed to\n+    feminine gender (e.g. \"doscientos\" -> \"doscientas\") and all forms of \"uno\" (including apocopated forms) are converted to \"una\".\n+    This has no boundary restriction and will perform shift across all values in output string.\n+    e.g.\n+        \"doscientos\" -> \"doscientas\"\n+        \"doscientos millones\" -> \"doscientas millones\"\n+        \"doscientos millones doscientos\" -> \"doscientas millones doscientas\"\n+\n+    Args:\n+        fst: Any fst. Composes conversion onto fst's output strings\n+    \"\"\"\n+    fem_allign = pynini.cdrewrite(fem_hundreds, \"\", \"\", NEMO_SIGMA)\n+    fem_allign @= pynini.cdrewrite(\n+        fem_ones, \"\", pynini.union(NEMO_SPACE, pynini.accep(\"[EOS]\"), pynini.accep(\"\\\"\")), NEMO_SIGMA\n+    )  # If before a quote or EOS, we know it's the end of a string\n+\n+    return fst @ fem_allign\n+\n+\n+def strip_cardinal_apocope(fst: 'pynini.FstLike') -> 'pynini.FstLike':\n+    \"\"\"\n+    Reverts apocope on cardinal strings in line with formation rules. e.g. \"un\" -> \"uno\". Due to cardinal formation rules, this in effect only\n+    affects strings where the final value is a variation of \"un\".\n+    e.g.\n+        \"un\" -> \"uno\"\n+        \"veinti\u00fan\" -> \"veintiuno\"\n+\n+    Args:\n+        fst: Any fst. Composes conversion onto fst's output strings\n+    \"\"\"\n+    # Since cardinals use apocope by default for large values (e.g. \"mill\u00f3n\"), this only needs to act on the last instance of one\n+    strip = pynini.cross(\"un\", \"uno\") | pynini.cross(\"\u00fan\", \"uno\")\n+    strip = pynini.cdrewrite(strip, \"\", pynini.union(\"[EOS]\", \"\\\"\"), NEMO_SIGMA)\n+    return fst @ strip\n+\n+\n+def roman_to_int(fst: 'pynini.FstLike') -> 'pynini.FstLike':\n+    \"\"\"\n+    Alters given fst to convert Roman integers (lower and upper cased) into Arabic numerals. Valid for values up to 1000.\n+    e.g.\n+        \"V\" -> \"5\"\n+        \"i\" -> \"1\"\n+\n+    Args:\n+        fst: Any fst. Composes fst onto Roman conversion outputs.\n+    \"\"\"\n+\n+    def _load_roman(file: str):\n+        roman = load_labels(get_abs_path(file))\n+        roman_numerals = [(x, y) for x, y in roman] + [(x.upper(), y) for x, y in roman]\n+        return pynini.string_map(roman_numerals)\n+\n+    digit = _load_roman(\"data/roman/digit.tsv\")\n+    ties = _load_roman(\"data/roman/ties.tsv\")\n+    hundreds = _load_roman(\"data/roman/hundreds.tsv\")\n+\n+    graph = (\n+        digit\n+        | ties + (digit | pynutil.add_weight(pynutil.insert(\"0\"), 0.01))\n+        | (\n+            hundreds\n+            + (ties | pynutil.add_weight(pynutil.insert(\"0\"), 0.01))\n+            + (digit | pynutil.add_weight(pynutil.insert(\"0\"), 0.01))\n+        )\n+    ).optimize()\n+\n+    return graph @ fst\ndiff --git a/nemo_text_processing/text_normalization/es/taggers/__init__.py b/nemo_text_processing/text_normalization/es/taggers/__init__.py\nnew file mode 100644\n--- /dev/null\n+++ b/nemo_text_processing/text_normalization/es/taggers/__init__.py\n@@ -0,0 +1,13 @@\n+# Copyright (c) 2022, NVIDIA CORPORATION & AFFILIATES.  All rights reserved.\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+#     http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\ndiff --git a/nemo_text_processing/text_normalization/es/taggers/cardinal.py b/nemo_text_processing/text_normalization/es/taggers/cardinal.py\nnew file mode 100644\n--- /dev/null\n+++ b/nemo_text_processing/text_normalization/es/taggers/cardinal.py\n@@ -0,0 +1,190 @@\n+# Copyright (c) 2022, NVIDIA CORPORATION & AFFILIATES.  All rights reserved.\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+#     http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+from nemo_text_processing.text_normalization.en.graph_utils import (\n+    NEMO_ALPHA,\n+    NEMO_DIGIT,\n+    NEMO_SIGMA,\n+    NEMO_SPACE,\n+    NEMO_WHITE_SPACE,\n+    GraphFst,\n+    delete_space,\n+    insert_space,\n+)\n+from nemo_text_processing.text_normalization.es.graph_utils import cardinal_separator\n+from nemo_text_processing.text_normalization.es.utils import get_abs_path\n+\n+try:\n+    import pynini\n+    from pynini.lib import pynutil\n+\n+    zero = pynini.invert(pynini.string_file(get_abs_path(\"data/numbers/zero.tsv\")))\n+    digit = pynini.invert(pynini.string_file(get_abs_path(\"data/numbers/digit.tsv\")))\n+    teen = pynini.invert(pynini.string_file(get_abs_path(\"data/numbers/teen.tsv\")))\n+    ties = pynini.invert(pynini.string_file(get_abs_path(\"data/numbers/ties.tsv\")))\n+    twenties = pynini.invert(pynini.string_file(get_abs_path(\"data/numbers/twenties.tsv\")))\n+    hundreds = pynini.invert(pynini.string_file(get_abs_path(\"data/numbers/hundreds.tsv\")))\n+\n+    PYNINI_AVAILABLE = True\n+\n+except (ModuleNotFoundError, ImportError):\n+    zero = None\n+    digit = None\n+    teen = None\n+    ties = None\n+    twenties = None\n+    hundreds = None\n+\n+    PYNINI_AVAILABLE = False\n+\n+\n+def filter_punctuation(fst: 'pynini.FstLike') -> 'pynini.FstLike':\n+    \"\"\"\n+    Helper function for parsing number strings. Converts common cardinal strings (groups of three digits delineated by 'cardinal_separator' - see graph_utils)\n+    and converts to a string of digits:\n+        \"1 000\" -> \"1000\"\n+        \"1.000.000\" -> \"1000000\"\n+    Args:\n+        fst: Any pynini.FstLike object. Function composes fst onto string parser fst\n+\n+    Returns:\n+        fst: A pynini.FstLike object\n+    \"\"\"\n+    exactly_three_digits = NEMO_DIGIT ** 3  # for blocks of three\n+    up_to_three_digits = pynini.closure(NEMO_DIGIT, 1, 3)  # for start of string\n+\n+    cardinal_string = pynini.closure(\n+        NEMO_DIGIT, 1\n+    )  # For string w/o punctuation (used for page numbers, thousand series)\n+\n+    cardinal_string |= (\n+        up_to_three_digits\n+        + pynutil.delete(cardinal_separator)\n+        + pynini.closure(exactly_three_digits + pynutil.delete(cardinal_separator))\n+        + exactly_three_digits\n+    )\n+\n+    return cardinal_string @ fst\n+\n+\n+class CardinalFst(GraphFst):\n+    \"\"\"\n+    Finite state transducer for classifying cardinals, e.g.\n+        \"1000\" ->  cardinal { integer: \"mil\" }\n+        \"2.000.000\" -> cardinal { integer: \"dos millones\" }\n+\n+    Args:\n+        deterministic: if True will provide a single transduction option,\n+            for False multiple transduction are generated (used for audio-based normalization)\n+    \"\"\"\n+\n+    def __init__(self, deterministic: bool = True):\n+        super().__init__(name=\"cardinal\", kind=\"classify\", deterministic=deterministic)\n+\n+        # Any single digit\n+        graph_digit = digit\n+        digits_no_one = (NEMO_DIGIT - \"1\") @ graph_digit\n+\n+        # Any double digit\n+        graph_tens = teen\n+        graph_tens |= ties + (pynutil.delete('0') | (pynutil.insert(\" y \") + graph_digit))\n+        graph_tens |= twenties\n+\n+        self.tens = graph_tens.optimize()\n+\n+        self.two_digit_non_zero = pynini.union(\n+            graph_digit, graph_tens, (pynini.cross(\"0\", NEMO_SPACE) + graph_digit)\n+        ).optimize()\n+\n+        # Three digit strings\n+        graph_hundreds = hundreds + pynini.union(\n+            pynutil.delete(\"00\"), (insert_space + graph_tens), (pynini.cross(\"0\", NEMO_SPACE) + graph_digit)\n+        )\n+        graph_hundreds |= pynini.cross(\"100\", \"cien\")\n+        graph_hundreds |= (\n+            pynini.cross(\"1\", \"ciento\") + insert_space + pynini.union(graph_tens, pynutil.delete(\"0\") + graph_digit)\n+        )\n+\n+        self.hundreds = graph_hundreds.optimize()\n+\n+        # For all three digit strings with leading zeroes (graph appends '0's to manage place in string)\n+        graph_hundreds_component = pynini.union(graph_hundreds, pynutil.delete(\"0\") + graph_tens)\n+\n+        graph_hundreds_component_at_least_one_none_zero_digit = graph_hundreds_component | (\n+            pynutil.delete(\"00\") + graph_digit\n+        )\n+        graph_hundreds_component_at_least_one_none_zero_digit_no_one = graph_hundreds_component | (\n+            pynutil.delete(\"00\") + digits_no_one\n+        )\n+\n+        graph_thousands_component_at_least_one_none_zero_digit = pynini.union(\n+            pynutil.delete(\"000\") + graph_hundreds_component_at_least_one_none_zero_digit,\n+            graph_hundreds_component_at_least_one_none_zero_digit_no_one\n+            + pynutil.insert(\" mil\")\n+            + ((insert_space + graph_hundreds_component_at_least_one_none_zero_digit) | pynutil.delete(\"000\")),\n+            pynini.cross(\"001\", \"mil\")\n+            + ((insert_space + graph_hundreds_component_at_least_one_none_zero_digit) | pynutil.delete(\"000\")),\n+        )\n+\n+        graph_thousands_component_at_least_one_none_zero_digit_no_one = pynini.union(\n+            pynutil.delete(\"000\") + graph_hundreds_component_at_least_one_none_zero_digit_no_one,\n+            graph_hundreds_component_at_least_one_none_zero_digit_no_one\n+            + pynutil.insert(\" mil\")\n+            + ((insert_space + graph_hundreds_component_at_least_one_none_zero_digit) | pynutil.delete(\"000\")),\n+            pynini.cross(\"001\", \"mil\")\n+            + ((insert_space + graph_hundreds_component_at_least_one_none_zero_digit) | pynutil.delete(\"000\")),\n+        )\n+\n+        graph_million = pynutil.add_weight(pynini.cross(\"000001\", \"un mill\u00f3n\"), -0.001)\n+        graph_million |= graph_thousands_component_at_least_one_none_zero_digit_no_one + pynutil.insert(\" millones\")\n+        graph_million |= pynutil.delete(\"000000\")\n+        graph_million += insert_space\n+\n+        graph_billion = pynutil.add_weight(pynini.cross(\"000001\", \"un bill\u00f3n\"), -0.001)\n+        graph_billion |= graph_thousands_component_at_least_one_none_zero_digit_no_one + pynutil.insert(\" billones\")\n+        graph_billion |= pynutil.delete(\"000000\")\n+        graph_billion += insert_space\n+\n+        graph_trillion = pynutil.add_weight(pynini.cross(\"000001\", \"un trill\u00f3n\"), -0.001)\n+        graph_trillion |= graph_thousands_component_at_least_one_none_zero_digit_no_one + pynutil.insert(\" trillones\")\n+        graph_trillion |= pynutil.delete(\"000000\")\n+        graph_trillion += insert_space\n+\n+        graph = (\n+            graph_trillion\n+            + graph_billion\n+            + graph_million\n+            + (graph_thousands_component_at_least_one_none_zero_digit | pynutil.delete(\"000000\"))\n+        )\n+\n+        self.graph = (\n+            ((NEMO_DIGIT - \"0\") + pynini.closure(NEMO_DIGIT, 0))\n+            @ pynini.cdrewrite(pynini.closure(pynutil.insert(\"0\")), \"[BOS]\", \"\", NEMO_SIGMA)\n+            @ NEMO_DIGIT ** 24\n+            @ graph\n+            @ pynini.cdrewrite(delete_space, \"[BOS]\", \"\", NEMO_SIGMA)\n+            @ pynini.cdrewrite(delete_space, \"\", \"[EOS]\", NEMO_SIGMA)\n+            @ pynini.cdrewrite(\n+                pynini.cross(pynini.closure(NEMO_WHITE_SPACE, 2), NEMO_SPACE), NEMO_ALPHA, NEMO_ALPHA, NEMO_SIGMA\n+            )\n+        )\n+        self.graph |= zero\n+\n+        self.graph = filter_punctuation(self.graph).optimize()\n+\n+        optional_minus_graph = pynini.closure(pynutil.insert(\"negative: \") + pynini.cross(\"-\", \"\\\"true\\\" \"), 0, 1)\n+\n+        final_graph = optional_minus_graph + pynutil.insert(\"integer: \\\"\") + self.graph + pynutil.insert(\"\\\"\")\n+\n+        final_graph = self.add_tokens(final_graph)\n+        self.fst = final_graph.optimize()\ndiff --git a/nemo_text_processing/text_normalization/es/taggers/date.py b/nemo_text_processing/text_normalization/es/taggers/date.py\nnew file mode 100644\n--- /dev/null\n+++ b/nemo_text_processing/text_normalization/es/taggers/date.py\n@@ -0,0 +1,107 @@\n+# Copyright (c) 2022, NVIDIA CORPORATION & AFFILIATES.  All rights reserved.\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+#     http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+from nemo_text_processing.text_normalization.en.graph_utils import NEMO_DIGIT, NEMO_SPACE, GraphFst, delete_extra_space\n+from nemo_text_processing.text_normalization.es.utils import get_abs_path\n+\n+try:\n+    import pynini\n+    from pynini.lib import pynutil\n+\n+    articles = pynini.union(\"de\", \"del\", \"el\", \"del a\u00f1o\")\n+    delete_leading_zero = (pynutil.delete(\"0\") | (NEMO_DIGIT - \"0\")) + NEMO_DIGIT\n+    month_numbers = pynini.string_file(get_abs_path(\"data/dates/months.tsv\"))\n+\n+    PYNINI_AVAILABLE = True\n+\n+except (ModuleNotFoundError, ImportError):\n+    articles = None\n+    delete_leading_zero = None\n+    month_numbers = None\n+\n+    PYNINI_AVAILABLE = False\n+\n+\n+class DateFst(GraphFst):\n+    \"\"\"\n+    Finite state transducer for classifying date, e.g.\n+        \"01.04.2010\" -> date { day: \"un\" month: \"enero\" year: \"dos mil diez\" preserve_order: true }\n+        \"marzo 4 2000\" -> date { month: \"marzo\" day: \"cuatro\" year: \"dos mil\" }\n+        \"1990-20-01\" -> date { year: \"mil novecientos noventa\" day: \"veinte\" month: \"enero\" }\n+\n+    Args:\n+        cardinal: cardinal GraphFst\n+        deterministic: if True will provide a single transduction option,\n+            for False multiple transduction are generated (used for audio-based normalization)\n+    \"\"\"\n+\n+    def __init__(self, cardinal: GraphFst, deterministic: bool):\n+        super().__init__(name=\"date\", kind=\"classify\", deterministic=deterministic)\n+\n+        number_to_month = month_numbers.optimize()\n+        month_graph = pynini.project(number_to_month, \"output\")\n+\n+        numbers = cardinal.graph\n+        optional_leading_zero = delete_leading_zero | NEMO_DIGIT\n+\n+        # 01, 31, 1\n+        digit_day = optional_leading_zero @ pynini.union(*[str(x) for x in range(1, 32)]) @ numbers\n+        day = (pynutil.insert(\"day: \\\"\") + digit_day + pynutil.insert(\"\\\"\")).optimize()\n+\n+        digit_month = optional_leading_zero @ pynini.union(*[str(x) for x in range(1, 13)])\n+        number_to_month = digit_month @ number_to_month\n+\n+        month_name = (pynutil.insert(\"month: \\\"\") + month_graph + pynutil.insert(\"\\\"\")).optimize()\n+        month_number = (pynutil.insert(\"month: \\\"\") + number_to_month + pynutil.insert(\"\\\"\")).optimize()\n+\n+        # prefer cardinal over year\n+        year = (NEMO_DIGIT - \"0\") + pynini.closure(NEMO_DIGIT, 1, 3)  # 90, 990, 1990\n+        year @= numbers\n+        self.year = year\n+\n+        year_only = pynutil.insert(\"year: \\\"\") + year + pynutil.insert(\"\\\"\")\n+        year_with_articles = (\n+            pynutil.insert(\"year: \\\"\") + pynini.closure(articles + NEMO_SPACE, 0, 1) + year + pynutil.insert(\"\\\"\")\n+        )\n+\n+        graph_dmy = (\n+            day\n+            + pynini.closure(pynutil.delete(\" de\"))\n+            + NEMO_SPACE\n+            + month_name\n+            + pynini.closure(NEMO_SPACE + year_with_articles, 0, 1)\n+        )\n+\n+        graph_mdy = (  # English influences on language\n+            month_name + delete_extra_space + day + pynini.closure(NEMO_SPACE + year_with_articles, 0, 1)\n+        )\n+\n+        separators = [\".\", \"-\", \"/\"]\n+        for sep in separators:\n+            year_optional = pynini.closure(pynini.cross(sep, NEMO_SPACE) + year_only, 0, 1)\n+            new_graph = day + pynini.cross(sep, NEMO_SPACE) + month_number + year_optional\n+            graph_dmy |= new_graph\n+            if not deterministic:\n+                new_graph = month_number + pynini.cross(sep, NEMO_SPACE) + day + year_optional\n+                graph_mdy |= new_graph\n+\n+        dash = \"-\"\n+        day_optional = pynini.closure(pynini.cross(dash, NEMO_SPACE) + day, 0, 1)\n+        graph_ymd = NEMO_DIGIT ** 4 @ year_only + pynini.cross(dash, NEMO_SPACE) + month_number + day_optional\n+\n+        final_graph = graph_dmy + pynutil.insert(\" preserve_order: true\")\n+        final_graph |= graph_ymd\n+        final_graph |= graph_mdy\n+\n+        self.final_graph = final_graph.optimize()\n+        self.fst = self.add_tokens(self.final_graph).optimize()\ndiff --git a/nemo_text_processing/text_normalization/es/taggers/decimals.py b/nemo_text_processing/text_normalization/es/taggers/decimals.py\nnew file mode 100644\n--- /dev/null\n+++ b/nemo_text_processing/text_normalization/es/taggers/decimals.py\n@@ -0,0 +1,138 @@\n+# Copyright (c) 2022, NVIDIA CORPORATION & AFFILIATES.  All rights reserved.\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+#     http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+from nemo_text_processing.text_normalization.en.graph_utils import (\n+    NEMO_DIGIT,\n+    NEMO_SIGMA,\n+    NEMO_SPACE,\n+    GraphFst,\n+    delete_space,\n+    insert_space,\n+)\n+from nemo_text_processing.text_normalization.es.graph_utils import (\n+    cardinal_separator,\n+    decimal_separator,\n+    strip_cardinal_apocope,\n+)\n+from nemo_text_processing.text_normalization.es.utils import get_abs_path\n+\n+try:\n+    import pynini\n+    from pynini.lib import pynutil\n+\n+    quantities = pynini.string_file(get_abs_path(\"data/numbers/quantities.tsv\"))\n+    digit = pynini.invert(pynini.string_file(get_abs_path(\"data/numbers/digit.tsv\")))\n+    zero = pynini.invert(pynini.string_file(get_abs_path(\"data/numbers/zero.tsv\")))\n+\n+    PYNINI_AVAILABLE = True\n+\n+except (ModuleNotFoundError, ImportError):\n+    quantities = None\n+    digit = None\n+    zero = None\n+\n+    PYNINI_AVAILABLE = False\n+\n+\n+def get_quantity(decimal_graph: 'pynini.FstLike', cardinal_graph: 'pynini.FstLike') -> 'pynini.FstLike':\n+    \"\"\"\n+    Returns FST that transforms either a cardinal or decimal followed by a quantity into a numeral,\n+    e.g. 2 millones -> integer_part: \"dos\" quantity: \"millones\"\n+    e.g. 2,4 millones -> integer_part: \"dos\" fractional_part: \"quatro\" quantity: \"millones\"\n+    e.g. 2,400 millones -> integer_part: \"dos mil cuatrocientos\" fractional_part: \"quatro\" quantity: \"millones\"\n+\n+    Args:\n+        decimal_graph: DecimalFST\n+        cardinal_graph: CardinalFST\n+    \"\"\"\n+    numbers = pynini.closure(NEMO_DIGIT, 1, 6) @ cardinal_graph\n+    numbers = pynini.cdrewrite(pynutil.delete(cardinal_separator), \"\", \"\", NEMO_SIGMA) @ numbers\n+\n+    res = (\n+        pynutil.insert(\"integer_part: \\\"\")\n+        + numbers  # The cardinal we're passing only produces 'un' for one, so gender agreement is safe (all quantities are masculine). Limit to 10^6 power.\n+        + pynutil.insert(\"\\\"\")\n+        + NEMO_SPACE\n+        + pynutil.insert(\"quantity: \\\"\")\n+        + quantities\n+        + pynutil.insert(\"\\\"\")\n+    )\n+    res |= decimal_graph + NEMO_SPACE + pynutil.insert(\"quantity: \\\"\") + quantities + pynutil.insert(\"\\\"\")\n+    return res\n+\n+\n+class DecimalFst(GraphFst):\n+    \"\"\"\n+    Finite state transducer for classifying decimal, e.g.\n+        -11,4006 billones -> decimal { negative: \"true\" integer_part: \"once\"  fractional_part: \"cuatro cero cero seis\" quantity: \"billones\" preserve_order: true }\n+        1 bill\u00f3n -> decimal { integer_part: \"un\" quantity: \"bill\u00f3n\" preserve_order: true }\n+    Args:\n+        cardinal: CardinalFst\n+        deterministic: if True will provide a single transduction option,\n+            for False multiple transduction are generated (used for audio-based normalization)\n+    \"\"\"\n+\n+    def __init__(self, cardinal: GraphFst, deterministic: bool = True):\n+        super().__init__(name=\"decimal\", kind=\"classify\", deterministic=deterministic)\n+        graph_digit = digit | zero\n+\n+        if not deterministic:\n+            graph = pynini.union(graph_digit, cardinal.hundreds, cardinal.tens)\n+            graph += pynini.closure(insert_space + graph)\n+\n+        else:\n+            # General pattern seems to be 1-3 digits: map as cardinal, default to digits otherwise \\\n+            graph = pynini.union(\n+                graph_digit,\n+                cardinal.tens,\n+                cardinal.hundreds,\n+                graph_digit + pynini.closure(insert_space + graph_digit, 3),\n+                zero\n+                + pynini.closure(insert_space + zero)\n+                + pynini.closure(insert_space + graph_digit),  # For cases such as \"1,010\"\n+            )\n+\n+        # Need to strip apocope everywhere BUT end of string\n+        reverse_apocope = pynini.string_map([(\"un\", \"uno\"), (\"\u00fan\", \"uno\")])\n+        apply_reverse_apocope = pynini.cdrewrite(reverse_apocope, \"\", NEMO_SPACE, NEMO_SIGMA)\n+        graph @= apply_reverse_apocope\n+\n+        # Technically decimals should be space delineated groups of three, e.g. (1,333 333). This removes any possible spaces\n+        strip_formatting = pynini.cdrewrite(delete_space, \"\", \"\", NEMO_SIGMA)\n+        graph = strip_formatting @ graph\n+\n+        self.graph = graph.optimize()\n+\n+        graph_separator = pynutil.delete(decimal_separator)\n+        optional_graph_negative = pynini.closure(pynutil.insert(\"negative: \") + pynini.cross(\"-\", \"\\\"true\\\" \"), 0, 1)\n+\n+        self.graph_fractional = pynutil.insert(\"fractional_part: \\\"\") + self.graph + pynutil.insert(\"\\\"\")\n+\n+        # Integer graph maintains apocope except for ones place\n+        graph_integer = (\n+            strip_cardinal_apocope(cardinal.graph)\n+            if deterministic\n+            else pynini.union(cardinal.graph, strip_cardinal_apocope(cardinal.graph))\n+        )  # Gives us forms w/ and w/o apocope\n+        self.graph_integer = pynutil.insert(\"integer_part: \\\"\") + graph_integer + pynutil.insert(\"\\\"\")\n+        final_graph_wo_sign = self.graph_integer + graph_separator + insert_space + self.graph_fractional\n+\n+        self.final_graph_wo_negative = (\n+            final_graph_wo_sign | get_quantity(final_graph_wo_sign, cardinal.graph).optimize()\n+        )\n+        final_graph = optional_graph_negative + self.final_graph_wo_negative\n+\n+        final_graph += pynutil.insert(\" preserve_order: true\")\n+        final_graph = self.add_tokens(final_graph)\n+\n+        self.fst = final_graph.optimize()\ndiff --git a/nemo_text_processing/text_normalization/es/taggers/electronic.py b/nemo_text_processing/text_normalization/es/taggers/electronic.py\nnew file mode 100644\n--- /dev/null\n+++ b/nemo_text_processing/text_normalization/es/taggers/electronic.py\n@@ -0,0 +1,84 @@\n+# Copyright (c) 2022, NVIDIA CORPORATION & AFFILIATES.  All rights reserved.\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+#     http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+from nemo_text_processing.text_normalization.en.graph_utils import NEMO_ALPHA, NEMO_DIGIT, GraphFst, insert_space\n+from nemo_text_processing.text_normalization.es.utils import get_abs_path, load_labels\n+\n+try:\n+    import pynini\n+    from pynini.lib import pynutil\n+\n+    common_domains = [x[0] for x in load_labels(get_abs_path(\"data/electronic/domain.tsv\"))]\n+    symbols = [x[0] for x in load_labels(get_abs_path(\"data/electronic/symbols.tsv\"))]\n+\n+    PYNINI_AVAILABLE = True\n+\n+except (ModuleNotFoundError, ImportError):\n+    common_domains = None\n+    symbols = None\n+\n+    PYNINI_AVAILABLE = False\n+\n+\n+class ElectronicFst(GraphFst):\n+    \"\"\"\n+    Finite state transducer for classifying electronic: email addresses\n+        e.g. \"abc@hotmail.com\" -> electronic { username: \"abc\" domain: \"hotmail.com\" preserve_order: true }\n+        e.g. \"www.abc.com/123\" -> electronic { protocol: \"www.\" domain: \"abc.com/123\" preserve_order: true }\n+\n+    Args:\n+        deterministic: if True will provide a single transduction option,\n+            for False multiple transduction are generated (used for audio-based normalization)\n+    \"\"\"\n+\n+    def __init__(self, deterministic: bool = True):\n+        super().__init__(name=\"electronic\", kind=\"classify\", deterministic=deterministic)\n+\n+        dot = pynini.accep(\".\")\n+        accepted_common_domains = pynini.union(*common_domains)\n+        accepted_symbols = pynini.union(*symbols) - dot\n+        accepted_characters = pynini.closure(NEMO_ALPHA | NEMO_DIGIT | accepted_symbols)\n+        acceepted_characters_with_dot = pynini.closure(NEMO_ALPHA | NEMO_DIGIT | accepted_symbols | dot)\n+\n+        # email\n+        username = (\n+            pynutil.insert(\"username: \\\"\")\n+            + acceepted_characters_with_dot\n+            + pynutil.insert(\"\\\"\")\n+            + pynini.cross('@', ' ')\n+        )\n+        domain_graph = accepted_characters + dot + accepted_characters\n+        domain_graph = pynutil.insert(\"domain: \\\"\") + domain_graph + pynutil.insert(\"\\\"\")\n+        domain_common_graph = (\n+            pynutil.insert(\"domain: \\\"\")\n+            + accepted_characters\n+            + accepted_common_domains\n+            + pynini.closure((accepted_symbols | dot) + pynini.closure(accepted_characters, 1), 0, 1)\n+            + pynutil.insert(\"\\\"\")\n+        )\n+        graph = (username + domain_graph) | domain_common_graph\n+\n+        # url\n+        protocol_start = pynini.accep(\"https://\") | pynini.accep(\"http://\")\n+        protocol_end = (\n+            pynini.accep(\"www.\")\n+            if deterministic\n+            else pynini.accep(\"www.\") | pynini.cross(\"www.\", \"doble ve doble ve doble ve.\")\n+        )\n+        protocol = protocol_start | protocol_end | (protocol_start + protocol_end)\n+        protocol = pynutil.insert(\"protocol: \\\"\") + protocol + pynutil.insert(\"\\\"\")\n+        graph |= protocol + insert_space + (domain_graph | domain_common_graph)\n+        self.graph = graph\n+\n+        final_graph = self.add_tokens(self.graph + pynutil.insert(\" preserve_order: true\"))\n+        self.fst = final_graph.optimize()\ndiff --git a/nemo_text_processing/text_normalization/es/taggers/fraction.py b/nemo_text_processing/text_normalization/es/taggers/fraction.py\nnew file mode 100644\n--- /dev/null\n+++ b/nemo_text_processing/text_normalization/es/taggers/fraction.py\n@@ -0,0 +1,124 @@\n+# Copyright (c) 2022, NVIDIA CORPORATION & AFFILIATES.  All rights reserved.\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+#     http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+from nemo_text_processing.text_normalization.en.graph_utils import (\n+    NEMO_CHAR,\n+    NEMO_DIGIT,\n+    NEMO_SIGMA,\n+    NEMO_SPACE,\n+    GraphFst,\n+)\n+from nemo_text_processing.text_normalization.es.utils import get_abs_path\n+\n+try:\n+    import pynini\n+    from pynini.lib import pynutil\n+\n+    ordinal_exceptions = pynini.string_file(get_abs_path(\"data/fractions/ordinal_exceptions.tsv\"))\n+    higher_powers_of_ten = pynini.string_file(get_abs_path(\"data/fractions/powers_of_ten.tsv\"))\n+\n+    PYNINI_AVAILABLE = True\n+\n+except (ModuleNotFoundError, ImportError):\n+    ordinal_exceptions = None\n+    higher_powers_of_ten = None\n+\n+    PYNINI_AVAILABLE = False\n+\n+\n+class FractionFst(GraphFst):\n+    \"\"\"\n+    Finite state transducer for classifying fraction\n+    \"23 4/5\" ->\n+    tokens { fraction { integer: \"veintitr\u00e9s\" numerator: \"cuatro\" denominator: \"quinto\" mophosyntactic_features: \"ordinal\" } }\n+\n+    Args:\n+        cardinal: CardinalFst\n+        ordinal: OrdinalFst\n+        deterministic: if True will provide a single transduction option,\n+            for False multiple transduction are generated (used for audio-based normalization)\n+    \"\"\"\n+\n+    def __init__(self, cardinal: GraphFst, ordinal: GraphFst, deterministic: bool = True):\n+        super().__init__(name=\"fraction\", kind=\"classify\", deterministic=deterministic)\n+        cardinal_graph = cardinal.graph\n+        ordinal_graph = ordinal.graph\n+\n+        # 2-10 are all ordinals\n+        three_to_ten = pynini.string_map([\"2\", \"3\", \"4\", \"5\", \"6\", \"7\", \"8\", \"9\", \"10\",])\n+        block_three_to_ten = pynutil.delete(three_to_ten)  # To block cardinal productions\n+        if not deterministic:  # Multiples of tens are sometimes rendered as ordinals\n+            three_to_ten |= pynini.string_map([\"20\", \"30\", \"40\", \"50\", \"60\", \"70\", \"80\", \"90\",])\n+        graph_three_to_ten = three_to_ten @ ordinal_graph\n+        graph_three_to_ten @= pynini.cdrewrite(ordinal_exceptions, \"\", \"\", NEMO_SIGMA)\n+\n+        # Higher powers of tens (and multiples) are converted to ordinals.\n+        hundreds = pynini.string_map([\"100\", \"200\", \"300\", \"400\", \"500\", \"600\", \"700\", \"800\", \"900\",])\n+        graph_hundreds = hundreds @ ordinal_graph\n+\n+        multiples_of_thousand = ordinal.multiples_of_thousand  # So we can have X mil\u00e9simos\n+\n+        graph_higher_powers_of_ten = (\n+            pynini.closure(ordinal.one_to_one_thousand + NEMO_SPACE, 0, 1)\n+            + pynini.closure(\"mil \", 0, 1)\n+            + pynini.closure(ordinal.one_to_one_thousand + NEMO_SPACE, 0, 1)\n+        )  # x millones / x mil millones / x mil z millones\n+        graph_higher_powers_of_ten += higher_powers_of_ten\n+        graph_higher_powers_of_ten = cardinal_graph @ graph_higher_powers_of_ten\n+        graph_higher_powers_of_ten @= pynini.cdrewrite(\n+            pynutil.delete(\"un \"), pynini.accep(\"[BOS]\"), pynini.project(higher_powers_of_ten, \"output\"), NEMO_SIGMA\n+        )  # we drop 'un' from these ordinals (millionths, not one-millionths)\n+\n+        graph_higher_powers_of_ten = multiples_of_thousand | graph_hundreds | graph_higher_powers_of_ten\n+        block_higher_powers_of_ten = pynutil.delete(\n+            pynini.project(graph_higher_powers_of_ten, \"input\")\n+        )  # For cardinal graph\n+\n+        graph_fractions_ordinals = graph_higher_powers_of_ten | graph_three_to_ten\n+        graph_fractions_ordinals += pynutil.insert(\n+            \"\\\" morphosyntactic_features: \\\"ordinal\\\"\"\n+        )  # We note the root for processing later\n+\n+        # Blocking the digits and hundreds from Cardinal graph\n+        graph_fractions_cardinals = pynini.cdrewrite(\n+            block_three_to_ten | block_higher_powers_of_ten, pynini.accep(\"[BOS]\"), pynini.accep(\"[EOS]\"), NEMO_SIGMA\n+        )\n+        graph_fractions_cardinals @= NEMO_CHAR.plus @ pynini.cdrewrite(\n+            pynutil.delete(\"0\"), pynini.accep(\"[BOS]\"), pynini.accep(\"[EOS]\"), NEMO_SIGMA\n+        )  # Empty characters become '0' for NEMO_CHAR fst, so ned to block\n+        graph_fractions_cardinals @= cardinal_graph\n+        graph_fractions_cardinals += pynutil.insert(\n+            \"\\\" morphosyntactic_features: \\\"add_root\\\"\"\n+        )  # blocking these entries to reduce erroneous possibilities in debugging\n+\n+        if deterministic:\n+            graph_fractions_cardinals = (\n+                pynini.closure(NEMO_DIGIT, 1, 2) @ graph_fractions_cardinals\n+            )  # Past hundreds the conventional scheme can be hard to read. For determinism we stop here\n+\n+        graph_denominator = pynini.union(\n+            graph_fractions_ordinals,\n+            graph_fractions_cardinals,\n+            pynutil.add_weight(cardinal_graph + pynutil.insert(\"\\\"\"), 0.001),\n+        )  # Last form is simply recording the cardinal. Weighting so last resort\n+\n+        integer = pynutil.insert(\"integer_part: \\\"\") + cardinal_graph + pynutil.insert(\"\\\"\") + NEMO_SPACE\n+        numerator = (\n+            pynutil.insert(\"numerator: \\\"\") + cardinal_graph + (pynini.cross(\"/\", \"\\\" \") | pynini.cross(\" / \", \"\\\" \"))\n+        )\n+        denominator = pynutil.insert(\"denominator: \\\"\") + graph_denominator\n+\n+        self.graph = pynini.closure(integer, 0, 1) + numerator + denominator\n+\n+        final_graph = self.add_tokens(self.graph)\n+        self.fst = final_graph.optimize()\ndiff --git a/nemo_text_processing/text_normalization/es/taggers/measure.py b/nemo_text_processing/text_normalization/es/taggers/measure.py\nnew file mode 100644\n--- /dev/null\n+++ b/nemo_text_processing/text_normalization/es/taggers/measure.py\n@@ -0,0 +1,184 @@\n+# Copyright (c) 2022, NVIDIA CORPORATION & AFFILIATES.  All rights reserved.\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+#     http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+from nemo_text_processing.text_normalization.en.graph_utils import (\n+    NEMO_ALPHA,\n+    NEMO_DIGIT,\n+    NEMO_NON_BREAKING_SPACE,\n+    NEMO_SPACE,\n+    GraphFst,\n+    convert_space,\n+    delete_space,\n+    insert_space,\n+)\n+from nemo_text_processing.text_normalization.es.graph_utils import strip_cardinal_apocope\n+from nemo_text_processing.text_normalization.es.utils import get_abs_path\n+\n+try:\n+    import pynini\n+    from pynini.lib import pynutil\n+\n+    unit = pynini.string_file(get_abs_path(\"data/measures/measurements.tsv\"))\n+    unit_plural_fem = pynini.string_file(get_abs_path(\"data/measures/measurements_plural_fem.tsv\"))\n+    unit_plural_masc = pynini.string_file(get_abs_path(\"data/measures/measurements_plural_masc.tsv\"))\n+\n+    PYNINI_AVAILABLE = True\n+\n+except (ModuleNotFoundError, ImportError):\n+    unit = None\n+    unit_plural_fem = None\n+    unit_plural_masc = None\n+\n+    PYNINI_AVAILABLE = False\n+\n+\n+class MeasureFst(GraphFst):\n+    \"\"\"\n+    Finite state transducer for classifying measure,  e.g.\n+        \"2,4 g\" -> measure { cardinal { integer_part: \"dos\" fractional_part: \"cuatro\" units: \"gramos\" preserve_order: true } }\n+        \"1 g\" -> measure { cardinal { integer: \"un\" units: \"gramo\" preserve_order: true } }\n+        \"1 mill\u00f3n g\" -> measure { cardinal { integer: \"un quantity: \"mill\u00f3n\" units: \"gramos\" preserve_order: true } }\n+        e.g. \"a-8\" \u2014> \"a ocho\"\n+        e.g. \"1,2-a\" \u2014> \"uno coma dos a\"\n+        This class also converts words containing numbers and letters\n+        e.g. \"a-8\" \u2014> \"a ocho\"\n+        e.g. \"1,2-a\" \u2014> \"uno coma dos a\"\n+\n+\n+    Args:\n+        cardinal: CardinalFst\n+        decimal: DecimalFst\n+        fraction: FractionFst\n+        deterministic: if True will provide a single transduction option,\n+            for False multiple transduction are generated (used for audio-based normalization)\n+    \"\"\"\n+\n+    def __init__(self, cardinal: GraphFst, decimal: GraphFst, fraction: GraphFst, deterministic: bool = True):\n+        super().__init__(name=\"measure\", kind=\"classify\", deterministic=deterministic)\n+        cardinal_graph = cardinal.graph\n+\n+        unit_singular = unit\n+        unit_plural = unit_singular @ (unit_plural_fem | unit_plural_masc)\n+\n+        graph_unit_singular = convert_space(unit_singular)\n+        graph_unit_plural = convert_space(unit_plural)\n+\n+        optional_graph_negative = pynini.closure(\"-\", 0, 1)\n+\n+        graph_unit_denominator = (\n+            pynini.cross(\"/\", \"por\") + pynutil.insert(NEMO_NON_BREAKING_SPACE) + graph_unit_singular\n+        )\n+\n+        optional_unit_denominator = pynini.closure(\n+            pynutil.insert(NEMO_NON_BREAKING_SPACE) + graph_unit_denominator, 0, 1,\n+        )\n+\n+        unit_plural = (\n+            pynutil.insert(\"units: \\\"\")\n+            + ((graph_unit_plural + optional_unit_denominator) | graph_unit_denominator)\n+            + pynutil.insert(\"\\\"\")\n+        )\n+\n+        unit_singular_graph = (\n+            pynutil.insert(\"units: \\\"\")\n+            + ((graph_unit_singular + optional_unit_denominator) | graph_unit_denominator)\n+            + pynutil.insert(\"\\\"\")\n+        )\n+\n+        subgraph_decimal = decimal.fst + insert_space + pynini.closure(NEMO_SPACE, 0, 1) + unit_plural\n+\n+        subgraph_cardinal = (\n+            (optional_graph_negative + (pynini.closure(NEMO_DIGIT) - \"1\")) @ cardinal.fst\n+            + insert_space\n+            + pynini.closure(delete_space, 0, 1)\n+            + unit_plural\n+        )\n+\n+        subgraph_cardinal |= (\n+            (optional_graph_negative + pynini.accep(\"1\")) @ cardinal.fst\n+            + insert_space\n+            + pynini.closure(delete_space, 0, 1)\n+            + unit_singular_graph\n+        )\n+\n+        subgraph_fraction = fraction.fst + insert_space + pynini.closure(delete_space, 0, 1) + unit_plural\n+\n+        decimal_times = (\n+            pynutil.insert(\"decimal { \")\n+            + decimal.final_graph_wo_negative\n+            + pynutil.insert(\" } units: \\\"\")\n+            + pynini.union('x', 'X')\n+            + pynutil.insert(\"\\\"\")\n+        )\n+\n+        cardinal_times = (\n+            pynutil.insert(\"cardinal { integer: \\\"\")\n+            + strip_cardinal_apocope(cardinal_graph)\n+            + pynutil.insert(\"\\\" } units: \\\"\")\n+            + pynini.union('x', 'X')\n+            + pynutil.insert(\"\\\"\")\n+        )\n+\n+        cardinal_dash_alpha = (\n+            pynutil.insert(\"cardinal { integer: \\\"\")\n+            + strip_cardinal_apocope(cardinal_graph)\n+            + pynutil.delete('-')\n+            + pynutil.insert(\"\\\" } units: \\\"\")\n+            + pynini.closure(NEMO_ALPHA, 1)\n+            + pynutil.insert(\"\\\"\")\n+        )\n+\n+        decimal_dash_alpha = (\n+            pynutil.insert(\"decimal { \")\n+            + decimal.final_graph_wo_negative\n+            + pynutil.delete('-')\n+            + pynutil.insert(\" } units: \\\"\")\n+            + pynini.closure(NEMO_ALPHA, 1)\n+            + pynutil.insert(\"\\\"\")\n+        )\n+\n+        alpha_dash_cardinal = (\n+            pynutil.insert(\"units: \\\"\")\n+            + pynini.closure(NEMO_ALPHA, 1)\n+            + pynutil.delete('-')\n+            + pynutil.insert(\"\\\"\")\n+            + pynutil.insert(\" cardinal { integer: \\\"\")\n+            + cardinal_graph\n+            + pynutil.insert(\"\\\" } preserve_order: true\")\n+        )\n+\n+        alpha_dash_decimal = (\n+            pynutil.insert(\"units: \\\"\")\n+            + pynini.closure(NEMO_ALPHA, 1)\n+            + pynutil.delete('-')\n+            + pynutil.insert(\"\\\"\")\n+            + pynutil.insert(\" decimal { \")\n+            + decimal.final_graph_wo_negative\n+            + pynutil.insert(\" } preserve_order: true\")\n+        )\n+\n+        final_graph = (\n+            subgraph_decimal\n+            | subgraph_cardinal\n+            | subgraph_fraction\n+            | cardinal_dash_alpha\n+            | alpha_dash_cardinal\n+            | decimal_dash_alpha\n+            | decimal_times\n+            | cardinal_times\n+            | alpha_dash_decimal\n+        )\n+        final_graph += pynutil.insert(\" preserve_order: true\")\n+        final_graph = self.add_tokens(final_graph)\n+\n+        self.fst = final_graph.optimize()\ndiff --git a/nemo_text_processing/text_normalization/es/taggers/money.py b/nemo_text_processing/text_normalization/es/taggers/money.py\nnew file mode 100644\n--- /dev/null\n+++ b/nemo_text_processing/text_normalization/es/taggers/money.py\n@@ -0,0 +1,194 @@\n+# Copyright (c) 2022, NVIDIA CORPORATION & AFFILIATES.  All rights reserved.\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+#     http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+from nemo_text_processing.text_normalization.en.graph_utils import (\n+    NEMO_ALPHA,\n+    NEMO_DIGIT,\n+    NEMO_SIGMA,\n+    NEMO_SPACE,\n+    GraphFst,\n+    delete_space,\n+    insert_space,\n+)\n+from nemo_text_processing.text_normalization.es.graph_utils import decimal_separator\n+from nemo_text_processing.text_normalization.es.utils import get_abs_path, load_labels\n+\n+try:\n+    import pynini\n+    from pynini.lib import pynutil\n+\n+    maj_singular_labels = load_labels(get_abs_path(\"data/money/currency_major.tsv\"))\n+    maj_singular = pynini.string_file((get_abs_path(\"data/money/currency_major.tsv\")))\n+    min_singular = pynini.string_file(get_abs_path(\"data/money/currency_minor.tsv\"))\n+    fem_plural = pynini.string_file((get_abs_path(\"data/money/currency_plural_fem.tsv\")))\n+    masc_plural = pynini.string_file((get_abs_path(\"data/money/currency_plural_masc.tsv\")))\n+\n+    PYNINI_AVAILABLE = True\n+\n+except (ModuleNotFoundError, ImportError):\n+    maj_singular_labels = None\n+    min_singular = None\n+    maj_singular = None\n+    fem_plural = None\n+    masc_plural = None\n+\n+    PYNINI_AVAILABLE = False\n+\n+\n+class MoneyFst(GraphFst):\n+    \"\"\"\n+    Finite state transducer for classifying money, e.g.\n+        \"\u20ac1\" -> money { currency_maj: \"euro\" integer_part: \"un\"}\n+        \"\u20ac1,000\" -> money { currency_maj: \"euro\" integer_part: \"un\" }\n+        \"\u20ac1,001\" -> money { currency_maj: \"euro\" integer_part: \"un\" fractional_part: \"cero cero un\" }\n+        \"\u00a31,4\" -> money { integer_part: \"una\" currency_maj: \"libra\" fractional_part: \"cuarenta\" preserve_order: true }\n+               -> money { integer_part: \"una\" currency_maj: \"libra\" fractional_part: \"cuarenta\" currency_min: \"penique\" preserve_order: true }\n+        \"0,01 \u00a3\" -> money { fractional_part: \"un\" currency_min: \"penique\" preserve_order: true }\n+        \"0,02 \u00a3\" -> money { fractional_part: \"dos\" currency_min: \"peniques\" preserve_order: true }\n+        \"\u00a30,01 million\" -> money { currency_maj: \"libra\" integer_part: \"cero\" fractional_part: \"cero un\" quantity: \"million\" }\n+\n+    Args:\n+        cardinal: CardinalFst\n+        decimal: DecimalFst\n+        deterministic: if True will provide a single transduction option,\n+            for False multiple transduction are generated (used for audio-based normalization)\n+    \"\"\"\n+\n+    def __init__(self, cardinal: GraphFst, decimal: GraphFst, deterministic: bool = True):\n+        super().__init__(name=\"money\", kind=\"classify\", deterministic=deterministic)\n+        cardinal_graph = cardinal.graph\n+        graph_decimal_final = decimal.final_graph_wo_negative\n+\n+        maj_singular_graph = maj_singular\n+        min_singular_graph = min_singular\n+        maj_plural_graph = maj_singular @ (fem_plural | masc_plural)\n+        min_plural_graph = min_singular @ (fem_plural | masc_plural)\n+\n+        graph_maj_singular = pynutil.insert(\"currency_maj: \\\"\") + maj_singular_graph + pynutil.insert(\"\\\"\")\n+        graph_maj_plural = pynutil.insert(\"currency_maj: \\\"\") + maj_plural_graph + pynutil.insert(\"\\\"\")\n+\n+        graph_integer_one = pynutil.insert(\"integer_part: \\\"\") + pynini.cross(\"1\", \"un\") + pynutil.insert(\"\\\"\")\n+\n+        decimal_with_quantity = (NEMO_SIGMA + NEMO_ALPHA) @ graph_decimal_final\n+\n+        graph_decimal_plural = pynini.union(\n+            graph_maj_plural + pynini.closure(delete_space, 0, 1) + insert_space + graph_decimal_final,  # $1,05\n+            graph_decimal_final + pynini.closure(delete_space, 0, 1) + insert_space + graph_maj_plural,  # 1,05 $\n+        )\n+        graph_decimal_plural = (\n+            (NEMO_SIGMA - \"1\") + decimal_separator + NEMO_SIGMA\n+        ) @ graph_decimal_plural  # Can't have \"un euros\"\n+\n+        graph_decimal_singular = pynini.union(\n+            graph_maj_singular + pynini.closure(delete_space, 0, 1) + insert_space + graph_decimal_final,  # $1,05\n+            graph_decimal_final + pynini.closure(delete_space, 0, 1) + insert_space + graph_maj_singular,  # 1,05 $\n+        )\n+        graph_decimal_singular = (pynini.accep(\"1\") + decimal_separator + NEMO_SIGMA) @ graph_decimal_singular\n+\n+        graph_decimal = pynini.union(\n+            graph_decimal_singular,\n+            graph_decimal_plural,\n+            graph_maj_plural + pynini.closure(delete_space, 0, 1) + insert_space + decimal_with_quantity,\n+        )\n+\n+        graph_integer = (\n+            pynutil.insert(\"integer_part: \\\"\") + ((NEMO_SIGMA - \"1\") @ cardinal_graph) + pynutil.insert(\"\\\"\")\n+        )\n+\n+        graph_integer_only = pynini.union(\n+            graph_maj_singular + pynini.closure(delete_space, 0, 1) + insert_space + graph_integer_one,\n+            graph_integer_one + pynini.closure(delete_space, 0, 1) + insert_space + graph_maj_singular,\n+        )\n+        graph_integer_only |= pynini.union(\n+            graph_maj_plural + pynini.closure(delete_space, 0, 1) + insert_space + graph_integer,\n+            graph_integer + pynini.closure(delete_space, 0, 1) + insert_space + graph_maj_plural,\n+        )\n+\n+        graph = graph_integer_only | graph_decimal\n+\n+        # remove trailing zeros of non zero number in the first 2 digits and fill up to 2 digits\n+        # e.g. 2000 -> 20, 0200->02, 01 -> 01, 10 -> 10\n+        # not accepted: 002, 00, 0,\n+        two_digits_fractional_part = (\n+            pynini.closure(NEMO_DIGIT) + (NEMO_DIGIT - \"0\") + pynini.closure(pynutil.delete(\"0\"))\n+        ) @ (\n+            (pynutil.delete(\"0\") + (NEMO_DIGIT - \"0\"))\n+            | ((NEMO_DIGIT - \"0\") + pynutil.insert(\"0\"))\n+            | ((NEMO_DIGIT - \"0\") + NEMO_DIGIT)\n+        )\n+\n+        graph_min_singular = pynutil.insert(\"currency_min: \\\"\") + min_singular_graph + pynutil.insert(\"\\\"\")\n+        graph_min_plural = pynutil.insert(\"currency_min: \\\"\") + min_plural_graph + pynutil.insert(\"\\\"\")\n+\n+        # format ** euro ** cent\n+        decimal_graph_with_minor = None\n+        for curr_symbol, _ in maj_singular_labels:\n+            preserve_order = pynutil.insert(\" preserve_order: true\")\n+\n+            integer_plus_maj = pynini.union(\n+                graph_integer + insert_space + pynutil.insert(curr_symbol) @ graph_maj_plural,\n+                graph_integer_one + insert_space + pynutil.insert(curr_symbol) @ graph_maj_singular,\n+            )\n+            # non zero integer part\n+            integer_plus_maj = (pynini.closure(NEMO_DIGIT) - \"0\") @ integer_plus_maj\n+\n+            graph_fractional_one = (\n+                pynutil.insert(\"fractional_part: \\\"\")\n+                + two_digits_fractional_part @ pynini.cross(\"1\", \"un\")\n+                + pynutil.insert(\"\\\"\")\n+            )\n+\n+            graph_fractional = (\n+                two_digits_fractional_part @ (pynini.closure(NEMO_DIGIT, 1, 2) - \"1\") @ cardinal.two_digit_non_zero\n+            )\n+            graph_fractional = pynutil.insert(\"fractional_part: \\\"\") + graph_fractional + pynutil.insert(\"\\\"\")\n+\n+            fractional_plus_min = pynini.union(\n+                graph_fractional + insert_space + pynutil.insert(curr_symbol) @ graph_min_plural,\n+                graph_fractional_one + insert_space + pynutil.insert(curr_symbol) @ graph_min_singular,\n+            )\n+\n+            decimal_graph_with_minor_curr = (\n+                integer_plus_maj + pynini.cross(decimal_separator, NEMO_SPACE) + fractional_plus_min\n+            )\n+            decimal_graph_with_minor_curr |= pynutil.add_weight(\n+                integer_plus_maj\n+                + pynini.cross(decimal_separator, NEMO_SPACE)\n+                + pynutil.insert(\"fractional_part: \\\"\")\n+                + two_digits_fractional_part @ cardinal.two_digit_non_zero\n+                + pynutil.insert(\"\\\"\"),\n+                weight=0.0001,\n+            )\n+\n+            decimal_graph_with_minor_curr |= pynutil.delete(\"0,\") + fractional_plus_min\n+            decimal_graph_with_minor_curr = pynini.union(\n+                pynutil.delete(curr_symbol)\n+                + pynini.closure(delete_space, 0, 1)\n+                + decimal_graph_with_minor_curr\n+                + preserve_order,\n+                decimal_graph_with_minor_curr\n+                + preserve_order\n+                + pynini.closure(delete_space, 0, 1)\n+                + pynutil.delete(curr_symbol),\n+            )\n+\n+            decimal_graph_with_minor = (\n+                decimal_graph_with_minor_curr\n+                if decimal_graph_with_minor is None\n+                else pynini.union(decimal_graph_with_minor, decimal_graph_with_minor_curr)\n+            )\n+\n+        final_graph = graph | pynutil.add_weight(decimal_graph_with_minor, -0.001)\n+\n+        final_graph = self.add_tokens(final_graph)\n+        self.fst = final_graph.optimize()\ndiff --git a/nemo_text_processing/text_normalization/es/taggers/ordinal.py b/nemo_text_processing/text_normalization/es/taggers/ordinal.py\nnew file mode 100644\n--- /dev/null\n+++ b/nemo_text_processing/text_normalization/es/taggers/ordinal.py\n@@ -0,0 +1,186 @@\n+# Copyright (c) 2022, NVIDIA CORPORATION & AFFILIATES.  All rights reserved.\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+#     http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+from nemo_text_processing.text_normalization.en.graph_utils import (\n+    NEMO_CHAR,\n+    NEMO_SIGMA,\n+    NEMO_SPACE,\n+    GraphFst,\n+    delete_space,\n+)\n+from nemo_text_processing.text_normalization.es.graph_utils import roman_to_int, strip_accent\n+from nemo_text_processing.text_normalization.es.utils import get_abs_path\n+\n+try:\n+    import pynini\n+    from pynini.lib import pynutil\n+\n+    digit = pynini.invert(pynini.string_file(get_abs_path(\"data/ordinals/digit.tsv\")))\n+    teens = pynini.invert(pynini.string_file(get_abs_path(\"data/ordinals/teen.tsv\")))\n+    twenties = pynini.invert(pynini.string_file(get_abs_path(\"data/ordinals/twenties.tsv\")))\n+    ties = pynini.invert(pynini.string_file(get_abs_path(\"data/ordinals/ties.tsv\")))\n+    hundreds = pynini.invert(pynini.string_file(get_abs_path(\"data/ordinals/hundreds.tsv\")))\n+\n+    PYNINI_AVAILABLE = True\n+\n+except (ImportError, ModuleNotFoundError):\n+    digit = None\n+    teens = None\n+    twenties = None\n+    ties = None\n+    hundreds = None\n+\n+    PYNINI_AVAILABLE = False\n+\n+\n+def get_one_to_one_thousand(cardinal: 'pynini.FstLike') -> 'pynini.FstLike':\n+    \"\"\"\n+    Produces an acceptor for verbalizations of all numbers from 1 to 1000. Needed for ordinals and fractions.\n+\n+    Args:\n+        cardinal: CardinalFst\n+\n+    Returns:\n+        fst: A pynini.FstLike object\n+    \"\"\"\n+    numbers = pynini.string_map([str(_) for _ in range(1, 1000)]) @ cardinal\n+    return pynini.project(numbers, \"output\").optimize()\n+\n+\n+class OrdinalFst(GraphFst):\n+    \"\"\"\n+    Finite state transducer for classifying ordinal\n+        \t\"21.\u00ba\" -> ordinal { integer: \"vig\u00e9simo primero\" morphosyntactic_features: \"gender_masc\" }\n+    This class converts ordinal up to the millionth (millon\u00e9simo) order (exclusive).\n+\n+    This FST also records the ending of the ordinal (called \"morphosyntactic_features\"):\n+    either as gender_masc, gender_fem, or apocope. Also introduces plural feature for non-deterministic graphs.\n+\n+    Args:\n+        cardinal: CardinalFst\n+        deterministic: if True will provide a single transduction option,\n+            for False multiple transduction are generated (used for audio-based normalization)\n+    \"\"\"\n+\n+    def __init__(self, cardinal: GraphFst, deterministic: bool = True):\n+        super().__init__(name=\"ordinal\", kind=\"classify\")\n+        cardinal_graph = cardinal.graph\n+\n+        graph_digit = digit.optimize()\n+        graph_teens = teens.optimize()\n+        graph_ties = ties.optimize()\n+        graph_twenties = twenties.optimize()\n+        graph_hundreds = hundreds.optimize()\n+\n+        if not deterministic:\n+            # Some alternative derivations\n+            graph_ties = graph_ties | pynini.cross(\"sesenta\", \"setuag\u00e9simo\")\n+\n+            graph_teens = graph_teens | pynini.cross(\"once\", \"decimoprimero\")\n+            graph_teens |= pynini.cross(\"doce\", \"decimosegundo\")\n+\n+            graph_digit = graph_digit | pynini.cross(\"nueve\", \"nono\")\n+            graph_digit |= pynini.cross(\"siete\", \"s\u00e9timo\")\n+\n+        graph_tens_component = (\n+            graph_teens\n+            | (graph_ties + pynini.closure(pynini.cross(\" y \", NEMO_SPACE) + graph_digit, 0, 1))\n+            | graph_twenties\n+        )\n+\n+        graph_hundred_component = pynini.union(\n+            graph_hundreds + pynini.closure(NEMO_SPACE + pynini.union(graph_tens_component, graph_digit), 0, 1),\n+            graph_tens_component,\n+            graph_digit,\n+        )\n+\n+        # Need to go up to thousands for fractions\n+        self.one_to_one_thousand = get_one_to_one_thousand(cardinal_graph)\n+\n+        thousands = pynini.cross(\"mil\", \"mil\u00e9simo\")\n+\n+        graph_thousands = (\n+            strip_accent(self.one_to_one_thousand) + NEMO_SPACE + thousands\n+        )  # Cardinals become prefix for thousands series. Snce accent on the powers of ten we strip accent from leading words\n+        graph_thousands @= pynini.cdrewrite(delete_space, \"\", \"mil\u00e9simo\", NEMO_SIGMA)  # merge as a prefix\n+        graph_thousands |= thousands\n+\n+        self.multiples_of_thousand = (cardinal_graph @ graph_thousands).optimize()\n+\n+        if (\n+            not deterministic\n+        ):  # Formally the words preceding the power of ten should be a prefix, but some maintain word boundaries.\n+            graph_thousands |= (self.one_to_one_thousand @ graph_hundred_component) + NEMO_SPACE + thousands\n+\n+        graph_thousands += pynini.closure(NEMO_SPACE + graph_hundred_component, 0, 1)\n+\n+        ordinal_graph = graph_thousands | graph_hundred_component\n+        ordinal_graph = cardinal_graph @ ordinal_graph\n+\n+        if not deterministic:\n+            # The 10's and 20's series can also be two words\n+            split_words = pynini.cross(\"decimo\", \"d\u00e9cimo \") | pynini.cross(\"vigesimo\", \"vig\u00e9simo \")\n+            split_words = pynini.cdrewrite(split_words, \"\", NEMO_CHAR, NEMO_SIGMA)\n+            ordinal_graph |= ordinal_graph @ split_words\n+\n+        # If \"octavo\" is preceeded by the \"o\" within string, it needs deletion\n+        ordinal_graph @= pynini.cdrewrite(pynutil.delete(\"o\"), \"\", \"octavo\", NEMO_SIGMA)\n+\n+        self.graph = ordinal_graph.optimize()\n+\n+        masc = pynini.accep(\"gender_masc\")\n+        fem = pynini.accep(\"gender_fem\")\n+        apocope = pynini.accep(\"apocope\")\n+\n+        delete_period = pynini.closure(pynutil.delete(\".\"), 0, 1)  # Sometimes the period is omitted f\n+\n+        accept_masc = delete_period + pynini.cross(\"\u00ba\", masc)\n+        accep_fem = delete_period + pynini.cross(\"\u00aa\", fem)\n+        accep_apocope = delete_period + pynini.cross(\"\u1d49\u02b3\", apocope)\n+\n+        # Managing Romanization\n+        graph_roman = pynutil.insert(\"integer: \\\"\") + roman_to_int(ordinal_graph) + pynutil.insert(\"\\\"\")\n+        if not deterministic:\n+            # Introduce plural\n+            plural = pynini.closure(pynutil.insert(\"/plural\"), 0, 1)\n+            accept_masc += plural\n+            accep_fem += plural\n+\n+            # Romanizations have no morphology marker, so in non-deterministic case we provide option for all\n+            insert_morphology = pynutil.insert(pynini.union(masc, fem)) + plural\n+            insert_morphology |= pynutil.insert(apocope)\n+            insert_morphology = (\n+                pynutil.insert(\" morphosyntactic_features: \\\"\") + insert_morphology + pynutil.insert(\"\\\"\")\n+            )\n+\n+            graph_roman += insert_morphology\n+\n+        else:\n+            # We assume masculine gender as default\n+            graph_roman += pynutil.insert(\" morphosyntactic_features: \\\"gender_masc\\\"\")\n+\n+        # Rest of graph\n+        convert_abbreviation = accept_masc | accep_fem | accep_apocope\n+\n+        graph = (\n+            pynutil.insert(\"integer: \\\"\")\n+            + ordinal_graph\n+            + pynutil.insert(\"\\\"\")\n+            + pynutil.insert(\" morphosyntactic_features: \\\"\")\n+            + convert_abbreviation\n+            + pynutil.insert(\"\\\"\")\n+        )\n+        graph = pynini.union(graph, graph_roman)\n+\n+        final_graph = self.add_tokens(graph)\n+        self.fst = final_graph.optimize()\ndiff --git a/nemo_text_processing/text_normalization/es/taggers/telephone.py b/nemo_text_processing/text_normalization/es/taggers/telephone.py\nnew file mode 100644\n--- /dev/null\n+++ b/nemo_text_processing/text_normalization/es/taggers/telephone.py\n@@ -0,0 +1,156 @@\n+# Copyright (c) 2022, NVIDIA CORPORATION & AFFILIATES.  All rights reserved.\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+#     http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+from nemo_text_processing.text_normalization.en.graph_utils import NEMO_SIGMA, GraphFst, insert_space\n+from nemo_text_processing.text_normalization.es.graph_utils import ones\n+from nemo_text_processing.text_normalization.es.utils import get_abs_path\n+\n+try:\n+    import pynini\n+    from pynini.lib import pynutil\n+\n+    graph_digit = pynini.string_file(get_abs_path(\"data/numbers/digit.tsv\"))\n+    graph_ties = pynini.string_file(get_abs_path(\"data/numbers/ties.tsv\"))\n+    graph_teen = pynini.string_file(get_abs_path(\"data/numbers/teen.tsv\"))\n+    graph_twenties = pynini.string_file(get_abs_path(\"data/numbers/twenties.tsv\"))\n+\n+    PYNINI_AVAILABLE = True\n+\n+except (ModuleNotFoundError, ImportError):\n+    graph_digit = None\n+    graph_ties = None\n+    graph_teen = None\n+    graph_twenties = None\n+\n+    PYNINI_AVAILABLE = False\n+\n+\n+class TelephoneFst(GraphFst):\n+    \"\"\"\n+    Finite state transducer for classifying telephone numbers, e.g.\n+        123-123-5678 -> { number_part: \"uno dos tres uno dos tres cinco seis siete ocho\" }.\n+        In Spanish, digits are generally read individually, or as 2-digit numbers,\n+        eg. \"123\" = \"uno dos tres\",\n+            \"1234\" = \"doce treinta y cuatro\".\n+        This will verbalize sequences of 10 (3+3+4 e.g. 123-456-7890).\n+        9 (3+3+3 e.g. 123-456-789) and 8 (4+4 e.g. 1234-5678) digits.\n+\n+        (we ignore more complicated cases such as \"doscientos y dos\" or \"tres nueves\").\n+\n+    Args:\n+\t\tdeterministic: if True will provide a single transduction option,\n+\t\t\tfor False multiple transduction are generated (used for audio-based normalization)\n+    \"\"\"\n+\n+    def __init__(self, deterministic: bool = True):\n+        super().__init__(name=\"telephone\", kind=\"classify\")\n+\n+        # create `single_digits` and `double_digits` graphs as these will be\n+        # the building blocks of possible telephone numbers\n+        single_digits = pynini.invert(graph_digit).optimize() | pynini.cross(\"0\", \"cero\")\n+\n+        double_digits = pynini.union(\n+            graph_twenties,\n+            graph_teen,\n+            (graph_ties + pynutil.delete(\"0\")),\n+            (graph_ties + insert_space + pynutil.insert(\"y\") + insert_space + graph_digit),\n+        )\n+        double_digits = pynini.invert(double_digits)\n+\n+        # define `ten_digit_graph`, `nine_digit_graph`, `eight_digit_graph`\n+        # which produces telephone numbers spoken (1) only with single digits,\n+        # or (2) spoken with double digits (and sometimes single digits)\n+\n+        # 10-digit option (1): all single digits\n+        ten_digit_graph = (\n+            pynini.closure(single_digits + insert_space, 3, 3)\n+            + pynutil.delete(\"-\")\n+            + pynini.closure(single_digits + insert_space, 3, 3)\n+            + pynutil.delete(\"-\")\n+            + pynini.closure(single_digits + insert_space, 3, 3)\n+            + single_digits\n+        )\n+\n+        # 9-digit option (1): all single digits\n+        nine_digit_graph = (\n+            pynini.closure(single_digits + insert_space, 3, 3)\n+            + pynutil.delete(\"-\")\n+            + pynini.closure(single_digits + insert_space, 3, 3)\n+            + pynutil.delete(\"-\")\n+            + pynini.closure(single_digits + insert_space, 2, 2)\n+            + single_digits\n+        )\n+\n+        # 8-digit option (1): all single digits\n+        eight_digit_graph = (\n+            pynini.closure(single_digits + insert_space, 4, 4)\n+            + pynutil.delete(\"-\")\n+            + pynini.closure(single_digits + insert_space, 3, 3)\n+            + single_digits\n+        )\n+\n+        if not deterministic:\n+            # 10-digit option (2): (1+2) + (1+2) + (2+2) digits\n+            ten_digit_graph |= (\n+                single_digits\n+                + insert_space\n+                + double_digits\n+                + insert_space\n+                + pynutil.delete(\"-\")\n+                + single_digits\n+                + insert_space\n+                + double_digits\n+                + insert_space\n+                + pynutil.delete(\"-\")\n+                + double_digits\n+                + insert_space\n+                + double_digits\n+            )\n+\n+            # 9-digit option (2): (1+2) + (1+2) + (1+2) digits\n+            nine_digit_graph |= (\n+                single_digits\n+                + insert_space\n+                + double_digits\n+                + insert_space\n+                + pynutil.delete(\"-\")\n+                + single_digits\n+                + insert_space\n+                + double_digits\n+                + insert_space\n+                + pynutil.delete(\"-\")\n+                + single_digits\n+                + insert_space\n+                + double_digits\n+            )\n+\n+            # 8-digit option (2): (2+2) + (2+2) digits\n+            eight_digit_graph |= (\n+                double_digits\n+                + insert_space\n+                + double_digits\n+                + insert_space\n+                + pynutil.delete(\"-\")\n+                + double_digits\n+                + insert_space\n+                + double_digits\n+            )\n+\n+        number_part = pynini.union(ten_digit_graph, nine_digit_graph, eight_digit_graph)\n+        number_part @= pynini.cdrewrite(pynini.cross(ones, \"uno\"), \"\", \"\", NEMO_SIGMA)\n+\n+        number_part = pynutil.insert(\"number_part: \\\"\") + number_part + pynutil.insert(\"\\\"\")\n+\n+        graph = number_part\n+        final_graph = self.add_tokens(graph)\n+        self.fst = final_graph.optimize()\ndiff --git a/nemo_text_processing/text_normalization/es/taggers/time.py b/nemo_text_processing/text_normalization/es/taggers/time.py\nnew file mode 100644\n--- /dev/null\n+++ b/nemo_text_processing/text_normalization/es/taggers/time.py\n@@ -0,0 +1,218 @@\n+# Copyright (c) 2022, NVIDIA CORPORATION & AFFILIATES.  All rights reserved.\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+#     http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+from nemo_text_processing.text_normalization.en.graph_utils import (\n+    NEMO_DIGIT,\n+    NEMO_SIGMA,\n+    GraphFst,\n+    delete_space,\n+    insert_space,\n+)\n+from nemo_text_processing.text_normalization.es.utils import get_abs_path\n+\n+try:\n+    import pynini\n+    from pynini.lib import pynutil\n+\n+    time_zone_graph = pynini.string_file(get_abs_path(\"data/time/time_zone.tsv\"))\n+    suffix = pynini.string_file(get_abs_path(\"data/time/time_suffix.tsv\"))\n+\n+    PYNINI_AVAILABLE = True\n+\n+except (ModuleNotFoundError, ImportError):\n+    time_zone_graph = None\n+    suffix = None\n+\n+    PYNINI_AVAILABLE = False\n+\n+\n+class TimeFst(GraphFst):\n+    \"\"\"\n+    Finite state transducer for classifying time, e.g.\n+        \"02:15 est\" -> time { hours: \"dos\" minutes: \"quince\" zone: \"e s t\"}\n+        \"2 h\" -> time { hours: \"dos\" }\n+        \"9 h\" -> time { hours: \"nueve\" }\n+        \"02:15:10 h\" -> time { hours: \"dos\" minutes: \"quince\" seconds: \"diez\"}\n+\n+    Args:\n+        cardinal: CardinalFst\n+        deterministic: if True will provide a single transduction option,\n+            for False multiple transduction are generated (used for audio-based normalization)\n+    \"\"\"\n+\n+    def __init__(self, cardinal: GraphFst, deterministic: bool = True):\n+        super().__init__(name=\"time\", kind=\"classify\", deterministic=deterministic)\n+\n+        delete_time_delimiter = pynutil.delete(pynini.union(\".\", \":\"))\n+\n+        one = pynini.string_map([(\"un\", \"una\"), (\"\u00fan\", \"una\")])\n+        change_one = pynini.cdrewrite(one, \"\", \"\", NEMO_SIGMA)\n+        cardinal_graph = cardinal.graph @ change_one\n+\n+        day_suffix = pynutil.insert(\"suffix: \\\"\") + suffix + pynutil.insert(\"\\\"\")\n+        day_suffix = delete_space + insert_space + day_suffix\n+\n+        delete_hora_suffix = delete_space + insert_space + pynutil.delete(\"h\")\n+        delete_minute_suffix = delete_space + insert_space + pynutil.delete(\"min\")\n+        delete_second_suffix = delete_space + insert_space + pynutil.delete(\"s\")\n+\n+        labels_hour_24 = [\n+            str(x) for x in range(0, 25)\n+        ]  # Can see both systems. Twelve hour requires am/pm for ambiguity resolution\n+        labels_hour_12 = [str(x) for x in range(1, 13)]\n+        labels_minute_single = [str(x) for x in range(1, 10)]\n+        labels_minute_double = [str(x) for x in range(10, 60)]\n+\n+        delete_leading_zero_to_double_digit = (\n+            pynini.closure(pynutil.delete(\"0\") | (NEMO_DIGIT - \"0\"), 0, 1) + NEMO_DIGIT\n+        )\n+\n+        graph_24 = (\n+            pynini.closure(NEMO_DIGIT, 1, 2) @ delete_leading_zero_to_double_digit @ pynini.union(*labels_hour_24)\n+        )\n+        graph_12 = (\n+            pynini.closure(NEMO_DIGIT, 1, 2) @ delete_leading_zero_to_double_digit @ pynini.union(*labels_hour_12)\n+        )\n+\n+        graph_hour_24 = graph_24 @ cardinal_graph\n+        graph_hour_12 = graph_12 @ cardinal_graph\n+\n+        graph_minute_single = delete_leading_zero_to_double_digit @ pynini.union(*labels_minute_single)\n+        graph_minute_double = pynini.union(*labels_minute_double)\n+\n+        graph_minute = pynini.union(graph_minute_single, graph_minute_double) @ cardinal_graph\n+\n+        final_graph_hour_only_24 = (\n+            pynutil.insert(\"hours: \\\"\") + graph_hour_24 + pynutil.insert(\"\\\"\") + delete_hora_suffix\n+        )\n+        final_graph_hour_only_12 = pynutil.insert(\"hours: \\\"\") + graph_hour_12 + pynutil.insert(\"\\\"\") + day_suffix\n+\n+        final_graph_hour_24 = pynutil.insert(\"hours: \\\"\") + graph_hour_24 + pynutil.insert(\"\\\"\")\n+        final_graph_hour_12 = pynutil.insert(\"hours: \\\"\") + graph_hour_12 + pynutil.insert(\"\\\"\")\n+\n+        final_graph_minute = pynutil.insert(\"minutes: \\\"\") + graph_minute + pynutil.insert(\"\\\"\")\n+        final_graph_second = pynutil.insert(\"seconds: \\\"\") + graph_minute + pynutil.insert(\"\\\"\")\n+        final_time_zone_optional = pynini.closure(\n+            delete_space + insert_space + pynutil.insert(\"zone: \\\"\") + time_zone_graph + pynutil.insert(\"\\\"\"), 0, 1,\n+        )\n+\n+        # 02.30 h\n+        graph_hm = (\n+            final_graph_hour_24\n+            + delete_time_delimiter\n+            + (pynutil.delete(\"00\") | (insert_space + final_graph_minute))\n+            + pynini.closure(\n+                delete_time_delimiter + (pynini.cross(\"00\", \" seconds: \\\"0\\\"\") | (insert_space + final_graph_second)),\n+                0,\n+                1,\n+            )  # For seconds 2.30.35 h\n+            + pynini.closure(delete_hora_suffix, 0, 1)  # 2.30 is valid if unambiguous\n+            + final_time_zone_optional\n+        )\n+\n+        # 2 h 30 min\n+        graph_hm |= (\n+            final_graph_hour_24\n+            + delete_hora_suffix\n+            + delete_space\n+            + (pynutil.delete(\"00\") | (insert_space + final_graph_minute))\n+            + delete_minute_suffix\n+            + pynini.closure(\n+                delete_space\n+                + (pynini.cross(\"00\", \" seconds: \\\"0\\\"\") | (insert_space + final_graph_second))\n+                + delete_second_suffix,\n+                0,\n+                1,\n+            )  # For seconds\n+            + final_time_zone_optional\n+        )\n+\n+        # 2.30 a. m. (Only for 12 hour clock)\n+        graph_hm |= (\n+            final_graph_hour_12\n+            + delete_time_delimiter\n+            + (pynutil.delete(\"00\") | (insert_space + final_graph_minute))\n+            + pynini.closure(\n+                delete_time_delimiter + (pynini.cross(\"00\", \" seconds: \\\"0\\\"\") | (insert_space + final_graph_second)),\n+                0,\n+                1,\n+            )  # For seconds 2.30.35 a. m.\n+            + day_suffix\n+            + final_time_zone_optional\n+        )\n+\n+        graph_h = (\n+            pynini.union(final_graph_hour_only_24, final_graph_hour_only_12) + final_time_zone_optional\n+        )  # Should always have a time indicator, else we'll pass to cardinals\n+\n+        if not deterministic:\n+            # This includes alternate vocalization (hour menos min, min para hour), here we shift the times and indicate a `style` tag\n+            hour_shift_24 = pynini.invert(pynini.string_file(get_abs_path(\"data/time/hour_to_24.tsv\")))\n+            hour_shift_12 = pynini.invert(pynini.string_file(get_abs_path(\"data/time/hour_to_12.tsv\")))\n+            minute_shift = pynini.string_file(get_abs_path(\"data/time/minute_to.tsv\"))\n+\n+            graph_hour_to_24 = graph_24 @ hour_shift_24 @ cardinal_graph\n+            graph_hour_to_12 = graph_12 @ hour_shift_12 @ cardinal_graph\n+\n+            graph_minute_to = pynini.union(graph_minute_single, graph_minute_double) @ minute_shift @ cardinal_graph\n+\n+            final_graph_hour_to_24 = pynutil.insert(\"hours: \\\"\") + graph_hour_to_24 + pynutil.insert(\"\\\"\")\n+            final_graph_hour_to_12 = pynutil.insert(\"hours: \\\"\") + graph_hour_to_12 + pynutil.insert(\"\\\"\")\n+\n+            final_graph_minute_to = pynutil.insert(\"minutes: \\\"\") + graph_minute_to + pynutil.insert(\"\\\"\")\n+\n+            graph_menos = pynutil.insert(\" style: \\\"1\\\"\")\n+            graph_para = pynutil.insert(\" style: \\\"2\\\"\")\n+\n+            final_graph_style = graph_menos | graph_para\n+\n+            # 02.30 h (omitting seconds since a bit awkward)\n+            graph_hm |= (\n+                final_graph_hour_to_24\n+                + delete_time_delimiter\n+                + insert_space\n+                + final_graph_minute_to\n+                + pynini.closure(delete_hora_suffix, 0, 1)  # 2.30 is valid if unambiguous\n+                + final_time_zone_optional\n+                + final_graph_style\n+            )\n+\n+            # 2 h 30 min\n+            graph_hm |= (\n+                final_graph_hour_to_24\n+                + delete_hora_suffix\n+                + delete_space\n+                + insert_space\n+                + final_graph_minute_to\n+                + delete_minute_suffix\n+                + final_time_zone_optional\n+                + final_graph_style\n+            )\n+\n+            # 2.30 a. m. (Only for 12 hour clock)\n+            graph_hm |= (\n+                final_graph_hour_to_12\n+                + delete_time_delimiter\n+                + insert_space\n+                + final_graph_minute_to\n+                + day_suffix\n+                + final_time_zone_optional\n+                + final_graph_style\n+            )\n+\n+        final_graph = graph_hm | graph_h\n+        if deterministic:\n+            final_graph = final_graph + pynutil.insert(\" preserve_order: true\")\n+        final_graph = final_graph.optimize()\n+        final_graph = self.add_tokens(final_graph)\n+        self.fst = final_graph.optimize()\ndiff --git a/nemo_text_processing/text_normalization/es/taggers/tokenize_and_classify.py b/nemo_text_processing/text_normalization/es/taggers/tokenize_and_classify.py\nnew file mode 100644\n--- /dev/null\n+++ b/nemo_text_processing/text_normalization/es/taggers/tokenize_and_classify.py\n@@ -0,0 +1,157 @@\n+# Copyright (c) 2022, NVIDIA CORPORATION & AFFILIATES.  All rights reserved.\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+#     http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+\n+import os\n+\n+from nemo_text_processing.text_normalization.en.graph_utils import (\n+    NEMO_WHITE_SPACE,\n+    GraphFst,\n+    delete_extra_space,\n+    delete_space,\n+    generator_main,\n+)\n+from nemo_text_processing.text_normalization.en.taggers.punctuation import PunctuationFst\n+from nemo_text_processing.text_normalization.es.taggers.cardinal import CardinalFst\n+from nemo_text_processing.text_normalization.es.taggers.date import DateFst\n+from nemo_text_processing.text_normalization.es.taggers.decimals import DecimalFst\n+from nemo_text_processing.text_normalization.es.taggers.electronic import ElectronicFst\n+from nemo_text_processing.text_normalization.es.taggers.fraction import FractionFst\n+from nemo_text_processing.text_normalization.es.taggers.measure import MeasureFst\n+from nemo_text_processing.text_normalization.es.taggers.money import MoneyFst\n+from nemo_text_processing.text_normalization.es.taggers.ordinal import OrdinalFst\n+from nemo_text_processing.text_normalization.es.taggers.telephone import TelephoneFst\n+from nemo_text_processing.text_normalization.es.taggers.time import TimeFst\n+from nemo_text_processing.text_normalization.es.taggers.whitelist import WhiteListFst\n+from nemo_text_processing.text_normalization.es.taggers.word import WordFst\n+\n+from nemo.utils import logging\n+\n+try:\n+    import pynini\n+    from pynini.lib import pynutil\n+\n+    PYNINI_AVAILABLE = True\n+\n+except (ModuleNotFoundError, ImportError):\n+    PYNINI_AVAILABLE = False\n+\n+\n+class ClassifyFst(GraphFst):\n+    \"\"\"\n+    Final class that composes all other classification grammars. This class can process an entire sentence, that is lower cased.\n+    For deployment, this grammar will be compiled and exported to OpenFst Finate State aRchive (FAR) File.\n+    More details to deployment at NeMo/tools/text_processing_deployment.\n+\n+    Args:\n+        input_case: accepting either \"lower_cased\" or \"cased\" input.\n+        deterministic: if True will provide a single transduction option,\n+            for False multiple options (used for audio-based normalization)\n+        cache_dir: path to a dir with .far grammar file. Set to None to avoid using cache.\n+        overwrite_cache: set to True to overwrite .far files\n+        whitelist: path to a file with whitelist replacements\n+    \"\"\"\n+\n+    def __init__(\n+        self,\n+        input_case: str,\n+        deterministic: bool = False,\n+        cache_dir: str = None,\n+        overwrite_cache: bool = False,\n+        whitelist: str = None,\n+    ):\n+        super().__init__(name=\"tokenize_and_classify\", kind=\"classify\", deterministic=deterministic)\n+        far_file = None\n+        if cache_dir is not None and cache_dir != \"None\":\n+            os.makedirs(cache_dir, exist_ok=True)\n+            whitelist_file = os.path.basename(whitelist) if whitelist else \"\"\n+            far_file = os.path.join(\n+                cache_dir, f\"_{input_case}_es_tn_{deterministic}_deterministic{whitelist_file}.far\"\n+            )\n+        if not overwrite_cache and far_file and os.path.exists(far_file):\n+            self.fst = pynini.Far(far_file, mode=\"r\")[\"tokenize_and_classify\"]\n+            logging.info(f\"ClassifyFst.fst was restored from {far_file}.\")\n+        else:\n+            logging.info(f\"Creating ClassifyFst grammars. This might take some time...\")\n+\n+            self.cardinal = CardinalFst(deterministic=deterministic)\n+            cardinal_graph = self.cardinal.fst\n+\n+            self.ordinal = OrdinalFst(cardinal=self.cardinal, deterministic=deterministic)\n+            ordinal_graph = self.ordinal.fst\n+\n+            self.decimal = DecimalFst(cardinal=self.cardinal, deterministic=deterministic)\n+            decimal_graph = self.decimal.fst\n+\n+            self.fraction = FractionFst(cardinal=self.cardinal, ordinal=self.ordinal, deterministic=deterministic)\n+            fraction_graph = self.fraction.fst\n+            self.measure = MeasureFst(\n+                cardinal=self.cardinal, decimal=self.decimal, fraction=self.fraction, deterministic=deterministic\n+            )\n+            measure_graph = self.measure.fst\n+            self.date = DateFst(cardinal=self.cardinal, deterministic=deterministic)\n+            date_graph = self.date.fst\n+            word_graph = WordFst(deterministic=deterministic).fst\n+            self.time = TimeFst(self.cardinal, deterministic=deterministic)\n+            time_graph = self.time.fst\n+            self.telephone = TelephoneFst(deterministic=deterministic)\n+            telephone_graph = self.telephone.fst\n+            self.electronic = ElectronicFst(deterministic=deterministic)\n+            electronic_graph = self.electronic.fst\n+            self.money = MoneyFst(cardinal=self.cardinal, decimal=self.decimal, deterministic=deterministic)\n+            money_graph = self.money.fst\n+            self.whitelist = WhiteListFst(input_case=input_case, deterministic=deterministic, input_file=whitelist)\n+            whitelist_graph = self.whitelist.fst\n+            punct_graph = PunctuationFst(deterministic=deterministic).fst\n+\n+            classify = (\n+                pynutil.add_weight(whitelist_graph, 1.01)\n+                | pynutil.add_weight(time_graph, 1.09)\n+                | pynutil.add_weight(measure_graph, 1.08)\n+                | pynutil.add_weight(cardinal_graph, 1.1)\n+                | pynutil.add_weight(fraction_graph, 1.09)\n+                | pynutil.add_weight(date_graph, 1.1)\n+                | pynutil.add_weight(ordinal_graph, 1.1)\n+                | pynutil.add_weight(decimal_graph, 1.1)\n+                | pynutil.add_weight(money_graph, 1.1)\n+                | pynutil.add_weight(telephone_graph, 1.1)\n+                | pynutil.add_weight(electronic_graph, 1.1)\n+                | pynutil.add_weight(word_graph, 200)\n+            )\n+            punct = pynutil.insert(\"tokens { \") + pynutil.add_weight(punct_graph, weight=2.1) + pynutil.insert(\" }\")\n+            punct = pynini.closure(\n+                pynini.compose(pynini.closure(NEMO_WHITE_SPACE, 1), delete_extra_space)\n+                | (pynutil.insert(\" \") + punct),\n+                1,\n+            )\n+            token = pynutil.insert(\"tokens { \") + classify + pynutil.insert(\" }\")\n+            token_plus_punct = (\n+                pynini.closure(punct + pynutil.insert(\" \")) + token + pynini.closure(pynutil.insert(\" \") + punct)\n+            )\n+\n+            graph = token_plus_punct + pynini.closure(\n+                (\n+                    pynini.compose(pynini.closure(NEMO_WHITE_SPACE, 1), delete_extra_space)\n+                    | (pynutil.insert(\" \") + punct + pynutil.insert(\" \"))\n+                )\n+                + token_plus_punct\n+            )\n+\n+            graph = delete_space + graph + delete_space\n+            graph |= punct\n+\n+            self.fst = graph.optimize()\n+\n+            if far_file:\n+                generator_main(far_file, {\"tokenize_and_classify\": self.fst})\n+                logging.info(f\"ClassifyFst grammars are saved to {far_file}.\")\ndiff --git a/nemo_text_processing/text_normalization/es/taggers/whitelist.py b/nemo_text_processing/text_normalization/es/taggers/whitelist.py\nnew file mode 100644\n--- /dev/null\n+++ b/nemo_text_processing/text_normalization/es/taggers/whitelist.py\n@@ -0,0 +1,69 @@\n+# Copyright (c) 2022, NVIDIA CORPORATION & AFFILIATES.  All rights reserved.\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+#     http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+from nemo_text_processing.text_normalization.en.graph_utils import GraphFst, convert_space\n+from nemo_text_processing.text_normalization.es.utils import get_abs_path, load_labels\n+\n+try:\n+    import pynini\n+    from pynini.lib import pynutil\n+\n+    PYNINI_AVAILABLE = True\n+\n+except (ModuleNotFoundError, ImportError):\n+    PYNINI_AVAILABLE = False\n+\n+\n+class WhiteListFst(GraphFst):\n+    \"\"\"\n+    Finite state transducer for classifying whitelist, e.g.\n+        \"sr.\" -> tokens { name: \"se\u00f1or\" }\n+    This class has highest priority among all classifier grammars. Whitelisted tokens are defined and loaded from \"data/whitelist.tsv\".\n+\n+    Args:\n+        input_case: accepting either \"lower_cased\" or \"cased\" input.\n+        deterministic: if True will provide a single transduction option,\n+            for False multiple options (used for audio-based normalization)\n+        input_file: path to a file with whitelist replacements\n+    \"\"\"\n+\n+    def __init__(self, input_case: str, deterministic: bool = True, input_file: str = None):\n+        super().__init__(name=\"whitelist\", kind=\"classify\", deterministic=deterministic)\n+\n+        def _get_whitelist_graph(input_case, file):\n+            whitelist = load_labels(file)\n+            if input_case == \"lower_cased\":\n+                whitelist = [[x[0].lower()] + x[1:] for x in whitelist]\n+            graph = pynini.string_map(whitelist)\n+            return graph\n+\n+        graph = _get_whitelist_graph(input_case, get_abs_path(\"data/whitelist.tsv\"))\n+        if not deterministic and input_case != \"lower_cased\":\n+            graph |= pynutil.add_weight(\n+                _get_whitelist_graph(\"lower_cased\", get_abs_path(\"data/whitelist.tsv\")), weight=0.0001\n+            )\n+\n+        if input_file:\n+            whitelist_provided = _get_whitelist_graph(input_case, input_file)\n+            if not deterministic:\n+                graph |= whitelist_provided\n+            else:\n+                graph = whitelist_provided\n+\n+        if not deterministic:\n+            units_graph = _get_whitelist_graph(input_case, file=get_abs_path(\"data/measures/measurements.tsv\"))\n+            graph |= units_graph\n+\n+        self.graph = graph\n+        self.final_graph = convert_space(self.graph).optimize()\n+        self.fst = (pynutil.insert(\"name: \\\"\") + self.final_graph + pynutil.insert(\"\\\"\")).optimize()\ndiff --git a/nemo_text_processing/text_normalization/es/taggers/word.py b/nemo_text_processing/text_normalization/es/taggers/word.py\nnew file mode 100644\n--- /dev/null\n+++ b/nemo_text_processing/text_normalization/es/taggers/word.py\n@@ -0,0 +1,39 @@\n+# Copyright (c) 2022, NVIDIA CORPORATION & AFFILIATES.  All rights reserved.\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+#     http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+from nemo_text_processing.text_normalization.en.graph_utils import NEMO_NOT_SPACE, GraphFst\n+\n+try:\n+    import pynini\n+    from pynini.lib import pynutil\n+\n+    PYNINI_AVAILABLE = True\n+\n+except (ModuleNotFoundError, ImportError):\n+    PYNINI_AVAILABLE = False\n+\n+\n+class WordFst(GraphFst):\n+    \"\"\"\n+    Finite state transducer for classifying word.\n+        e.g. dormir -> tokens { name: \"dormir\" }\n+\n+    Args:\n+        deterministic: if True will provide a single transduction option,\n+            for False multiple transduction are generated (used for audio-based normalization)\n+    \"\"\"\n+\n+    def __init__(self, deterministic: bool = True):\n+        super().__init__(name=\"word\", kind=\"classify\")\n+        word = pynutil.insert(\"name: \\\"\") + pynini.closure(NEMO_NOT_SPACE, 1) + pynutil.insert(\"\\\"\")\n+        self.fst = word.optimize()\ndiff --git a/nemo_text_processing/text_normalization/es/utils.py b/nemo_text_processing/text_normalization/es/utils.py\nnew file mode 100644\n--- /dev/null\n+++ b/nemo_text_processing/text_normalization/es/utils.py\n@@ -0,0 +1,42 @@\n+# Copyright (c) 2022, NVIDIA CORPORATION & AFFILIATES.  All rights reserved.\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+#     http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+\n+import csv\n+import os\n+\n+\n+def get_abs_path(rel_path):\n+    \"\"\"\n+    Get absolute path\n+\n+    Args:\n+        rel_path: relative path to this file\n+\n+    Returns absolute path\n+    \"\"\"\n+    return os.path.dirname(os.path.abspath(__file__)) + '/' + rel_path\n+\n+\n+def load_labels(abs_path):\n+    \"\"\"\n+    loads relative path file as dictionary\n+\n+    Args:\n+        abs_path: absolute path\n+\n+    Returns dictionary of mappings\n+    \"\"\"\n+    label_tsv = open(abs_path)\n+    labels = list(csv.reader(label_tsv, delimiter=\"\\t\"))\n+    return labels\ndiff --git a/nemo_text_processing/text_normalization/es/verbalizers/__init__.py b/nemo_text_processing/text_normalization/es/verbalizers/__init__.py\nnew file mode 100644\n--- /dev/null\n+++ b/nemo_text_processing/text_normalization/es/verbalizers/__init__.py\n@@ -0,0 +1,13 @@\n+# Copyright (c) 2022, NVIDIA CORPORATION & AFFILIATES.  All rights reserved.\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+#     http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\ndiff --git a/nemo_text_processing/text_normalization/es/verbalizers/cardinal.py b/nemo_text_processing/text_normalization/es/verbalizers/cardinal.py\nnew file mode 100644\n--- /dev/null\n+++ b/nemo_text_processing/text_normalization/es/verbalizers/cardinal.py\n@@ -0,0 +1,57 @@\n+# Copyright (c) 2022, NVIDIA CORPORATION & AFFILIATES.  All rights reserved.\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+#     http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+from nemo_text_processing.text_normalization.en.graph_utils import NEMO_NOT_QUOTE, GraphFst\n+from nemo_text_processing.text_normalization.es.graph_utils import shift_cardinal_gender, strip_cardinal_apocope\n+\n+try:\n+    import pynini\n+    from pynini.lib import pynutil\n+\n+    PYNINI_AVAILABLE = True\n+\n+except (ModuleNotFoundError, ImportError):\n+    PYNINI_AVAILABLE = False\n+\n+\n+class CardinalFst(GraphFst):\n+    \"\"\"\n+\tFinite state transducer for verbalizing cardinals\n+\t\te.g. cardinal { integer: \"dos\" } -> \"dos\"\n+\n+\tArgs:\n+\t\tdeterministic: if True will provide a single transduction option,\n+\t\t\tfor False multiple transduction are generated (used for audio-based normalization)\n+\t\"\"\"\n+\n+    def __init__(self, deterministic: bool = True):\n+        super().__init__(name=\"cardinal\", kind=\"verbalize\", deterministic=deterministic)\n+        optional_sign = pynini.closure(pynini.cross(\"negative: \\\"true\\\" \", \"menos \"), 0, 1)\n+        self.optional_sign = optional_sign\n+\n+        integer = pynini.closure(NEMO_NOT_QUOTE, 1)\n+        self.integer = pynutil.delete(\" \\\"\") + integer + pynutil.delete(\"\\\"\")\n+\n+        integer = pynutil.delete(\"integer:\") + self.integer\n+        self.numbers = integer\n+        graph = optional_sign + self.numbers\n+\n+        if not deterministic:\n+            # For alternate renderings\n+            no_adjust = graph\n+            fem_adjust = shift_cardinal_gender(graph)\n+            apocope_adjust = strip_cardinal_apocope(graph)\n+            graph = no_adjust | fem_adjust | apocope_adjust\n+\n+        delete_tokens = self.delete_tokens(graph)\n+        self.fst = delete_tokens.optimize()\ndiff --git a/nemo_text_processing/text_normalization/es/verbalizers/date.py b/nemo_text_processing/text_normalization/es/verbalizers/date.py\nnew file mode 100644\n--- /dev/null\n+++ b/nemo_text_processing/text_normalization/es/verbalizers/date.py\n@@ -0,0 +1,86 @@\n+# Copyright (c) 2022, NVIDIA CORPORATION & AFFILIATES.  All rights reserved.\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+#     http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+from nemo_text_processing.text_normalization.en.graph_utils import (\n+    NEMO_NOT_QUOTE,\n+    NEMO_SIGMA,\n+    NEMO_SPACE,\n+    GraphFst,\n+    delete_preserve_order,\n+)\n+from nemo_text_processing.text_normalization.es.graph_utils import strip_cardinal_apocope\n+from nemo_text_processing.text_normalization.es.taggers.date import articles\n+\n+try:\n+    import pynini\n+    from pynini.lib import pynutil\n+\n+    PYNINI_AVAILABLE = True\n+\n+except (ModuleNotFoundError, ImportError):\n+    PYNINI_AVAILABLE = False\n+\n+\n+class DateFst(GraphFst):\n+    \"\"\"\n+    Finite state transducer for verbalizing date, e.g.\n+        date { day: \"treinta y uno\" month: \"marzo\" year: \"dos mil\" } -> \"treinta y uno de marzo de dos mil\"\n+        date { day: \"uno\" month: \"mayo\" year: \"del mil novecientos noventa\" } -> \"primero de mayo del mil novecientos noventa\"\n+\n+    Args:\n+        deterministic: if True will provide a single transduction option,\n+            for False multiple transduction are generated (used for audio-based normalization)\n+    \"\"\"\n+\n+    def __init__(self, deterministic: bool = True):\n+        super().__init__(name=\"date\", kind=\"verbalize\", deterministic=deterministic)\n+\n+        day_cardinal = pynutil.delete(\"day: \\\"\") + pynini.closure(NEMO_NOT_QUOTE, 1) + pynutil.delete(\"\\\"\")\n+        day = strip_cardinal_apocope(day_cardinal)\n+\n+        primero = pynini.cdrewrite(pynini.cross(\"uno\", \"primero\"), \"[BOS]\", \"[EOS]\", NEMO_SIGMA)\n+        day = (\n+            (day @ primero) if deterministic else pynini.union(day, day @ primero)\n+        )  # Primero for first day is traditional, but will vary depending on region\n+\n+        month = pynutil.delete(\"month: \\\"\") + pynini.closure(NEMO_NOT_QUOTE, 1) + pynutil.delete(\"\\\"\")\n+\n+        year = (\n+            pynutil.delete(\"year: \\\"\")\n+            + articles\n+            + NEMO_SPACE\n+            + pynini.closure(NEMO_NOT_QUOTE, 1)\n+            + pynutil.delete(\"\\\"\")\n+        )\n+\n+        # Insert preposition if wasn't originally with the year. This would mean a space was present\n+        year = pynutil.add_weight(year, -0.001)\n+        year |= (\n+            pynutil.delete(\"year: \\\"\")\n+            + pynutil.insert(\"de \")\n+            + pynini.closure(NEMO_NOT_QUOTE, 1)\n+            + pynutil.delete(\"\\\"\")\n+        )\n+\n+        # day month year\n+        graph_dmy = day + pynini.cross(NEMO_SPACE, \" de \") + month + pynini.closure(pynini.accep(\" \") + year, 0, 1)\n+\n+        graph_mdy = month + NEMO_SPACE + day + pynini.closure(NEMO_SPACE + year, 0, 1)\n+        if deterministic:\n+            graph_mdy += pynutil.delete(\" preserve_order: true\")  # Only accepts this if was explicitly passed\n+\n+        self.graph = graph_dmy | graph_mdy\n+        final_graph = self.graph + delete_preserve_order\n+\n+        delete_tokens = self.delete_tokens(final_graph)\n+        self.fst = delete_tokens.optimize()\ndiff --git a/nemo_text_processing/text_normalization/es/verbalizers/decimals.py b/nemo_text_processing/text_normalization/es/verbalizers/decimals.py\nnew file mode 100644\n--- /dev/null\n+++ b/nemo_text_processing/text_normalization/es/verbalizers/decimals.py\n@@ -0,0 +1,87 @@\n+# Copyright (c) 2022, NVIDIA CORPORATION & AFFILIATES.  All rights reserved.\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+#     http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+from nemo_text_processing.text_normalization.en.graph_utils import (\n+    NEMO_NOT_QUOTE,\n+    GraphFst,\n+    delete_preserve_order,\n+    delete_space,\n+    insert_space,\n+)\n+from nemo_text_processing.text_normalization.es import LOCALIZATION\n+from nemo_text_processing.text_normalization.es.graph_utils import (\n+    shift_cardinal_gender,\n+    shift_number_gender,\n+    strip_cardinal_apocope,\n+)\n+\n+try:\n+    import pynini\n+    from pynini.lib import pynutil\n+\n+    PYNINI_AVAILABLE = True\n+\n+except (ModuleNotFoundError, ImportError):\n+    PYNINI_AVAILABLE = False\n+\n+\n+class DecimalFst(GraphFst):\n+    \"\"\"\n+\tFinite state transducer for classifying decimal, e.g.\n+\t\tdecimal { negative: \"true\" integer_part: \"dos\"  fractional_part: \"cuatro cero\" quantity: \"billones\" } -> menos dos coma quatro cero billones\n+\t\tdecimal { integer_part: \"un\" quantity: \"bill\u00f3n\" } -> un bill\u00f3n\n+\n+    Args:\n+\t\tdeterministic: if True will provide a single transduction option,\n+\t\t\tfor False multiple transduction are generated (used for audio-based normalization)\n+\t\"\"\"\n+\n+    def __init__(self, deterministic: bool = True):\n+        super().__init__(name=\"decimal\", kind=\"classify\", deterministic=deterministic)\n+\n+        self.optional_sign = pynini.closure(pynini.cross(\"negative: \\\"true\\\"\", \"menos \") + delete_space, 0, 1)\n+        self.integer = pynutil.delete(\"integer_part: \\\"\") + pynini.closure(NEMO_NOT_QUOTE, 1) + pynutil.delete(\"\\\"\")\n+        self.fractional_default = (\n+            pynutil.delete(\"fractional_part: \\\"\") + pynini.closure(NEMO_NOT_QUOTE, 1) + pynutil.delete(\"\\\"\")\n+        )\n+\n+        conjunction = pynutil.insert(\" punto \") if LOCALIZATION == \"am\" else pynutil.insert(\" coma \")\n+        if not deterministic:\n+            conjunction |= pynutil.insert(pynini.union(\" con \", \" y \"))\n+            self.fractional_default |= strip_cardinal_apocope(self.fractional_default)\n+        self.fractional = conjunction + self.fractional_default\n+\n+        self.quantity = (\n+            delete_space\n+            + insert_space\n+            + pynutil.delete(\"quantity: \\\"\")\n+            + pynini.closure(NEMO_NOT_QUOTE, 1)\n+            + pynutil.delete(\"\\\"\")\n+        )\n+        self.optional_quantity = pynini.closure(self.quantity, 0, 1)\n+\n+        graph = self.optional_sign + pynini.union(\n+            (self.integer + self.quantity), (self.integer + delete_space + self.fractional + self.optional_quantity)\n+        )\n+\n+        self.numbers = graph.optimize()\n+        self.numbers_no_quantity = self.integer + delete_space + self.fractional + self.optional_quantity\n+\n+        if not deterministic:\n+            graph |= self.optional_sign + (\n+                shift_cardinal_gender(self.integer + delete_space) + shift_number_gender(self.fractional)\n+            )\n+\n+        graph += delete_preserve_order\n+        delete_tokens = self.delete_tokens(graph)\n+        self.fst = delete_tokens.optimize()\ndiff --git a/nemo_text_processing/text_normalization/es/verbalizers/electronic.py b/nemo_text_processing/text_normalization/es/verbalizers/electronic.py\nnew file mode 100644\n--- /dev/null\n+++ b/nemo_text_processing/text_normalization/es/verbalizers/electronic.py\n@@ -0,0 +1,91 @@\n+# Copyright (c) 2022, NVIDIA CORPORATION & AFFILIATES.  All rights reserved.\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+#     http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+from nemo_text_processing.text_normalization.en.graph_utils import (\n+    NEMO_NOT_QUOTE,\n+    NEMO_SIGMA,\n+    GraphFst,\n+    delete_preserve_order,\n+    insert_space,\n+)\n+from nemo_text_processing.text_normalization.es.utils import get_abs_path\n+\n+try:\n+    import pynini\n+    from pynini.lib import pynutil\n+\n+    digit_no_zero = pynini.invert(pynini.string_file(get_abs_path(\"data/numbers/digit.tsv\")))\n+    zero = pynini.invert(pynini.string_file(get_abs_path(\"data/numbers/zero.tsv\")))\n+\n+    graph_symbols = pynini.string_file(get_abs_path(\"data/electronic/symbols.tsv\"))\n+    server_common = pynini.string_file(get_abs_path(\"data/electronic/server_name.tsv\"))\n+    domain_common = pynini.string_file(get_abs_path(\"data/electronic/domain.tsv\"))\n+\n+    PYNINI_AVAILABLE = True\n+\n+except (ModuleNotFoundError, ImportError):\n+    digit_no_zero = None\n+    zero = None\n+\n+    graph_symbols = None\n+    server_common = None\n+    domain_common = None\n+\n+    PYNINI_AVAILABLE = False\n+\n+\n+class ElectronicFst(GraphFst):\n+    \"\"\"\n+    Finite state transducer for verbalizing electronic\n+        e.g. electronic { username: \"abc\" domain: \"hotmail.com\" } -> \"a b c arroba hotmail punto com\"\n+                                                           -> \"a b c arroba h o t m a i l punto c o m\"\n+                                                           -> \"a b c arroba hotmail punto c o m\"\n+                                                           -> \"a b c at h o t m a i l punto com\"\n+    Args:\n+        deterministic: if True will provide a single transduction option,\n+            for False multiple transduction are generated (used for audio-based normalization)\n+    \"\"\"\n+\n+    def __init__(self, deterministic: bool = True):\n+        super().__init__(name=\"electronic\", kind=\"verbalize\", deterministic=deterministic)\n+\n+        graph_digit_no_zero = (\n+            digit_no_zero @ pynini.cdrewrite(pynini.cross(\"un\", \"uno\"), \"\", \"\", NEMO_SIGMA).optimize()\n+        )\n+        graph_digit = graph_digit_no_zero | zero\n+\n+        def add_space_after_char():\n+            return pynini.closure(NEMO_NOT_QUOTE - pynini.accep(\" \") + insert_space) + (\n+                NEMO_NOT_QUOTE - pynini.accep(\" \")\n+            )\n+\n+        verbalize_characters = pynini.cdrewrite(graph_symbols | graph_digit, \"\", \"\", NEMO_SIGMA)\n+\n+        user_name = pynutil.delete(\"username: \\\"\") + add_space_after_char() + pynutil.delete(\"\\\"\")\n+        user_name @= verbalize_characters\n+\n+        convert_defaults = pynutil.add_weight(NEMO_NOT_QUOTE, weight=0.0001) | domain_common | server_common\n+        domain = convert_defaults + pynini.closure(insert_space + convert_defaults)\n+        domain @= verbalize_characters\n+\n+        domain = pynutil.delete(\"domain: \\\"\") + domain + pynutil.delete(\"\\\"\")\n+        protocol = (\n+            pynutil.delete(\"protocol: \\\"\")\n+            + add_space_after_char() @ pynini.cdrewrite(graph_symbols, \"\", \"\", NEMO_SIGMA)\n+            + pynutil.delete(\"\\\"\")\n+        )\n+        self.graph = (pynini.closure(protocol + pynini.accep(\" \"), 0, 1) + domain) | (\n+            user_name + pynini.accep(\" \") + pynutil.insert(\"arroba \") + domain\n+        )\n+        delete_tokens = self.delete_tokens(self.graph + delete_preserve_order)\n+        self.fst = delete_tokens.optimize()\ndiff --git a/nemo_text_processing/text_normalization/es/verbalizers/fraction.py b/nemo_text_processing/text_normalization/es/verbalizers/fraction.py\nnew file mode 100644\n--- /dev/null\n+++ b/nemo_text_processing/text_normalization/es/verbalizers/fraction.py\n@@ -0,0 +1,184 @@\n+# Copyright (c) 2022, NVIDIA CORPORATION & AFFILIATES.  All rights reserved.\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+#     http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+from nemo_text_processing.text_normalization.en.graph_utils import (\n+    NEMO_CHAR,\n+    NEMO_NOT_QUOTE,\n+    NEMO_NOT_SPACE,\n+    NEMO_SIGMA,\n+    NEMO_SPACE,\n+    GraphFst,\n+    delete_space,\n+    insert_space,\n+)\n+from nemo_text_processing.text_normalization.es.graph_utils import (\n+    accents,\n+    shift_cardinal_gender,\n+    strip_cardinal_apocope,\n+)\n+\n+try:\n+    import pynini\n+    from pynini.lib import pynutil\n+\n+    PYNINI_AVAILABLE = True\n+\n+except (ModuleNotFoundError, ImportError):\n+    PYNINI_AVAILABLE = False\n+\n+\n+class FractionFst(GraphFst):\n+    \"\"\"\n+\tFinite state transducer for verbalizing fraction\n+\t\te.g. tokens { fraction { integer: \"treinta y tres\" numerator: \"cuatro\" denominator: \"quinto\" } } ->\n+            treinta y tres y cuatro quintos\n+\t\t\n+\n+\tArgs:\n+\t\tdeterministic: if True will provide a single transduction option,\n+\t\t\tfor False multiple transduction are generated (used for audio-based normalization)\n+\t\"\"\"\n+\n+    def __init__(self, deterministic: bool = True):\n+        super().__init__(name=\"fraction\", kind=\"verbalize\", deterministic=deterministic)\n+\n+        # Derivational strings append 'avo' as a suffix. Adding space for processing aid\n+        fraction_stem = pynutil.insert(\" avo\")\n+        plural = pynutil.insert(\"s\")\n+\n+        integer = (\n+            pynutil.delete(\"integer_part: \\\"\")\n+            + strip_cardinal_apocope(pynini.closure(NEMO_NOT_QUOTE))\n+            + pynutil.delete(\"\\\"\")\n+        )\n+\n+        numerator_one = pynutil.delete(\"numerator: \\\"\") + pynini.accep(\"un\") + pynutil.delete(\"\\\" \")\n+        numerator = (\n+            pynutil.delete(\"numerator: \\\"\")\n+            + pynini.difference(pynini.closure(NEMO_NOT_QUOTE), \"un\")\n+            + pynutil.delete(\"\\\" \")\n+        )\n+\n+        denominator_add_stem = pynutil.delete(\"denominator: \\\"\") + (\n+            pynini.closure(NEMO_NOT_QUOTE)\n+            + fraction_stem\n+            + pynutil.delete(\"\\\" morphosyntactic_features: \\\"add_root\\\"\")\n+        )\n+        denominator_ordinal = pynutil.delete(\"denominator: \\\"\") + (\n+            pynini.closure(NEMO_NOT_QUOTE) + pynutil.delete(\"\\\" morphosyntactic_features: \\\"ordinal\\\"\")\n+        )\n+        denominator_cardinal = pynutil.delete(\"denominator: \\\"\") + (\n+            pynini.closure(NEMO_NOT_QUOTE) + pynutil.delete(\"\\\"\")\n+        )\n+\n+        denominator_singular = pynini.union(denominator_add_stem, denominator_ordinal)\n+        denominator_plural = denominator_singular + plural\n+\n+        if not deterministic:\n+            # Occasional exceptions\n+            denominator_singular |= denominator_add_stem @ pynini.string_map(\n+                [(\"once avo\", \"und\u00e9cimo\"), (\"doce avo\", \"duod\u00e9cimo\")]\n+            )\n+\n+        # Merging operations\n+        merge = pynini.cdrewrite(\n+            pynini.cross(\" y \", \"i\"), \"\", \"\", NEMO_SIGMA\n+        )  # The denominator must be a single word, with the conjunction \"y\" replaced by i\n+        merge @= pynini.cdrewrite(delete_space, \"\", pynini.difference(NEMO_CHAR, \"parte\"), NEMO_SIGMA)\n+\n+        # The merger can produce duplicate vowels. This is not allowed in orthography\n+        delete_duplicates = pynini.string_map([(\"aa\", \"a\"), (\"oo\", \"o\")])  # Removes vowels\n+        delete_duplicates = pynini.cdrewrite(delete_duplicates, \"\", \"\", NEMO_SIGMA)\n+\n+        remove_accents = pynini.cdrewrite(\n+            accents,\n+            pynini.union(NEMO_SPACE, pynini.accep(\"[BOS]\")) + pynini.closure(NEMO_NOT_SPACE),\n+            pynini.closure(NEMO_NOT_SPACE) + pynini.union(\"avo\", \"ava\", \"\u00e9simo\", \"\u00e9sima\"),\n+            NEMO_SIGMA,\n+        )\n+        merge_into_single_word = merge @ remove_accents @ delete_duplicates\n+\n+        fraction_default = numerator + delete_space + insert_space + (denominator_plural @ merge_into_single_word)\n+        fraction_with_one = (\n+            numerator_one + delete_space + insert_space + (denominator_singular @ merge_into_single_word)\n+        )\n+\n+        fraction_with_cardinal = strip_cardinal_apocope(numerator | numerator_one)\n+        fraction_with_cardinal += (\n+            delete_space + pynutil.insert(\" sobre \") + strip_cardinal_apocope(denominator_cardinal)\n+        )\n+\n+        conjunction = pynutil.insert(\" y \")\n+\n+        if not deterministic:\n+            # There is an alternative rendering where ordinals act as adjectives for 'parte'. This requires use of the feminine\n+            # Other rules will manage use of \"un\" at end, so just worry about endings\n+            exceptions = pynini.string_map([(\"tercia\", \"tercera\")])\n+            apply_exceptions = pynini.cdrewrite(exceptions, \"\", \"\", NEMO_SIGMA)\n+            vowel_change = pynini.cdrewrite(pynini.cross(\"o\", \"a\"), \"\", pynini.accep(\"[EOS]\"), NEMO_SIGMA)\n+\n+            denominator_singular_fem = shift_cardinal_gender(denominator_singular) @ vowel_change @ apply_exceptions\n+            denominator_plural_fem = denominator_singular_fem + plural\n+\n+            numerator_one_fem = shift_cardinal_gender(numerator_one)\n+            numerator_fem = shift_cardinal_gender(numerator)\n+\n+            fraction_with_cardinal |= (\n+                (numerator_one_fem | numerator_fem)\n+                + delete_space\n+                + pynutil.insert(\" sobre \")\n+                + shift_cardinal_gender(denominator_cardinal)\n+            )\n+\n+            # Still need to manage stems\n+            merge_stem = pynini.cdrewrite(\n+                delete_space, \"\", pynini.union(\"avo\", \"ava\", \"avos\", \"avas\"), NEMO_SIGMA\n+            )  # For managing alternative spacing\n+            merge_stem @= remove_accents @ delete_duplicates\n+\n+            fraction_with_one_fem = numerator_one_fem + delete_space + insert_space\n+            fraction_with_one_fem += pynini.union(\n+                denominator_singular_fem @ merge_stem, denominator_singular_fem @ merge_into_single_word\n+            )  # Both forms exists\n+            fraction_with_one_fem @= pynini.cdrewrite(pynini.cross(\"una media\", \"media\"), \"\", \"\", NEMO_SIGMA)\n+            fraction_with_one_fem += pynutil.insert(\" parte\")\n+\n+            fraction_default_fem = numerator_fem + delete_space + insert_space\n+            fraction_default_fem += pynini.union(\n+                denominator_plural_fem @ merge_stem, denominator_plural_fem @ merge_into_single_word\n+            )\n+            fraction_default_fem += pynutil.insert(\" partes\")\n+\n+            fraction_default |= (\n+                numerator + delete_space + insert_space + denominator_plural @ merge_stem\n+            )  # Case of no merger\n+            fraction_default |= fraction_default_fem\n+\n+            fraction_with_one |= numerator_one + delete_space + insert_space + denominator_singular @ merge_stem\n+            fraction_with_one |= fraction_with_one_fem\n+\n+            # Integers are influenced by dominant noun, need to allow feminine forms as well\n+            integer |= shift_cardinal_gender(integer)\n+\n+        # Remove 'un medio'\n+        fraction_with_one @= pynini.cdrewrite(pynini.cross(\"un medio\", \"medio\"), \"\", \"\", NEMO_SIGMA)\n+\n+        integer = pynini.closure(integer + delete_space + conjunction, 0, 1)\n+\n+        fraction = fraction_with_one | fraction_default | fraction_with_cardinal\n+\n+        graph = integer + fraction\n+\n+        self.graph = graph\n+        delete_tokens = self.delete_tokens(self.graph)\n+        self.fst = delete_tokens.optimize()\ndiff --git a/nemo_text_processing/text_normalization/es/verbalizers/measure.py b/nemo_text_processing/text_normalization/es/verbalizers/measure.py\nnew file mode 100644\n--- /dev/null\n+++ b/nemo_text_processing/text_normalization/es/verbalizers/measure.py\n@@ -0,0 +1,110 @@\n+# Copyright (c) 2022, NVIDIA CORPORATION & AFFILIATES.  All rights reserved.\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+#     http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+from nemo_text_processing.text_normalization.en.graph_utils import (\n+    NEMO_NOT_QUOTE,\n+    NEMO_SIGMA,\n+    NEMO_SPACE,\n+    NEMO_WHITE_SPACE,\n+    GraphFst,\n+    delete_extra_space,\n+    delete_preserve_order,\n+)\n+from nemo_text_processing.text_normalization.es.graph_utils import ones, shift_cardinal_gender\n+from nemo_text_processing.text_normalization.es.utils import get_abs_path\n+\n+try:\n+    import pynini\n+    from pynini.lib import pynutil\n+\n+    unit_plural_fem = pynini.string_file(get_abs_path(\"data/measures/measurements_plural_fem.tsv\"))\n+    unit_plural_masc = pynini.string_file(get_abs_path(\"data/measures/measurements_plural_masc.tsv\"))\n+\n+    unit_singular_fem = pynini.project(unit_plural_fem, \"input\")\n+    unit_singular_masc = pynini.project(unit_plural_masc, \"input\")\n+\n+    unit_plural_fem = pynini.project(unit_plural_fem, \"output\")\n+    unit_plural_masc = pynini.project(unit_plural_masc, \"output\")\n+\n+    PYNINI_AVAILABLE = True\n+\n+except (ModuleNotFoundError, ImportError):\n+    unit_plural_fem = None\n+    unit_plural_masc = None\n+\n+    unit_singular_fem = None\n+    unit_singular_masc = None\n+\n+    PYNINI_AVAILABLE = False\n+\n+\n+class MeasureFst(GraphFst):\n+    \"\"\"\n+    Finite state transducer for verbalizing measure, e.g.\n+        measure { cardinal { integer: \"dos\" units: \"gramos\" } } -> \"dos gramos\"\n+        measure { cardinal { integer_part: \"dos\" quantity: \"millones\" units: \"gramos\" } } -> \"dos millones de gramos\"\n+\n+    Args:\n+        decimal: DecimalFst\n+        cardinal: CardinalFst\n+        fraction: FractionFst\n+        deterministic: if True will provide a single transduction option,\n+            for False multiple transduction are generated (used for audio-based normalization)\n+    \"\"\"\n+\n+    def __init__(self, decimal: GraphFst, cardinal: GraphFst, fraction: GraphFst, deterministic: bool):\n+        super().__init__(name=\"measure\", kind=\"verbalize\", deterministic=deterministic)\n+\n+        graph_decimal = decimal.fst\n+        graph_cardinal = cardinal.fst\n+        graph_fraction = fraction.fst\n+\n+        unit_masc = (unit_plural_masc | unit_singular_masc) + pynini.closure(\n+            NEMO_WHITE_SPACE + \"por\" + pynini.closure(NEMO_NOT_QUOTE, 1), 0, 1\n+        )\n+        unit_masc |= \"por\" + pynini.closure(NEMO_NOT_QUOTE, 1)\n+        unit_masc = pynutil.delete(\"units: \\\"\") + (pynini.closure(NEMO_NOT_QUOTE) @ unit_masc) + pynutil.delete(\"\\\"\")\n+\n+        unit_fem = (unit_plural_fem | unit_singular_fem) + pynini.closure(\n+            NEMO_WHITE_SPACE + \"por\" + pynini.closure(NEMO_NOT_QUOTE, 1), 0, 1\n+        )\n+        unit_fem = pynutil.delete(\"units: \\\"\") + (pynini.closure(NEMO_NOT_QUOTE) @ unit_fem) + pynutil.delete(\"\\\"\")\n+\n+        graph_masc = (graph_cardinal | graph_decimal | graph_fraction) + NEMO_WHITE_SPACE + unit_masc\n+        graph_fem = (\n+            shift_cardinal_gender(graph_cardinal | graph_decimal | graph_fraction) + NEMO_WHITE_SPACE + unit_fem\n+        )\n+        graph = graph_masc | graph_fem\n+\n+        graph = (\n+            pynini.cdrewrite(\n+                pynutil.insert(\" de\"), \"quantity: \\\"\" + pynini.closure(NEMO_NOT_QUOTE, 1), \"\\\"\", NEMO_SIGMA\n+            )\n+            @ graph\n+        )  # billones de xyz\n+\n+        graph @= pynini.cdrewrite(pynini.cross(ones, \"uno\"), \"\", NEMO_WHITE_SPACE + \"por\", NEMO_SIGMA)\n+\n+        # To manage alphanumeric combonations (\"a-8, 5x\"), we let them use a weighted default path.\n+        alpha_num_unit = pynutil.delete(\"units: \\\"\") + pynini.closure(NEMO_NOT_QUOTE) + pynutil.delete(\"\\\"\")\n+        graph_alpha_num = pynini.union(\n+            (graph_cardinal | graph_decimal) + NEMO_SPACE + alpha_num_unit,\n+            alpha_num_unit + delete_extra_space + (graph_cardinal | graph_decimal),\n+        )\n+\n+        graph |= pynutil.add_weight(graph_alpha_num, 0.01)\n+\n+        graph += delete_preserve_order\n+\n+        delete_tokens = self.delete_tokens(graph)\n+        self.fst = delete_tokens.optimize()\ndiff --git a/nemo_text_processing/text_normalization/es/verbalizers/money.py b/nemo_text_processing/text_normalization/es/verbalizers/money.py\nnew file mode 100644\n--- /dev/null\n+++ b/nemo_text_processing/text_normalization/es/verbalizers/money.py\n@@ -0,0 +1,195 @@\n+# Copyright (c) 2022, NVIDIA CORPORATION & AFFILIATES.  All rights reserved.\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+#     http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+from nemo_text_processing.text_normalization.en.graph_utils import (\n+    NEMO_NOT_QUOTE,\n+    NEMO_SIGMA,\n+    NEMO_SPACE,\n+    GraphFst,\n+    delete_preserve_order,\n+)\n+from nemo_text_processing.text_normalization.es.graph_utils import (\n+    shift_cardinal_gender,\n+    shift_number_gender,\n+    strip_cardinal_apocope,\n+)\n+from nemo_text_processing.text_normalization.es.utils import get_abs_path\n+\n+try:\n+    import pynini\n+    from pynini.lib import pynutil\n+\n+    fem = pynini.string_file((get_abs_path(\"data/money/currency_plural_fem.tsv\")))\n+    masc = pynini.string_file((get_abs_path(\"data/money/currency_plural_masc.tsv\")))\n+\n+    fem_singular = pynini.project(fem, \"input\")\n+    masc_singular = pynini.project(masc, \"input\")\n+\n+    fem_plural = pynini.project(fem, \"output\")\n+    masc_plural = pynini.project(masc, \"output\")\n+\n+    PYNINI_AVAILABLE = True\n+\n+except (ModuleNotFoundError, ImportError):\n+    fem_plural = None\n+    masc_plural = None\n+\n+    fem_singular = None\n+    masc_singular = None\n+\n+    PYNINI_AVAILABLE = False\n+\n+\n+class MoneyFst(GraphFst):\n+    \"\"\"\n+    Finite state transducer for verbalizing money, e.g.\n+        money { currency_maj: \"euro\" integer_part: \"un\"} -> \"un euro\"\n+        money { currency_maj: \"euro\" integer_part: \"un\" fractional_part: \"cero cero un\"} -> \"uno coma cero cero uno euros\"\n+        money { integer_part: \"un\" currency_maj: \"libra\" fractional_part: \"cuarenta\" preserve_order: true} -> \"una libra cuarenta\"\n+        money { integer_part: \"un\" currency_maj: \"libra\" fractional_part: \"cuarenta\" currency_min: \"peniques\" preserve_order: true} -> \"una libra con cuarenta peniques\"\n+        money { fractional_part: \"un\" currency_min: \"penique\" preserve_order: true} -> \"un penique\"\n+\n+    Args:\n+        decimal: GraphFst\n+        deterministic: if True will provide a single transduction option,\n+            for False multiple transduction are generated (used for audio-based normalization)\n+    \"\"\"\n+\n+    def __init__(self, decimal: GraphFst, deterministic: bool = True):\n+        super().__init__(name=\"money\", kind=\"verbalize\", deterministic=deterministic)\n+\n+        maj_singular_masc = (\n+            pynutil.delete(\"currency_maj: \\\"\")\n+            + (pynini.closure(NEMO_NOT_QUOTE, 1) @ masc_singular)\n+            + pynutil.delete(\"\\\"\")\n+        )\n+        maj_singular_fem = (\n+            pynutil.delete(\"currency_maj: \\\"\")\n+            + (pynini.closure(NEMO_NOT_QUOTE, 1) @ fem_singular)\n+            + pynutil.delete(\"\\\"\")\n+        )\n+\n+        maj_plural_masc = (\n+            pynutil.delete(\"currency_maj: \\\"\")\n+            + (pynini.closure(NEMO_NOT_QUOTE, 1) @ masc_plural)\n+            + pynutil.delete(\"\\\"\")\n+        )\n+        maj_plural_fem = (\n+            pynutil.delete(\"currency_maj: \\\"\")\n+            + (pynini.closure(NEMO_NOT_QUOTE, 1) @ fem_plural)\n+            + pynutil.delete(\"\\\"\")\n+        )\n+\n+        maj_masc = maj_plural_masc | maj_singular_masc  # Tagger kept quantity resolution stable\n+        maj_fem = maj_plural_fem | maj_singular_fem\n+\n+        min_singular_masc = (\n+            pynutil.delete(\"currency_min: \\\"\")\n+            + (pynini.closure(NEMO_NOT_QUOTE, 1) @ masc_singular)\n+            + pynutil.delete(\"\\\"\")\n+        )\n+        min_singular_fem = (\n+            pynutil.delete(\"currency_min: \\\"\")\n+            + (pynini.closure(NEMO_NOT_QUOTE, 1) @ fem_singular)\n+            + pynutil.delete(\"\\\"\")\n+        )\n+\n+        min_plural_masc = (\n+            pynutil.delete(\"currency_min: \\\"\")\n+            + (pynini.closure(NEMO_NOT_QUOTE, 1) @ masc_plural)\n+            + pynutil.delete(\"\\\"\")\n+        )\n+        min_plural_fem = (\n+            pynutil.delete(\"currency_min: \\\"\")\n+            + (pynini.closure(NEMO_NOT_QUOTE, 1) @ fem_plural)\n+            + pynutil.delete(\"\\\"\")\n+        )\n+\n+        min_masc = min_plural_masc | min_singular_masc\n+        min_fem = min_plural_fem | min_singular_fem\n+\n+        fractional_part = (\n+            pynutil.delete(\"fractional_part: \\\"\") + pynini.closure(NEMO_NOT_QUOTE, 1) + pynutil.delete(\"\\\"\")\n+        )\n+\n+        integer_part = pynutil.delete(\"integer_part: \\\"\") + pynini.closure(NEMO_NOT_QUOTE, 1) + pynutil.delete(\"\\\"\")\n+        optional_add_and = pynini.closure(pynutil.insert(pynini.union(\"con \", \"y \")), 0, 1)\n+\n+        #  *** currency_maj\n+        graph_integer_masc = integer_part + NEMO_SPACE + maj_masc\n+        graph_integer_fem = shift_cardinal_gender(integer_part) + NEMO_SPACE + maj_fem\n+        graph_integer = graph_integer_fem | graph_integer_masc\n+\n+        #  *** currency_maj + (***) | ((con) *** current_min)\n+        graph_integer_with_minor_masc = (\n+            integer_part\n+            + NEMO_SPACE\n+            + maj_masc\n+            + NEMO_SPACE\n+            + pynini.union(\n+                optional_add_and + strip_cardinal_apocope(fractional_part),\n+                (optional_add_and + fractional_part + NEMO_SPACE + min_masc),\n+                (optional_add_and + shift_cardinal_gender(fractional_part) + NEMO_SPACE + min_fem),\n+            )  # Could be minor currency that is different gender\n+            + delete_preserve_order\n+        )\n+\n+        graph_integer_with_minor_fem = (\n+            shift_cardinal_gender(integer_part)\n+            + NEMO_SPACE\n+            + maj_fem\n+            + NEMO_SPACE\n+            + pynini.union(\n+                optional_add_and + shift_cardinal_gender(fractional_part),\n+                (optional_add_and + fractional_part + NEMO_SPACE + min_masc),\n+                (optional_add_and + shift_cardinal_gender(fractional_part) + NEMO_SPACE + min_fem),\n+            )  # Could be minor currency that is different gender\n+            + delete_preserve_order\n+        )\n+\n+        graph_integer_with_minor = graph_integer_with_minor_fem | graph_integer_with_minor_masc\n+\n+        # *** coma *** currency_maj\n+        graph_decimal_masc = decimal.numbers + NEMO_SPACE + maj_masc\n+\n+        # Need to fix some of the inner parts, so don't use decimal here (note: quantities covered by masc)\n+        graph_decimal_fem = (\n+            pynini.accep(\"integer_part: \\\"\")\n+            + shift_cardinal_gender(pynini.closure(NEMO_NOT_QUOTE, 1))\n+            + pynini.accep(\"\\\"\")\n+            + NEMO_SPACE\n+            + pynini.accep(\"fractional_part: \\\"\")\n+            + shift_number_gender(pynini.closure(NEMO_NOT_QUOTE, 1))\n+            + pynini.accep(\"\\\"\")\n+            + NEMO_SIGMA\n+        )\n+        graph_decimal_fem @= decimal.numbers_no_quantity\n+        graph_decimal_fem += NEMO_SPACE + maj_fem\n+\n+        graph_decimal = graph_decimal_fem | graph_decimal_masc\n+        graph_decimal = (\n+            pynini.cdrewrite(\n+                pynutil.insert(\" de\"), \"quantity: \\\"\" + pynini.closure(NEMO_NOT_QUOTE, 1), \"\\\"\", NEMO_SIGMA\n+            )\n+            @ graph_decimal\n+        )  # formally it's millones/billones de ***\n+\n+        # *** current_min\n+        graph_minor_masc = fractional_part + NEMO_SPACE + min_masc + delete_preserve_order\n+        graph_minor_fem = shift_cardinal_gender(fractional_part) + NEMO_SPACE + min_fem + delete_preserve_order\n+        graph_minor = graph_minor_fem | graph_minor_masc\n+\n+        graph = graph_integer | graph_integer_with_minor | graph_decimal | graph_minor\n+\n+        delete_tokens = self.delete_tokens(graph)\n+        self.fst = delete_tokens.optimize()\ndiff --git a/nemo_text_processing/text_normalization/es/verbalizers/ordinal.py b/nemo_text_processing/text_normalization/es/verbalizers/ordinal.py\nnew file mode 100644\n--- /dev/null\n+++ b/nemo_text_processing/text_normalization/es/verbalizers/ordinal.py\n@@ -0,0 +1,76 @@\n+# Copyright (c) 2022, NVIDIA CORPORATION & AFFILIATES.  All rights reserved.\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+#     http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+from nemo_text_processing.text_normalization.en.graph_utils import NEMO_NOT_QUOTE, NEMO_SIGMA, NEMO_SPACE, GraphFst\n+from nemo_text_processing.text_normalization.es.graph_utils import shift_number_gender\n+\n+try:\n+    import pynini\n+    from pynini.lib import pynutil\n+\n+    PYNINI_AVAILABLE = True\n+\n+except (ModuleNotFoundError, ImportError):\n+    PYNINI_AVAILABLE = False\n+\n+\n+class OrdinalFst(GraphFst):\n+    \"\"\"\n+    Finite state transducer for verbalizing ordinals\n+        e.g. ordinal { integer: \"tercer\" } } -> \"tercero\"\n+                                           -> \"tercera\"\n+\t\t\t\t\t\t\t\t\t\t   -> \"tercer\"\n+\n+    Args:\n+        deterministic: if True will provide a single transduction option,\n+            for False multiple transduction are generated (used for audio-based normalization)\n+    \"\"\"\n+\n+    def __init__(self, deterministic: bool = True):\n+        super().__init__(name=\"ordinal\", kind=\"verbalize\", deterministic=deterministic)\n+\n+        graph = pynutil.delete(\"integer: \\\"\") + pynini.closure(NEMO_NOT_QUOTE, 1) + pynutil.delete(\"\\\"\")\n+\n+        # masculne gender we leave as is\n+        graph_masc = graph + pynutil.delete(\" morphosyntactic_features: \\\"gender_masc\")\n+\n+        # shift gender\n+        graph_fem_ending = graph @ pynini.cdrewrite(\n+            pynini.cross(\"o\", \"a\"), \"\", NEMO_SPACE | pynini.accep(\"[EOS]\"), NEMO_SIGMA\n+        )\n+        graph_fem = shift_number_gender(graph_fem_ending) + pynutil.delete(\" morphosyntactic_features: \\\"gender_fem\")\n+\n+        # Apocope just changes tercero and primero. May occur if someone wrote 11.er (uncommon)\n+        graph_apocope = (\n+            pynini.cross(\"tercero\", \"tercer\")\n+            | pynini.cross(\"primero\", \"primer\")\n+            | pynini.cross(\"und\u00e9cimo\", \"decimoprimer\")\n+        )  # In case someone wrote 11.er with deterministic\n+        graph_apocope = (graph @ pynini.cdrewrite(graph_apocope, \"\", \"\", NEMO_SIGMA)) + pynutil.delete(\n+            \" morphosyntactic_features: \\\"apocope\"\n+        )\n+\n+        graph = graph_apocope | graph_masc | graph_fem\n+\n+        if not deterministic:\n+            # Plural graph\n+            graph_plural = pynini.cdrewrite(\n+                pynutil.insert(\"s\"), pynini.union(\"o\", \"a\"), NEMO_SPACE | pynini.accep(\"[EOS]\"), NEMO_SIGMA\n+            )\n+\n+            graph |= (graph @ graph_plural) + pynutil.delete(\"/plural\")\n+\n+        self.graph = graph + pynutil.delete(\"\\\"\")\n+\n+        delete_tokens = self.delete_tokens(self.graph)\n+        self.fst = delete_tokens.optimize()\ndiff --git a/nemo_text_processing/text_normalization/es/verbalizers/telephone.py b/nemo_text_processing/text_normalization/es/verbalizers/telephone.py\nnew file mode 100644\n--- /dev/null\n+++ b/nemo_text_processing/text_normalization/es/verbalizers/telephone.py\n@@ -0,0 +1,42 @@\n+# Copyright (c) 2022, NVIDIA CORPORATION & AFFILIATES.  All rights reserved.\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+#     http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+from nemo_text_processing.text_normalization.en.graph_utils import NEMO_NOT_QUOTE, GraphFst\n+\n+try:\n+    import pynini\n+    from pynini.lib import pynutil\n+\n+    PYNINI_AVAILABLE = True\n+\n+except (ModuleNotFoundError, ImportError):\n+    PYNINI_AVAILABLE = False\n+\n+\n+class TelephoneFst(GraphFst):\n+    \"\"\"\n+    Finite state transducer for verbalizing telephone, e.g.\n+        telephone { number_part: \"uno dos tres uno dos tres cinco seis siete ocho\" }\n+        -> uno dos tres uno dos tres cinco seis siete ocho\n+\n+    Args:\n+        deterministic: if True will provide a single transduction option,\n+            for False multiple transduction are generated (used for audio-based normalization)\n+    \"\"\"\n+\n+    def __init__(self, deterministic: bool = True):\n+        super().__init__(name=\"telephone\", kind=\"verbalize\")\n+\n+        number_part = pynutil.delete(\"number_part: \\\"\") + pynini.closure(NEMO_NOT_QUOTE, 1) + pynutil.delete(\"\\\"\")\n+        delete_tokens = self.delete_tokens(number_part)\n+        self.fst = delete_tokens.optimize()\ndiff --git a/nemo_text_processing/text_normalization/es/verbalizers/time.py b/nemo_text_processing/text_normalization/es/verbalizers/time.py\nnew file mode 100644\n--- /dev/null\n+++ b/nemo_text_processing/text_normalization/es/verbalizers/time.py\n@@ -0,0 +1,269 @@\n+# Copyright (c) 2022, NVIDIA CORPORATION & AFFILIATES.  All rights reserved.\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+#     http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+from nemo_text_processing.text_normalization.en.graph_utils import (\n+    NEMO_NOT_QUOTE,\n+    NEMO_SIGMA,\n+    GraphFst,\n+    delete_preserve_order,\n+    delete_space,\n+    insert_space,\n+)\n+from nemo_text_processing.text_normalization.es.utils import get_abs_path\n+\n+try:\n+    import pynini\n+    from pynini.lib import pynutil\n+\n+    alt_minutes = pynini.string_file(get_abs_path(\"data/time/alt_minutes.tsv\"))\n+\n+    morning_times = pynini.string_file(get_abs_path(\"data/time/morning_times.tsv\"))\n+    afternoon_times = pynini.string_file(get_abs_path(\"data/time/afternoon_times.tsv\"))\n+    evening_times = pynini.string_file(get_abs_path(\"data/time/evening_times.tsv\"))\n+\n+    PYNINI_AVAILABLE = True\n+\n+except (ModuleNotFoundError, ImportError):\n+    alt_minutes = None\n+\n+    morning_times = None\n+    afternoon_times = None\n+    evening_times = None\n+\n+    PYNINI_AVAILABLE = False\n+\n+\n+class TimeFst(GraphFst):\n+    \"\"\"\n+    Finite state transducer for verbalizing time, e.g.\n+        time { hours: \"doce\" minutes: \"media\" suffix: \"a m\" } -> doce y media de la noche\n+        time { hours: \"doce\" } -> twelve o'clock\n+\n+    Args:\n+        deterministic: if True will provide a single transduction option,\n+            for False multiple transduction are generated (used for audio-based normalization)\n+    \"\"\"\n+\n+    def __init__(self, deterministic: bool = True):\n+        super().__init__(name=\"time\", kind=\"verbalize\", deterministic=deterministic)\n+\n+        change_minutes = pynini.cdrewrite(alt_minutes, pynini.accep(\"[BOS]\"), pynini.accep(\"[EOS]\"), NEMO_SIGMA)\n+\n+        morning_phrases = pynini.cross(\"am\", \"de la ma\u00f1ana\")\n+        afternoon_phrases = pynini.cross(\"pm\", \"de la tarde\")\n+        evening_phrases = pynini.cross(\"pm\", \"de la noche\")\n+\n+        # For the 12's\n+        mid_times = pynini.accep(\"doce\")\n+        mid_phrases = (\n+            pynini.string_map([(\"pm\", \"del mediod\u00eda\"), (\"am\", \"de la noche\")])\n+            if deterministic\n+            else pynini.string_map(\n+                [\n+                    (\"pm\", \"de la ma\u00f1ana\"),\n+                    (\"pm\", \"del d\u00eda\"),\n+                    (\"pm\", \"del mediod\u00eda\"),\n+                    (\"am\", \"de la noche\"),\n+                    (\"am\", \"de la medianoche\"),\n+                ]\n+            )\n+        )\n+\n+        hour = (\n+            pynutil.delete(\"hours:\")\n+            + delete_space\n+            + pynutil.delete(\"\\\"\")\n+            + pynini.closure(NEMO_NOT_QUOTE, 1)\n+            + pynutil.delete(\"\\\"\")\n+        )\n+        minute = (\n+            pynutil.delete(\"minutes:\")\n+            + delete_space\n+            + pynutil.delete(\"\\\"\")\n+            + pynini.closure(NEMO_NOT_QUOTE, 1)\n+            + pynutil.delete(\"\\\"\")\n+        )\n+        minute = (minute @ change_minutes) if deterministic else pynini.union(minute, minute @ change_minutes)\n+\n+        suffix = (\n+            pynutil.delete(\"suffix:\")\n+            + delete_space\n+            + pynutil.delete(\"\\\"\")\n+            + pynini.closure(NEMO_NOT_QUOTE, 1)\n+            + pynutil.delete(\"\\\"\")\n+        )\n+        zone = (\n+            pynutil.delete(\"zone:\")\n+            + delete_space\n+            + pynutil.delete(\"\\\"\")\n+            + pynini.closure(NEMO_NOT_QUOTE, 1)\n+            + pynutil.delete(\"\\\"\")\n+        )\n+        optional_zone = pynini.closure(delete_space + insert_space + zone, 0, 1)\n+        second = (\n+            pynutil.delete(\"seconds:\")\n+            + delete_space\n+            + pynutil.delete(\"\\\"\")\n+            + pynini.closure(NEMO_NOT_QUOTE, 1)\n+            + pynutil.delete(\"\\\"\")\n+        )\n+\n+        graph_hms = (\n+            hour\n+            + pynutil.insert(\" horas \")\n+            + delete_space\n+            + minute\n+            + pynutil.insert(\" minutos y \")\n+            + delete_space\n+            + second\n+            + pynutil.insert(\" segundos\")\n+        )\n+\n+        graph_hm = hour + delete_space + pynutil.insert(\" y \") + minute\n+        graph_hm |= pynini.union(\n+            (hour @ morning_times)\n+            + delete_space\n+            + pynutil.insert(\" y \")\n+            + minute\n+            + delete_space\n+            + insert_space\n+            + (suffix @ morning_phrases),\n+            (hour @ afternoon_times)\n+            + delete_space\n+            + pynutil.insert(\" y \")\n+            + minute\n+            + delete_space\n+            + insert_space\n+            + (suffix @ afternoon_phrases),\n+            (hour @ evening_times)\n+            + delete_space\n+            + pynutil.insert(\" y \")\n+            + minute\n+            + delete_space\n+            + insert_space\n+            + (suffix @ evening_phrases),\n+            (hour @ mid_times)\n+            + delete_space\n+            + pynutil.insert(\" y \")\n+            + minute\n+            + delete_space\n+            + insert_space\n+            + (suffix @ mid_phrases),\n+        )\n+\n+        graph_h = pynini.union(\n+            hour,\n+            (hour @ morning_times) + delete_space + insert_space + (suffix @ morning_phrases),\n+            (hour @ afternoon_times) + delete_space + insert_space + (suffix @ afternoon_phrases),\n+            (hour @ evening_times) + delete_space + insert_space + (suffix @ evening_phrases),\n+            (hour @ mid_times) + delete_space + insert_space + (suffix @ mid_phrases),\n+        )\n+\n+        graph = (graph_hms | graph_hm | graph_h) + optional_zone\n+\n+        if not deterministic:\n+            graph_style_1 = pynutil.delete(\" style: \\\"1\\\"\")\n+            graph_style_2 = pynutil.delete(\" style: \\\"2\\\"\")\n+\n+            graph_menos = hour + delete_space + pynutil.insert(\" menos \") + minute + graph_style_1\n+            graph_menos |= (\n+                (hour @ morning_times)\n+                + delete_space\n+                + pynutil.insert(\" menos \")\n+                + minute\n+                + delete_space\n+                + insert_space\n+                + (suffix @ morning_phrases)\n+                + graph_style_1\n+            )\n+            graph_menos |= (\n+                (hour @ afternoon_times)\n+                + delete_space\n+                + pynutil.insert(\" menos \")\n+                + minute\n+                + delete_space\n+                + insert_space\n+                + (suffix @ afternoon_phrases)\n+                + graph_style_1\n+            )\n+            graph_menos |= (\n+                (hour @ evening_times)\n+                + delete_space\n+                + pynutil.insert(\" menos \")\n+                + minute\n+                + delete_space\n+                + insert_space\n+                + (suffix @ evening_phrases)\n+                + graph_style_1\n+            )\n+            graph_menos |= (\n+                (hour @ mid_times)\n+                + delete_space\n+                + pynutil.insert(\" menos \")\n+                + minute\n+                + delete_space\n+                + insert_space\n+                + (suffix @ mid_phrases)\n+                + graph_style_1\n+            )\n+            graph_menos += optional_zone\n+\n+            graph_para = minute + pynutil.insert(\" para las \") + delete_space + hour + graph_style_2\n+            graph_para |= (\n+                minute\n+                + pynutil.insert(\" para las \")\n+                + delete_space\n+                + (hour @ morning_times)\n+                + delete_space\n+                + insert_space\n+                + (suffix @ morning_phrases)\n+                + graph_style_2\n+            )\n+            graph_para |= (\n+                minute\n+                + pynutil.insert(\" para las \")\n+                + delete_space\n+                + (hour @ afternoon_times)\n+                + delete_space\n+                + insert_space\n+                + (suffix @ afternoon_phrases)\n+                + graph_style_2\n+            )\n+            graph_para |= (\n+                minute\n+                + pynutil.insert(\" para las \")\n+                + delete_space\n+                + (hour @ evening_times)\n+                + delete_space\n+                + insert_space\n+                + (suffix @ evening_phrases)\n+                + graph_style_2\n+            )\n+            graph_para |= (\n+                minute\n+                + pynutil.insert(\" para las \")\n+                + delete_space\n+                + (hour @ mid_times)\n+                + delete_space\n+                + insert_space\n+                + (suffix @ mid_phrases)\n+                + graph_style_2\n+            )\n+            graph_para += optional_zone\n+            graph_para @= pynini.cdrewrite(\n+                pynini.cross(\" las \", \" la \"), \"para\", \"una\", NEMO_SIGMA\n+            )  # Need agreement with one\n+\n+            graph |= graph_menos | graph_para\n+        delete_tokens = self.delete_tokens(graph + delete_preserve_order)\n+        self.fst = delete_tokens.optimize()\ndiff --git a/nemo_text_processing/text_normalization/es/verbalizers/verbalize.py b/nemo_text_processing/text_normalization/es/verbalizers/verbalize.py\nnew file mode 100644\n--- /dev/null\n+++ b/nemo_text_processing/text_normalization/es/verbalizers/verbalize.py\n@@ -0,0 +1,73 @@\n+# Copyright (c) 2022, NVIDIA CORPORATION & AFFILIATES.  All rights reserved.\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+#     http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+from nemo_text_processing.text_normalization.en.graph_utils import GraphFst\n+from nemo_text_processing.text_normalization.en.verbalizers.whitelist import WhiteListFst\n+from nemo_text_processing.text_normalization.es.verbalizers.cardinal import CardinalFst\n+from nemo_text_processing.text_normalization.es.verbalizers.date import DateFst\n+from nemo_text_processing.text_normalization.es.verbalizers.decimals import DecimalFst\n+from nemo_text_processing.text_normalization.es.verbalizers.electronic import ElectronicFst\n+from nemo_text_processing.text_normalization.es.verbalizers.fraction import FractionFst\n+from nemo_text_processing.text_normalization.es.verbalizers.measure import MeasureFst\n+from nemo_text_processing.text_normalization.es.verbalizers.money import MoneyFst\n+from nemo_text_processing.text_normalization.es.verbalizers.ordinal import OrdinalFst\n+from nemo_text_processing.text_normalization.es.verbalizers.telephone import TelephoneFst\n+from nemo_text_processing.text_normalization.es.verbalizers.time import TimeFst\n+\n+\n+class VerbalizeFst(GraphFst):\n+    \"\"\"\n+    Composes other verbalizer grammars.\n+    For deployment, this grammar will be compiled and exported to OpenFst Finate State Archiv (FAR) File.\n+    More details to deployment at NeMo/tools/text_processing_deployment.\n+\n+    Args:\n+        deterministic: if True will provide a single transduction option,\n+            for False multiple options (used for audio-based normalization)\n+    \"\"\"\n+\n+    def __init__(self, deterministic: bool = True):\n+        super().__init__(name=\"verbalize\", kind=\"verbalize\", deterministic=deterministic)\n+        cardinal = CardinalFst(deterministic=deterministic)\n+        cardinal_graph = cardinal.fst\n+        ordinal = OrdinalFst(deterministic=deterministic)\n+        ordinal_graph = ordinal.fst\n+        decimal = DecimalFst(deterministic=deterministic)\n+        decimal_graph = decimal.fst\n+        fraction = FractionFst(deterministic=deterministic)\n+        fraction_graph = fraction.fst\n+        date = DateFst(deterministic=deterministic)\n+        date_graph = date.fst\n+        measure = MeasureFst(cardinal=cardinal, decimal=decimal, fraction=fraction, deterministic=deterministic)\n+        measure_graph = measure.fst\n+        electronic = ElectronicFst(deterministic=deterministic)\n+        electronic_graph = electronic.fst\n+        whitelist_graph = WhiteListFst(deterministic=deterministic).fst\n+        money_graph = MoneyFst(decimal=decimal, deterministic=deterministic).fst\n+        telephone_graph = TelephoneFst(deterministic=deterministic).fst\n+        time_graph = TimeFst(deterministic=deterministic).fst\n+\n+        graph = (\n+            cardinal_graph\n+            | measure_graph\n+            | decimal_graph\n+            | ordinal_graph\n+            | date_graph\n+            | electronic_graph\n+            | money_graph\n+            | fraction_graph\n+            | whitelist_graph\n+            | telephone_graph\n+            | time_graph\n+        )\n+        self.fst = graph\ndiff --git a/nemo_text_processing/text_normalization/es/verbalizers/verbalize_final.py b/nemo_text_processing/text_normalization/es/verbalizers/verbalize_final.py\nnew file mode 100644\n--- /dev/null\n+++ b/nemo_text_processing/text_normalization/es/verbalizers/verbalize_final.py\n@@ -0,0 +1,52 @@\n+# Copyright (c) 2022, NVIDIA CORPORATION & AFFILIATES.  All rights reserved.\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+#     http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+from nemo_text_processing.text_normalization.en.graph_utils import GraphFst, delete_extra_space, delete_space\n+from nemo_text_processing.text_normalization.en.verbalizers.word import WordFst\n+from nemo_text_processing.text_normalization.es.verbalizers.verbalize import VerbalizeFst\n+\n+try:\n+    import pynini\n+    from pynini.lib import pynutil\n+\n+    PYNINI_AVAILABLE = True\n+\n+except (ModuleNotFoundError, ImportError):\n+    PYNINI_AVAILABLE = False\n+\n+\n+class VerbalizeFinalFst(GraphFst):\n+    \"\"\"\n+    Finite state transducer that verbalizes an entire sentence\n+\n+    Args:\n+        deterministic: if True will provide a single transduction option,\n+            for False multiple options (used for audio-based normalization)\n+    \"\"\"\n+\n+    def __init__(self, deterministic: bool = True):\n+        super().__init__(name=\"verbalize_final\", kind=\"verbalize\", deterministic=deterministic)\n+        verbalize = VerbalizeFst(deterministic=deterministic).fst\n+        word = WordFst(deterministic=deterministic).fst\n+        types = verbalize | word\n+        graph = (\n+            pynutil.delete(\"tokens\")\n+            + delete_space\n+            + pynutil.delete(\"{\")\n+            + delete_space\n+            + types\n+            + delete_space\n+            + pynutil.delete(\"}\")\n+        )\n+        graph = delete_space + pynini.closure(graph + delete_extra_space) + graph + delete_space\n+        self.fst = graph\ndiff --git a/nemo_text_processing/text_normalization/normalize.py b/nemo_text_processing/text_normalization/normalize.py\n--- a/nemo_text_processing/text_normalization/normalize.py\n+++ b/nemo_text_processing/text_normalization/normalize.py\n@@ -46,8 +46,8 @@\n \n class Normalizer:\n     \"\"\"\n-    Normalizer class that converts text from written to spoken form. \n-    Useful for TTS preprocessing. \n+    Normalizer class that converts text from written to spoken form.\n+    Useful for TTS preprocessing.\n \n     Args:\n         input_case: expected input capitalization\n@@ -83,10 +83,11 @@ def __init__(\n             from nemo_text_processing.text_normalization.ru.taggers.tokenize_and_classify import ClassifyFst\n             from nemo_text_processing.text_normalization.ru.verbalizers.verbalize_final import VerbalizeFinalFst\n         elif lang == 'de':\n-            # Ru TN only support non-deterministic cases and produces multiple normalization options\n-            # use normalize_with_audio.py\n             from nemo_text_processing.text_normalization.de.taggers.tokenize_and_classify import ClassifyFst\n             from nemo_text_processing.text_normalization.de.verbalizers.verbalize_final import VerbalizeFinalFst\n+        elif lang == 'es':\n+            from nemo_text_processing.text_normalization.es.taggers.tokenize_and_classify import ClassifyFst\n+            from nemo_text_processing.text_normalization.es.verbalizers.verbalize_final import VerbalizeFinalFst\n         self.tagger = ClassifyFst(\n             input_case=input_case,\n             deterministic=deterministic,\n@@ -106,7 +107,7 @@ def __init__(\n \n     def normalize_list(self, texts: List[str], verbose=False, punct_post_process: bool = False) -> List[str]:\n         \"\"\"\n-        NeMo text normalizer \n+        NeMo text normalizer\n \n         Args:\n             texts: list of input strings\n@@ -357,7 +358,7 @@ def select_verbalizer(self, lattice: 'pynini.FstLike') -> str:\n def parse_args():\n     parser = ArgumentParser()\n     parser.add_argument(\"input_string\", help=\"input string\", type=str)\n-    parser.add_argument(\"--language\", help=\"language\", choices=[\"en\", \"de\"], default=\"en\", type=str)\n+    parser.add_argument(\"--language\", help=\"language\", choices=[\"en\", \"de\", \"es\"], default=\"en\", type=str)\n     parser.add_argument(\n         \"--input_case\", help=\"input capitalization\", choices=[\"lower_cased\", \"cased\"], default=\"cased\", type=str\n     )\ndiff --git a/nemo_text_processing/text_normalization/normalize_with_audio.py b/nemo_text_processing/text_normalization/normalize_with_audio.py\n--- a/nemo_text_processing/text_normalization/normalize_with_audio.py\n+++ b/nemo_text_processing/text_normalization/normalize_with_audio.py\n@@ -55,15 +55,15 @@\n     \"audio_data\" - path to the audio file\n     \"text\" - raw text\n     \"pred_text\" - ASR model prediction\n-    \n+\n     See https://github.com/NVIDIA/NeMo/blob/main/examples/asr/transcribe_speech.py on how to add ASR predictions\n-        \n+\n     When the manifest is ready, run:\n         python normalize_with_audio.py \\\n                --audio_data PATH/TO/MANIFEST.JSON \\\n-               --language en \n-     \n-        \n+               --language en\n+\n+\n To run with a single audio file, specify path to audio and text with:\n     python normalize_with_audio.py \\\n            --audio_data PATH/TO/AUDIO.WAV \\\n@@ -71,18 +71,18 @@\n            --text raw text OR PATH/TO/.TXT/FILE\n            --model QuartzNet15x5Base-En \\\n            --verbose\n-    \n+\n To see possible normalization options for a text input without an audio file (could be used for debugging), run:\n     python python normalize_with_audio.py --text \"RAW TEXT\"\n-    \n+\n Specify `--cache_dir` to generate .far grammars once and re-used them for faster inference\n \"\"\"\n \n \n class NormalizerWithAudio(Normalizer):\n     \"\"\"\n-    Normalizer class that converts text from written to spoken form. \n-    Useful for TTS preprocessing. \n+    Normalizer class that converts text from written to spoken form.\n+    Useful for TTS preprocessing.\n \n     Args:\n         input_case: expected input capitalization\n@@ -282,7 +282,7 @@ def parse_args():\n         \"--input_case\", help=\"input capitalization\", choices=[\"lower_cased\", \"cased\"], default=\"cased\", type=str\n     )\n     parser.add_argument(\n-        \"--language\", help=\"Select target language\", choices=[\"en\", \"ru\", \"de\"], default=\"en\", type=str\n+        \"--language\", help=\"Select target language\", choices=[\"en\", \"ru\", \"de\", \"es\"], default=\"en\", type=str\n     )\n     parser.add_argument(\"--audio_data\", default=None, help=\"path to an audio file or .json manifest\")\n     parser.add_argument(\ndiff --git a/tools/text_processing_deployment/pynini_export.py b/tools/text_processing_deployment/pynini_export.py\n--- a/tools/text_processing_deployment/pynini_export.py\n+++ b/tools/text_processing_deployment/pynini_export.py\n@@ -67,7 +67,7 @@ def tn_grammars(**kwargs):\n \n def export_grammars(output_dir, grammars):\n     \"\"\"\n-    Exports tokenizer_and_classify and verbalize Fsts as OpenFst finite state archive (FAR) files. \n+    Exports tokenizer_and_classify and verbalize Fsts as OpenFst finite state archive (FAR) files.\n \n     Args:\n         output_dir: directory to export FAR files to. Subdirectories will be created for tagger and verbalizer respectively.\n@@ -109,7 +109,7 @@ def parse_args():\n if __name__ == '__main__':\n     args = parse_args()\n \n-    if args.language in ['ru', 'fr', 'es', 'vi'] and args.grammars == 'tn_grammars':\n+    if args.language in ['ru', 'fr', 'vi'] and args.grammars == 'tn_grammars':\n         raise ValueError('Only ITN grammars could be deployed in Sparrowhawk for the selected languages.')\n \n     if args.language == 'en':\n@@ -148,6 +148,10 @@ def parse_args():\n         from nemo_text_processing.inverse_text_normalization.es.verbalizers.verbalize import (\n             VerbalizeFst as ITNVerbalizeFst,\n         )\n+        from nemo_text_processing.text_normalization.es.taggers.tokenize_and_classify import (\n+            ClassifyFst as TNClassifyFst,\n+        )\n+        from nemo_text_processing.text_normalization.es.verbalizers.verbalize import VerbalizeFst as TNVerbalizeFst\n     elif args.language == 'fr':\n         from nemo_text_processing.inverse_text_normalization.fr.taggers.tokenize_and_classify import (\n             ClassifyFst as ITNClassifyFst,\n",
-    "test_patch": "diff --git a/tests/nemo_text_processing/es/data_text_normalization/test_cases_cardinal.txt b/tests/nemo_text_processing/es/data_text_normalization/test_cases_cardinal.txt\nnew file mode 100644\n--- /dev/null\n+++ b/tests/nemo_text_processing/es/data_text_normalization/test_cases_cardinal.txt\n@@ -0,0 +1,86 @@\n+1~un\n+2~dos\n+3~tres\n+4~cuatro\n+5~cinco\n+6~seis\n+7~siete\n+8~ocho\n+9~nueve\n+10~diez\n+11~once\n+12~doce\n+13~trece\n+14~catorce\n+15~quince\n+16~diecis\u00e9is\n+17~diecisiete\n+18~dieciocho\n+19~diecinueve\n+20~veinte\n+21~veinti\u00fan\n+22~veintid\u00f3s\n+23~veintitr\u00e9s\n+24~veinticuatro\n+25~veinticinco\n+26~veintis\u00e9is\n+27~veintisiete\n+28~veintiocho\n+29~veintinueve\n+30~treinta\n+31~treinta y un\n+40~cuarenta\n+41~cuarenta y un\n+50~cincuenta\n+51~cincuenta y un\n+60~sesenta\n+70~setenta\n+80~ochenta\n+90~noventa\n+100~cien\n+101~ciento un\n+120~ciento veinte\n+121~ciento veinti\u00fan\n+130~ciento treinta\n+131~ciento treinta y un\n+200~doscientos\n+201~doscientos un\n+300~trescientos\n+301~trescientos un\n+1000~mil\n+1 000~mil\n+1.000~mil\n+1001~mil un\n+1010~mil diez\n+1020~mil veinte\n+1021~mil veinti\u00fan\n+1100~mil cien\n+1101~mil ciento un\n+1110~mil ciento diez\n+1111~mil ciento once\n+1234~mil doscientos treinta y cuatro\n+2000~dos mil\n+2001~dos mil un\n+2010~dos mil diez\n+2020~dos mil veinte\n+2100~dos mil cien\n+2101~dos mil ciento un\n+2110~dos mil ciento diez\n+2111~dos mil ciento once\n+2222~dos mil doscientos veintid\u00f3s\n+10000~diez mil\n+10 000~diez mil\n+10.000~diez mil\n+100000~cien mil\n+100 000~cien mil\n+100.000~cien mil\n+1 000 000~un mill\u00f3n\n+1.000.000~un mill\u00f3n\n+1 234 568~un mill\u00f3n doscientos treinta y cuatro mil quinientos sesenta y ocho\n+2.000.000~dos millones\n+1.000.000.000~mil millones\n+2.000.000.000~dos mil millones\n+3 000 000 000 000~tres billones\n+3.000.000.000.000~tres billones\n+100 000 000 000 000 000 000 000~cien mil trillones\n+100 000 000 000 000 000 000 001~cien mil trillones un\ndiff --git a/tests/nemo_text_processing/es/data_text_normalization/test_cases_date.txt b/tests/nemo_text_processing/es/data_text_normalization/test_cases_date.txt\nnew file mode 100644\n--- /dev/null\n+++ b/tests/nemo_text_processing/es/data_text_normalization/test_cases_date.txt\n@@ -0,0 +1,13 @@\n+1 enero~primero de enero\n+5 febrero~cinco de febrero\n+20 de marzo~veinte de marzo\n+abril 30~treinta de abril\n+31 marzo~treinta y uno de marzo\n+10 mayo 1990~diez de mayo de mil novecientos noventa\n+junio 11 2000~once de junio de dos mil\n+30 julio del 2020~treinta de julio del dos mil veinte\n+30-2-1990~treinta de febrero de mil novecientos noventa\n+30/2/1990~treinta de febrero de mil novecientos noventa\n+30.2.1990~treinta de febrero de mil novecientos noventa\n+1990-2-30~treinta de febrero de mil novecientos noventa\n+1990-02-30~treinta de febrero de mil novecientos noventa\n\\ No newline at end of file\ndiff --git a/tests/nemo_text_processing/es/data_text_normalization/test_cases_decimal.txt b/tests/nemo_text_processing/es/data_text_normalization/test_cases_decimal.txt\nnew file mode 100644\n--- /dev/null\n+++ b/tests/nemo_text_processing/es/data_text_normalization/test_cases_decimal.txt\n@@ -0,0 +1,27 @@\n+0,1~cero coma un\n+0,01~cero coma cero un\n+0,010~cero coma cero uno cero\n+1,0101~uno coma cero uno cero un\n+0,0~cero coma cero\n+1,0~uno coma cero\n+1,00~uno coma cero cero\n+1,1~uno coma un\n+233,32~doscientos treinta y tres coma treinta y dos\n+32,22 millones~treinta y dos coma veintid\u00f3s millones\n+320 320,22 millones~trescientos veinte mil trescientos veinte coma veintid\u00f3s millones\n+5.002,232~cinco mil dos coma doscientos treinta y dos\n+3,2 trillones~tres coma dos trillones\n+3 millones~tres millones\n+3 000 millones~tres mil millones\n+3000 millones~tres mil millones\n+3.000 millones~tres mil millones\n+3.001 millones~tres mil un millones\n+1 mill\u00f3n~un mill\u00f3n\n+1 000 millones~mil millones\n+1000 millones~mil millones\n+1.000 millones~mil millones\n+2,33302 millones~dos coma tres tres tres cero dos millones\n+1,5332 mill\u00f3n~uno coma cinco tres tres dos mill\u00f3n\n+1,53322 mill\u00f3n~uno coma cinco tres tres dos dos mill\u00f3n\n+1,53321 mill\u00f3n~uno coma cinco tres tres dos un mill\u00f3n\n+101,010101 millones~ciento uno coma cero uno cero uno cero un millones\n\\ No newline at end of file\ndiff --git a/tests/nemo_text_processing/es/data_text_normalization/test_cases_electronic.txt b/tests/nemo_text_processing/es/data_text_normalization/test_cases_electronic.txt\nnew file mode 100644\n--- /dev/null\n+++ b/tests/nemo_text_processing/es/data_text_normalization/test_cases_electronic.txt\n@@ -0,0 +1,12 @@\n+a.bc@gmail.com~a punto b c arroba gmail punto com\n+cdf@abc.edu~c d f arroba a b c punto e d u\n+abc@gmail.abc~a b c arroba gmail punto a b c\n+abc@abc.com~a b c arroba a b c punto com\n+asdf123@abc.com~a s d f uno dos tres arroba a b c punto com\n+a1b2@abc.com~a uno b dos arroba a b c punto com\n+ab3.sdd.3@gmail.com~a b tres punto s d d punto tres arroba gmail punto com\n+https://www.nvidia.com~h t t p s dos puntos barra barra w w w punto nvidia punto com\n+www.nvidia.com~w w w punto nvidia punto com\n+www.abc.es/efg~w w w punto a b c punto es barra e f g\n+www.abc.es~w w w punto a b c punto es\n+http://www.ourdailynews.com.sm~h t t p dos puntos barra barra w w w punto o u r d a i l y n e w s punto com punto s m\n\\ No newline at end of file\ndiff --git a/tests/nemo_text_processing/es/data_text_normalization/test_cases_fraction.txt b/tests/nemo_text_processing/es/data_text_normalization/test_cases_fraction.txt\nnew file mode 100644\n--- /dev/null\n+++ b/tests/nemo_text_processing/es/data_text_normalization/test_cases_fraction.txt\n@@ -0,0 +1,76 @@\n+1/2~medio\n+1 1/2~uno y medio\n+3/2~tres medios\n+1 3/2~uno y tres medios\n+1/3~un tercio\n+2/3~dos tercios\n+1/4~un cuarto\n+2/4~dos cuartos\n+1/5~un quinto\n+2/5~dos quintos\n+1/6~un sexto\n+2/6~dos sextos\n+1/7~un s\u00e9ptimo\n+2/7~dos s\u00e9ptimos\n+1/8~un octavo\n+2/8~dos octavos\n+1/9~un noveno\n+2/9~dos novenos\n+1/10~un d\u00e9cimo\n+2/10~dos d\u00e9cimos\n+1/11~un onceavo\n+1/12~un doceavo\n+1/13~un treceavo\n+1/14~un catorceavo\n+1/15~un quinceavo\n+1/16~un dieciseisavo\n+1/17~un diecisieteavo\n+1/18~un dieciochoavo\n+1/19~un diecinueveavo\n+1/20~un veinteavo\n+1/21~un veintiunavo\n+1/22~un veintidosavo\n+1/30~un treintavo\n+1/31~un treintaiunavo\n+1/40~un cuarentavo\n+1/41~un cuarentaiunavo\n+1/50~un cincuentavo\n+1/60~un sesentavo\n+1/70~un setentavo\n+1/80~un ochentavo\n+1/90~un noventavo\n+1/100~un cent\u00e9simo\n+2/100~dos cent\u00e9simos\n+1 2/100~uno y dos cent\u00e9simos\n+1/101~uno sobre ciento uno\n+1/110~uno sobre ciento diez\n+1/111~uno sobre ciento once\n+1/112~uno sobre ciento doce\n+1/123~uno sobre ciento veintitr\u00e9s\n+1/134~uno sobre ciento treinta y cuatro\n+1/200~un ducent\u00e9simo\n+1/201~uno sobre doscientos uno\n+1/234~uno sobre doscientos treinta y cuatro\n+1/300~un tricent\u00e9simo\n+1/345~uno sobre trescientos cuarenta y cinco\n+1/400~un cuadringent\u00e9simo\n+1/456~uno sobre cuatrocientos cincuenta y seis\n+1/500~un quingent\u00e9simo\n+1/600~un sexcent\u00e9simo\n+1/700~un septingent\u00e9simo\n+1/800~un octingent\u00e9simo\n+1/900~un noningent\u00e9simo\n+1/1000~un mil\u00e9simo\n+2/1000~dos mil\u00e9simos\n+1 2/1000~uno y dos mil\u00e9simos\n+1/1001~uno sobre mil uno\n+1/1100~uno sobre mil cien\n+1/1200~uno sobre mil doscientos\n+1/1234~uno sobre mil doscientos treinta y cuatro\n+1/2000~un dosmil\u00e9simo\n+1/5000~un cincomil\u00e9simo\n+1/10000~un diezmil\u00e9simo\n+1/100.000~un cienmil\u00e9simo\n+1/1.000.000~un millon\u00e9simo\n+1/100.000.000~un cienmillon\u00e9simo\n+1/1.200.000.000~un mildoscientosmillon\u00e9simo\n\\ No newline at end of file\ndiff --git a/tests/nemo_text_processing/es/data_text_normalization/test_cases_measure.txt b/tests/nemo_text_processing/es/data_text_normalization/test_cases_measure.txt\nnew file mode 100644\n--- /dev/null\n+++ b/tests/nemo_text_processing/es/data_text_normalization/test_cases_measure.txt\n@@ -0,0 +1,17 @@\n+1,2-a~uno coma dos a\n+a-5~a cinco\n+200 m~doscientos metros\n+3 h~tres horas\n+1 h~una hora\n+245 mph~doscientas cuarenta y cinco millas por hora\n+2 kg~dos kilogramos\n+60,2400 kg~sesenta coma dos cuatro cero cero kilogramos\n+-60,2400 kg~menos sesenta coma dos cuatro cero cero kilogramos\n+8,52 %~ocho coma cincuenta y dos por ciento\n+-8,52 %~menos ocho coma cincuenta y dos por ciento\n+1 %~uno por ciento\n+3 cm~tres cent\u00edmetros\n+4 s~cuatro segundos\n+5 l~cinco litros\n+4,51/s~cuatro coma cincuenta y uno por segundo\n+0,0101 s~cero coma cero uno cero un segundos\n\\ No newline at end of file\ndiff --git a/tests/nemo_text_processing/es/data_text_normalization/test_cases_money.txt b/tests/nemo_text_processing/es/data_text_normalization/test_cases_money.txt\nnew file mode 100644\n--- /dev/null\n+++ b/tests/nemo_text_processing/es/data_text_normalization/test_cases_money.txt\n@@ -0,0 +1,24 @@\n+$1~un d\u00f3lar\n+1 $~un d\u00f3lar\n+$1,50~un d\u00f3lar cincuenta centavos\n+1,50 $~un d\u00f3lar cincuenta centavos\n+\u00a3200.000.001~doscientos millones una libras\n+200.000.001 \u00a3~doscientos millones una libras\n+2 billones de euros~dos billones de euros\n+\u20ac2 billones~dos billones de euros\n+\u20ac 2 billones~dos billones de euros\n+\u20ac 2,3 billones~dos coma tres billones de euros\n+2,3 billones de euros~dos coma tres billones de euros\n+\u20ac5,50~cinco euros cincuenta c\u00e9ntimos\n+5,50 \u20ac~cinco euros cincuenta c\u00e9ntimos\n+5,01 \u20ac~cinco euros un c\u00e9ntimo\n+5,01 \u00a3~cinco libras un penique\n+21 czk~veintiuna coronas checas\n+czk21~veintiuna coronas checas\n+czk21,1 millones~veintiuna coma una millones de coronas checas\n+czk 5,50 billones~cinco coma cincuenta billones de coronas checas\n+rs 5,50 billones~cinco coma cincuenta billones de rupias\n+czk5,50 billones~cinco coma cincuenta billones de coronas checas\n+0,55 $~cincuenta y cinco centavos\n+1,01 $~un d\u00f3lar un centavo\n+\u00a512,05~doce yenes cinco centavos\n\\ No newline at end of file\ndiff --git a/tests/nemo_text_processing/es/data_text_normalization/test_cases_normalize_with_audio.txt b/tests/nemo_text_processing/es/data_text_normalization/test_cases_normalize_with_audio.txt\nnew file mode 100644\n--- /dev/null\n+++ b/tests/nemo_text_processing/es/data_text_normalization/test_cases_normalize_with_audio.txt\n@@ -0,0 +1,120 @@\n+~121\n+ciento veinti\u00fan\n+ciento veintiuno\n+ciento veintiuna\n+121\n+~200\n+doscientos\n+doscientas\n+200\n+~201\n+doscientos un\n+doscientos uno\n+doscientas una\n+201\n+~1\n+un\n+uno\n+una\n+1\n+~550.000.001\n+quinientos cincuenta millones un\n+quinientos cincuenta millones una\n+quinientos cincuenta millones uno\n+550.000.001\n+~500.501\n+quinientos mil quinientos un\n+quinientos mil quinientos uno\n+quinientas mil quinientas una\n+500.501\n+~500.001.\u00ba\n+quinientosmil\u00e9simo primero\n+quingent\u00e9simo mil\u00e9simo primero\n+quinientosmil\u00e9simos primeros\n+quingent\u00e9simos mil\u00e9simos primeros\n+500.001.\u00ba\n+~500.001.\u00aa\n+quinientasmil\u00e9sima primera\n+quingent\u00e9sima mil\u00e9sima primera\n+quinientasmil\u00e9simas primeras\n+quingent\u00e9simas mil\u00e9simas primeras\n+500.001.\u00aa\n+~11.\u00aa\n+d\u00e9cima primera\n+decimoprimera\n+d\u00e9cimas primeras\n+decimoprimeras\n+und\u00e9cima\n+und\u00e9cimas\n+11.\u00aa\n+~11.\u00ba\n+d\u00e9cimo primero\n+decimoprimero\n+d\u00e9cimos primeros\n+decimoprimeros\n+und\u00e9cimo\n+und\u00e9cimos\n+11.\u00ba\n+~12.\u00ba\n+d\u00e9cimo segundo\n+decimosegundo\n+d\u00e9cimos segundos\n+decimosegundos\n+duod\u00e9cimo\n+duod\u00e9cimos\n+12.\u00ba\n+~200,0101\n+doscientos coma cero uno cero un\n+doscientos coma cero uno cero uno\n+doscientas coma cero una cero una\n+200,0101\n+~1.000.200,21\n+un mill\u00f3n doscientos coma veinti\u00fan\n+un mill\u00f3n doscientos coma veintiuno\n+un mill\u00f3n doscientas coma veintiuna\n+un mill\u00f3n doscientos coma dos un\n+un mill\u00f3n doscientos coma dos uno\n+un mill\u00f3n doscientas coma dos una\n+1.000.200,21\n+~1/12\n+un doceavo\n+una doceava parte\n+un duod\u00e9cimo\n+una duod\u00e9cima parte\n+uno sobre doce\n+1/12\n+~5/200\n+cinco ducent\u00e9simos\n+cinco ducent\u00e9simas partes\n+cinco sobre doscientos\n+5/200\n+~1 5/3\n+uno y cinco tercios\n+una y cinco terceras partes\n+uno y cinco sobre tres\n+una y cinco sobre tres\n+~1/5/2020\n+primero de mayo de dos mil veinte\n+uno de mayo de dos mil veinte\n+cinco de enero de dos mil veinte\n+~$5,50\n+cinco d\u00f3lares con cincuenta\n+cinco d\u00f3lares y cincuenta\n+cinco d\u00f3lares cincuenta\n+cinco d\u00f3lares con cincuenta centavos\n+cinco d\u00f3lares y cincuenta centavos\n+cinco d\u00f3lares cincuenta centavos\n+~2.30 h\n+dos y treinta\n+dos y media\n+tres menos treinta\n+tres menos media\n+treinta para las tres\n+~12.30 a.m.\n+doce y treinta de la medianoche\n+doce y treinta de la noche\n+doce y media de la medianoche\n+doce y media de la noche\n+una menos treinta de la ma\u00f1ana\n+una menos media de la ma\u00f1ana\n+treinta para la una de la ma\u00f1ana\n\\ No newline at end of file\ndiff --git a/tests/nemo_text_processing/es/data_text_normalization/test_cases_ordinal.txt b/tests/nemo_text_processing/es/data_text_normalization/test_cases_ordinal.txt\nnew file mode 100644\n--- /dev/null\n+++ b/tests/nemo_text_processing/es/data_text_normalization/test_cases_ordinal.txt\n@@ -0,0 +1,137 @@\n+1.\u1d49\u02b3~primer\n+1.\u00ba~primero\n+1.\u00aa~primera\n+2.\u00ba~segundo\n+2.\u00aa~segunda\n+ii~segundo\n+II~segundo\n+3.\u1d49\u02b3~tercer\n+3.\u00ba~tercero\n+3.\u00aa~tercera\n+4.\u00ba~cuarto\n+4.\u00aa~cuarta\n+5.\u00ba~quinto\n+5.\u00aa~quinta\n+6.\u00ba~sexto\n+6.\u00aa~sexta\n+7.\u00ba~s\u00e9ptimo\n+7.\u00aa~s\u00e9ptima\n+8.\u00ba~octavo\n+8.\u00aa~octava\n+9.\u00ba~noveno\n+9.\u00aa~novena\n+10.\u00ba~d\u00e9cimo\n+10.\u00aa~d\u00e9cima\n+11.\u1d49\u02b3~decimoprimer\n+11.\u00ba~und\u00e9cimo\n+11.\u00aa~und\u00e9cima\n+12.\u00ba~duod\u00e9cimo\n+12.\u00aa~duod\u00e9cima\n+13.\u1d49\u02b3~decimotercer\n+13.\u00ba~decimotercero\n+13.\u00aa~decimotercera\n+14.\u00ba~decimocuarto\n+14.\u00aa~decimocuarta\n+15.\u00ba~decimoquinto\n+15.\u00aa~decimoquinta\n+16.\u00ba~decimosexto\n+16.\u00aa~decimosexta\n+17.\u00ba~decimos\u00e9ptimo\n+17.\u00aa~decimos\u00e9ptima\n+18.\u00ba~decimoctavo\n+18.\u00aa~decimoctava\n+19.\u00ba~decimonoveno\n+19.\u00aa~decimonovena\n+20.\u00ba~vig\u00e9simo\n+20.\u00aa~vig\u00e9sima\n+21.\u1d49\u02b3~vigesimoprimer\n+21.\u00ba~vigesimoprimero\n+21.\u00aa~vigesimoprimera\n+30.\u00ba~trig\u00e9simo\n+30.\u00aa~trig\u00e9sima\n+31.\u1d49\u02b3~trig\u00e9simo primer\n+31.\u00ba~trig\u00e9simo primero\n+31.\u00aa~trig\u00e9sima primera\n+40.\u00ba~cuadrag\u00e9simo\n+40.\u00aa~cuadrag\u00e9sima\n+41.\u1d49\u02b3~cuadrag\u00e9simo primer\n+41.\u00ba~cuadrag\u00e9simo primero\n+41.\u00aa~cuadrag\u00e9sima primera\n+50.\u00ba~quincuag\u00e9simo\n+50.\u00aa~quincuag\u00e9sima\n+51.\u1d49\u02b3~quincuag\u00e9simo primer\n+51.\u00ba~quincuag\u00e9simo primero\n+51.\u00aa~quincuag\u00e9sima primera\n+60.\u00ba~sexag\u00e9simo\n+60.\u00aa~sexag\u00e9sima\n+70.\u00ba~septuag\u00e9simo\n+70.\u00aa~septuag\u00e9sima\n+80.\u00ba~octog\u00e9simo\n+80.\u00aa~octog\u00e9sima\n+90.\u00ba~nonag\u00e9simo\n+90.\u00aa~nonag\u00e9sima\n+100.\u00ba~cent\u00e9simo\n+100.\u00aa~cent\u00e9sima\n+101.\u1d49\u02b3~cent\u00e9simo primer\n+101.\u00ba~cent\u00e9simo primero\n+101.\u00aa~cent\u00e9sima primera\n+134.\u00ba~cent\u00e9simo trig\u00e9simo cuarto\n+134.\u00aa~cent\u00e9sima trig\u00e9sima cuarta\n+200.\u00ba~ducent\u00e9simo\n+200.\u00aa~ducent\u00e9sima\n+300.\u00ba~tricent\u00e9simo\n+300.\u00aa~tricent\u00e9sima\n+400.\u00ba~cuadringent\u00e9simo\n+400.\u00aa~cuadringent\u00e9sima\n+500.\u00ba~quingent\u00e9simo\n+500.\u00aa~quingent\u00e9sima\n+600.\u00ba~sexcent\u00e9simo\n+600.\u00aa~sexcent\u00e9sima\n+700.\u00ba~septingent\u00e9simo\n+700.\u00aa~septingent\u00e9sima\n+800.\u00ba~octingent\u00e9simo\n+800.\u00aa~octingent\u00e9sima\n+900.\u00ba~noningent\u00e9simo\n+900.\u00aa~noningent\u00e9sima\n+1000.\u00ba~mil\u00e9simo\n+1000.\u00aa~mil\u00e9sima\n+1001.\u1d49\u02b3~mil\u00e9simo primer\n+1 000.\u00ba~mil\u00e9simo\n+1 000.\u00aa~mil\u00e9sima\n+1 001.\u1d49\u02b3~mil\u00e9simo primer\n+1.000.\u00ba~mil\u00e9simo\n+1.000.\u00aa~mil\u00e9sima\n+1.001.\u1d49\u02b3~mil\u00e9simo primer\n+1248.\u00ba~mil\u00e9simo ducent\u00e9simo cuadrag\u00e9simo octavo\n+1248.\u00aa~mil\u00e9sima ducent\u00e9sima cuadrag\u00e9sima octava\n+2000.\u00ba~dosmil\u00e9simo\n+100 000.\u00ba~cienmil\u00e9simo\n+i~primero\n+I~primero\n+ii~segundo\n+II~segundo\n+iii~tercero\n+III~tercero\n+iv~cuarto\n+IV~cuarto\n+V~quinto\n+VI~sexto\n+VII~s\u00e9ptimo\n+VIII~octavo\n+IX~noveno\n+X~d\u00e9cimo\n+XI~und\u00e9cimo\n+XII~duod\u00e9cimo\n+XIII~decimotercero\n+XX~vig\u00e9simo\n+XXI~vigesimoprimero\n+XXX~trig\u00e9simo\n+XL~cuadrag\u00e9simo\n+L~quincuag\u00e9simo\n+XC~nonag\u00e9simo\n+C~cent\u00e9simo\n+CD~cuadringent\u00e9simo\n+D~quingent\u00e9simo\n+CM~noningent\u00e9simo\n+999.\u00ba~noningent\u00e9simo nonag\u00e9simo noveno\n+cmxcix~noningent\u00e9simo nonag\u00e9simo noveno\n\\ No newline at end of file\ndiff --git a/tests/nemo_text_processing/es/data_text_normalization/test_cases_telephone.txt b/tests/nemo_text_processing/es/data_text_normalization/test_cases_telephone.txt\nnew file mode 100644\n--- /dev/null\n+++ b/tests/nemo_text_processing/es/data_text_normalization/test_cases_telephone.txt\n@@ -0,0 +1,3 @@\n+123-123-5678~uno dos tres uno dos tres cinco seis siete ocho\n+123-456-789~uno dos tres cuatro cinco seis siete ocho nueve\n+1234-5678~uno dos tres cuatro cinco seis siete ocho\n\\ No newline at end of file\ndiff --git a/tests/nemo_text_processing/es/data_text_normalization/test_cases_time.txt b/tests/nemo_text_processing/es/data_text_normalization/test_cases_time.txt\nnew file mode 100644\n--- /dev/null\n+++ b/tests/nemo_text_processing/es/data_text_normalization/test_cases_time.txt\n@@ -0,0 +1,26 @@\n+1.00~una\n+1:00~una\n+01:00~una\n+01 h~una\n+3 h~tres horas\n+1 h~una hora\n+1.05 h~una y cinco\n+01.05 h~una y cinco\n+1.00 h~una\n+1.00 a.m.~una de la ma\u00f1ana\n+1.00 a.m~una de la ma\u00f1ana\n+1.00 p.m.~una de la tarde\n+1.00 p.m est~una de la tarde e s t\n+1.00 est~una e s t\n+5:02 est~cinco y dos e s t\n+5:02 p.m pst~cinco y dos de la noche p s t\n+5:02 p.m.~cinco y dos de la noche\n+12.15~doce y cuarto\n+12.15 a.m.~doce y cuarto de la noche\n+12.15 p.m.~doce y cuarto del mediod\u00eda\n+13.30~trece y media\n+14.05~catorce y cinco\n+24:50~veinticuatro y cincuenta\n+3:02:32 pst~tres horas dos minutos y treinta y dos segundos p s t\n+00:52~cero y cincuenta y dos\n+0:52~cero y cincuenta y dos\ndiff --git a/tests/nemo_text_processing/es/data_text_normalization/test_cases_whitelist.txt b/tests/nemo_text_processing/es/data_text_normalization/test_cases_whitelist.txt\nnew file mode 100644\n--- /dev/null\n+++ b/tests/nemo_text_processing/es/data_text_normalization/test_cases_whitelist.txt\n@@ -0,0 +1,3 @@\n+el dr.~el doctor\n+sr. rodriguez~se\u00f1or rodriguez\n+182 esq. toledo~ciento ochenta y dos esquina toledo\n\\ No newline at end of file\ndiff --git a/tests/nemo_text_processing/es/data_text_normalization/test_cases_word.txt b/tests/nemo_text_processing/es/data_text_normalization/test_cases_word.txt\nnew file mode 100644\n--- /dev/null\n+++ b/tests/nemo_text_processing/es/data_text_normalization/test_cases_word.txt\n@@ -0,0 +1,48 @@\n+~\n+yahoo!~yahoo!\n+veinte!~veinte!\n+\u2014~\u2014\n+aaa~aaa\n+aabach~aabach\n+aabenraa~aabenraa\n+aabye~aabye\n+aaccessed~aaccessed\n+aach~aach\n+aachen's~aachen's\n+aadri~aadri\n+aafia~aafia\n+aagaard~aagaard\n+aagadu~aagadu\n+aagard~aagard\n+aagathadi~aagathadi\n+aaghart's~aaghart's\n+aagnes~aagnes\n+aagomoni~aagomoni\n+aagon~aagon\n+aagoo~aagoo\n+aagot~aagot\n+aahar~aahar\n+aahh~aahh\n+aahperd~aahperd\n+aaibinterstate~aaibinterstate\n+aajab~aajab\n+aakasa~aakasa\n+aakervik~aakervik\n+aakirkeby~aakirkeby\n+aalam~aalam\n+aalbaek~aalbaek\n+aaldiu~aaldiu\n+aalem~aalem\n+a'ali~a'ali\n+aalilaassamthey~aalilaassamthey\n+aalin~aalin\n+aaliyan~aaliyan\n+aaliyan's~aaliyan's\n+aamadu~aamadu\n+aamara~aamara\n+aambala~aambala\n+aamera~aamera\n+aamer's~aamer's\n+aamina~aamina\n+aaminah~aaminah\n+aamjiwnaang~aamjiwnaang\ndiff --git a/tests/nemo_text_processing/es/test_cardinal.py b/tests/nemo_text_processing/es/test_cardinal.py\n--- a/tests/nemo_text_processing/es/test_cardinal.py\n+++ b/tests/nemo_text_processing/es/test_cardinal.py\n@@ -22,7 +22,8 @@\n \n \n class TestCardinal:\n-    inverse_normalizer_es = (\n+\n+    inverse_normalizer = (\n         InverseNormalizer(lang='es', cache_dir=CACHE_DIR, overwrite_cache=False) if PYNINI_AVAILABLE else None\n     )\n \n@@ -32,6 +33,34 @@ class TestCardinal:\n     )\n     @pytest.mark.run_only_on('CPU')\n     @pytest.mark.unit\n-    def test_denorm_es(self, test_input, expected):\n-        pred = self.inverse_normalizer_es.inverse_normalize(test_input, verbose=False)\n+    def test_denorm(self, test_input, expected):\n+        pred = self.inverse_normalizer.inverse_normalize(test_input, verbose=False)\n         assert pred == expected\n+\n+    normalizer = (\n+        Normalizer(input_case='cased', lang='es', cache_dir=CACHE_DIR, overwrite_cache=False)\n+        if PYNINI_AVAILABLE\n+        else None\n+    )\n+\n+    normalizer_with_audio = (\n+        NormalizerWithAudio(input_case='cased', lang='es', cache_dir=CACHE_DIR, overwrite_cache=False)\n+        if PYNINI_AVAILABLE and CACHE_DIR\n+        else None\n+    )\n+\n+    @parameterized.expand(parse_test_case_file('es/data_text_normalization/test_cases_cardinal.txt'))\n+    @pytest.mark.skipif(\n+        not PYNINI_AVAILABLE, reason=\"`pynini` not installed, please install via nemo_text_processing/setup.sh\"\n+    )\n+    @pytest.mark.run_only_on('CPU')\n+    @pytest.mark.unit\n+    def test_norm(self, test_input, expected):\n+        pred = self.normalizer.normalize(test_input, verbose=False)\n+        assert pred == expected\n+\n+        if self.normalizer_with_audio:\n+            pred_non_deterministic = self.normalizer_with_audio.normalize(\n+                test_input, n_tagged=1000, punct_post_process=False\n+            )\n+            assert expected in pred_non_deterministic\ndiff --git a/tests/nemo_text_processing/es/test_date.py b/tests/nemo_text_processing/es/test_date.py\n--- a/tests/nemo_text_processing/es/test_date.py\n+++ b/tests/nemo_text_processing/es/test_date.py\n@@ -22,7 +22,7 @@\n \n \n class TestDate:\n-    inverse_normalizer_es = (\n+    inverse_normalizer = (\n         InverseNormalizer(lang='es', cache_dir=CACHE_DIR, overwrite_cache=False) if PYNINI_AVAILABLE else None\n     )\n \n@@ -32,6 +32,34 @@ class TestDate:\n     )\n     @pytest.mark.run_only_on('CPU')\n     @pytest.mark.unit\n-    def test_denorm_es(self, test_input, expected):\n-        pred = self.inverse_normalizer_es.inverse_normalize(test_input, verbose=False)\n+    def test_denorm(self, test_input, expected):\n+        pred = self.inverse_normalizer.inverse_normalize(test_input, verbose=False)\n         assert pred == expected\n+\n+    normalizer = (\n+        Normalizer(input_case='cased', lang='es', cache_dir=CACHE_DIR, overwrite_cache=False)\n+        if PYNINI_AVAILABLE\n+        else None\n+    )\n+\n+    normalizer_with_audio = (\n+        NormalizerWithAudio(input_case='cased', lang='es', cache_dir=CACHE_DIR, overwrite_cache=False)\n+        if PYNINI_AVAILABLE and CACHE_DIR\n+        else None\n+    )\n+\n+    @parameterized.expand(parse_test_case_file('es/data_text_normalization/test_cases_date.txt'))\n+    @pytest.mark.skipif(\n+        not PYNINI_AVAILABLE, reason=\"`pynini` not installed, please install via nemo_text_processing/setup.sh\"\n+    )\n+    @pytest.mark.run_only_on('CPU')\n+    @pytest.mark.unit\n+    def test_norm(self, test_input, expected):\n+        pred = self.normalizer.normalize(test_input, verbose=False)\n+        assert pred == expected\n+\n+        if self.normalizer_with_audio:\n+            pred_non_deterministic = self.normalizer_with_audio.normalize(\n+                test_input, n_tagged=1000, punct_post_process=False\n+            )\n+            assert expected in pred_non_deterministic\ndiff --git a/tests/nemo_text_processing/es/test_decimal.py b/tests/nemo_text_processing/es/test_decimal.py\n--- a/tests/nemo_text_processing/es/test_decimal.py\n+++ b/tests/nemo_text_processing/es/test_decimal.py\n@@ -22,7 +22,7 @@\n \n \n class TestDecimal:\n-    inverse_normalizer_es = (\n+    inverse_normalizer = (\n         InverseNormalizer(lang='es', cache_dir=CACHE_DIR, overwrite_cache=False) if PYNINI_AVAILABLE else None\n     )\n \n@@ -32,6 +32,34 @@ class TestDecimal:\n     )\n     @pytest.mark.run_only_on('CPU')\n     @pytest.mark.unit\n-    def test_denorm_es(self, test_input, expected):\n-        pred = self.inverse_normalizer_es.inverse_normalize(test_input, verbose=False)\n+    def test_denorm(self, test_input, expected):\n+        pred = self.inverse_normalizer.inverse_normalize(test_input, verbose=False)\n         assert pred == expected\n+\n+    normalizer = (\n+        Normalizer(input_case='cased', lang='es', cache_dir=CACHE_DIR, overwrite_cache=False)\n+        if PYNINI_AVAILABLE\n+        else None\n+    )\n+\n+    normalizer_with_audio = (\n+        NormalizerWithAudio(input_case='cased', lang='es', cache_dir=CACHE_DIR, overwrite_cache=False)\n+        if PYNINI_AVAILABLE and CACHE_DIR\n+        else None\n+    )\n+\n+    @parameterized.expand(parse_test_case_file('es/data_text_normalization/test_cases_decimal.txt'))\n+    @pytest.mark.skipif(\n+        not PYNINI_AVAILABLE, reason=\"`pynini` not installed, please install via nemo_text_processing/setup.sh\"\n+    )\n+    @pytest.mark.run_only_on('CPU')\n+    @pytest.mark.unit\n+    def test_norm(self, test_input, expected):\n+        pred = self.normalizer.normalize(test_input, verbose=False)\n+        assert pred == expected\n+\n+        if self.normalizer_with_audio:\n+            pred_non_deterministic = self.normalizer_with_audio.normalize(\n+                test_input, n_tagged=1000, punct_post_process=False\n+            )\n+            assert expected in pred_non_deterministic\ndiff --git a/tests/nemo_text_processing/es/test_electronic.py b/tests/nemo_text_processing/es/test_electronic.py\n--- a/tests/nemo_text_processing/es/test_electronic.py\n+++ b/tests/nemo_text_processing/es/test_electronic.py\n@@ -35,3 +35,31 @@ class TestElectronic:\n     def test_denorm_es(self, test_input, expected):\n         pred = self.inverse_normalizer_es.inverse_normalize(test_input, verbose=False)\n         assert pred == expected\n+\n+    normalizer = (\n+        Normalizer(input_case='cased', lang='es', cache_dir=CACHE_DIR, overwrite_cache=False)\n+        if PYNINI_AVAILABLE\n+        else None\n+    )\n+\n+    normalizer_with_audio = (\n+        NormalizerWithAudio(input_case='cased', lang='es', cache_dir=CACHE_DIR, overwrite_cache=False)\n+        if PYNINI_AVAILABLE and CACHE_DIR\n+        else None\n+    )\n+\n+    @parameterized.expand(parse_test_case_file('es/data_text_normalization/test_cases_electronic.txt'))\n+    @pytest.mark.skipif(\n+        not PYNINI_AVAILABLE, reason=\"`pynini` not installed, please install via nemo_text_processing/setup.sh\"\n+    )\n+    @pytest.mark.run_only_on('CPU')\n+    @pytest.mark.unit\n+    def test_norm(self, test_input, expected):\n+        pred = self.normalizer.normalize(test_input, verbose=False)\n+        assert pred == expected\n+\n+        if self.normalizer_with_audio:\n+            pred_non_deterministic = self.normalizer_with_audio.normalize(\n+                test_input, n_tagged=1000, punct_post_process=False\n+            )\n+            assert expected in pred_non_deterministic\ndiff --git a/tests/nemo_text_processing/es/test_fraction.py b/tests/nemo_text_processing/es/test_fraction.py\nnew file mode 100644\n--- /dev/null\n+++ b/tests/nemo_text_processing/es/test_fraction.py\n@@ -0,0 +1,51 @@\n+# Copyright (c) 2022, NVIDIA CORPORATION & AFFILIATES.  All rights reserved.\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+#     http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+\n+\n+import pytest\n+from nemo_text_processing.text_normalization.normalize import Normalizer\n+from nemo_text_processing.text_normalization.normalize_with_audio import NormalizerWithAudio\n+from parameterized import parameterized\n+\n+from ..utils import CACHE_DIR, PYNINI_AVAILABLE, parse_test_case_file\n+\n+\n+class TestFraction:\n+    normalizer = (\n+        Normalizer(input_case='cased', lang='es', cache_dir=CACHE_DIR, overwrite_cache=False)\n+        if PYNINI_AVAILABLE\n+        else None\n+    )\n+\n+    normalizer_with_audio = (\n+        NormalizerWithAudio(input_case='cased', lang='es', cache_dir=CACHE_DIR, overwrite_cache=False)\n+        if PYNINI_AVAILABLE and CACHE_DIR\n+        else None\n+    )\n+\n+    @parameterized.expand(parse_test_case_file('es/data_text_normalization/test_cases_fraction.txt'))\n+    @pytest.mark.skipif(\n+        not PYNINI_AVAILABLE, reason=\"`pynini` not installed, please install via nemo_text_processing/setup.sh\"\n+    )\n+    @pytest.mark.run_only_on('CPU')\n+    @pytest.mark.unit\n+    def test_norm(self, test_input, expected):\n+        pred = self.normalizer.normalize(test_input, verbose=False)\n+        assert pred == expected\n+\n+        if self.normalizer_with_audio:\n+            pred_non_deterministic = self.normalizer_with_audio.normalize(\n+                test_input, n_tagged=1000, punct_post_process=False\n+            )\n+            assert expected in pred_non_deterministic\ndiff --git a/tests/nemo_text_processing/es/test_measure.py b/tests/nemo_text_processing/es/test_measure.py\n--- a/tests/nemo_text_processing/es/test_measure.py\n+++ b/tests/nemo_text_processing/es/test_measure.py\n@@ -36,3 +36,31 @@ class TestMeasure:\n     def test_denorm_es(self, test_input, expected):\n         pred = self.inverse_normalizer_es.inverse_normalize(test_input, verbose=False)\n         assert pred == expected\n+\n+    normalizer = (\n+        Normalizer(input_case='cased', lang='es', cache_dir=CACHE_DIR, overwrite_cache=False)\n+        if PYNINI_AVAILABLE\n+        else None\n+    )\n+\n+    normalizer_with_audio = (\n+        NormalizerWithAudio(input_case='cased', lang='es', cache_dir=CACHE_DIR, overwrite_cache=False)\n+        if PYNINI_AVAILABLE and CACHE_DIR\n+        else None\n+    )\n+\n+    @parameterized.expand(parse_test_case_file('es/data_text_normalization/test_cases_measure.txt'))\n+    @pytest.mark.skipif(\n+        not PYNINI_AVAILABLE, reason=\"`pynini` not installed, please install via nemo_text_processing/setup.sh\"\n+    )\n+    @pytest.mark.run_only_on('CPU')\n+    @pytest.mark.unit\n+    def test_norm(self, test_input, expected):\n+        pred = self.normalizer.normalize(test_input, verbose=False)\n+        assert pred == expected\n+\n+        if self.normalizer_with_audio:\n+            pred_non_deterministic = self.normalizer_with_audio.normalize(\n+                test_input, n_tagged=1000, punct_post_process=False\n+            )\n+            assert expected in pred_non_deterministic\ndiff --git a/tests/nemo_text_processing/es/test_money.py b/tests/nemo_text_processing/es/test_money.py\n--- a/tests/nemo_text_processing/es/test_money.py\n+++ b/tests/nemo_text_processing/es/test_money.py\n@@ -23,7 +23,7 @@\n \n \n class TestMoney:\n-    inverse_normalizer_es = (\n+    inverse_normalizer = (\n         InverseNormalizer(lang='es', cache_dir=CACHE_DIR, overwrite_cache=False) if PYNINI_AVAILABLE else None\n     )\n \n@@ -33,6 +33,34 @@ class TestMoney:\n     )\n     @pytest.mark.run_only_on('CPU')\n     @pytest.mark.unit\n-    def test_denorm_es(self, test_input, expected):\n-        pred = self.inverse_normalizer_es.inverse_normalize(test_input, verbose=False)\n+    def test_denorm(self, test_input, expected):\n+        pred = self.inverse_normalizer.inverse_normalize(test_input, verbose=False)\n         assert pred == expected\n+\n+    normalizer = (\n+        Normalizer(input_case='cased', lang='es', cache_dir=CACHE_DIR, overwrite_cache=False)\n+        if PYNINI_AVAILABLE\n+        else None\n+    )\n+\n+    normalizer_with_audio = (\n+        NormalizerWithAudio(input_case='cased', lang='es', cache_dir=CACHE_DIR, overwrite_cache=False)\n+        if PYNINI_AVAILABLE and CACHE_DIR\n+        else None\n+    )\n+\n+    @parameterized.expand(parse_test_case_file('es/data_text_normalization/test_cases_money.txt'))\n+    @pytest.mark.skipif(\n+        not PYNINI_AVAILABLE, reason=\"`pynini` not installed, please install via nemo_text_processing/setup.sh\"\n+    )\n+    @pytest.mark.run_only_on('CPU')\n+    @pytest.mark.unit\n+    def test_norm(self, test_input, expected):\n+        pred = self.normalizer.normalize(test_input, verbose=False)\n+        assert pred == expected\n+\n+        if self.normalizer_with_audio:\n+            pred_non_deterministic = self.normalizer_with_audio.normalize(\n+                test_input, n_tagged=1000, punct_post_process=False\n+            )\n+            assert expected in pred_non_deterministic\ndiff --git a/tests/nemo_text_processing/es/test_normalization_with_audio.py b/tests/nemo_text_processing/es/test_normalization_with_audio.py\nnew file mode 100644\n--- /dev/null\n+++ b/tests/nemo_text_processing/es/test_normalization_with_audio.py\n@@ -0,0 +1,43 @@\n+# Copyright (c) 2022, NVIDIA CORPORATION & AFFILIATES.  All rights reserved.\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+#     http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+\n+import pytest\n+from nemo_text_processing.text_normalization.normalize_with_audio import NormalizerWithAudio\n+from parameterized import parameterized\n+\n+from ..utils import CACHE_DIR, PYNINI_AVAILABLE, get_test_cases_multiple\n+\n+\n+class TestNormalizeWithAudio:\n+\n+    normalizer_es = (\n+        NormalizerWithAudio(input_case='cased', lang='es', cache_dir=CACHE_DIR, overwrite_cache=False)\n+        if PYNINI_AVAILABLE\n+        else None\n+    )\n+\n+    @parameterized.expand(get_test_cases_multiple('es/data_text_normalization/test_cases_normalize_with_audio.txt'))\n+    @pytest.mark.skipif(\n+        not PYNINI_AVAILABLE, reason=\"`pynini` not installed, please install via nemo_text_processing/setup.sh\"\n+    )\n+    @pytest.mark.run_only_on('CPU')\n+    @pytest.mark.unit\n+    def test_norm(self, test_input, expected):\n+        pred = self.normalizer_es.normalize(test_input, n_tagged=1000, punct_post_process=False)\n+        print(expected)\n+        print(\"pred\")\n+        print(pred)\n+        assert len(set(pred).intersection(set(expected))) == len(\n+            expected\n+        ), f'missing: {set(expected).difference(set(pred))}'\ndiff --git a/tests/nemo_text_processing/es/test_ordinal.py b/tests/nemo_text_processing/es/test_ordinal.py\n--- a/tests/nemo_text_processing/es/test_ordinal.py\n+++ b/tests/nemo_text_processing/es/test_ordinal.py\n@@ -23,7 +23,7 @@\n \n \n class TestOrdinal:\n-    inverse_normalizer_es = (\n+    inverse_normalizer = (\n         InverseNormalizer(lang='es', cache_dir=CACHE_DIR, overwrite_cache=False) if PYNINI_AVAILABLE else None\n     )\n \n@@ -33,6 +33,33 @@ class TestOrdinal:\n     )\n     @pytest.mark.run_only_on('CPU')\n     @pytest.mark.unit\n-    def test_denorm_es(self, test_input, expected):\n-        pred = self.inverse_normalizer_es.inverse_normalize(test_input, verbose=False)\n+    def test_denorm(self, test_input, expected):\n+        pred = self.inverse_normalizer.inverse_normalize(test_input, verbose=False)\n         assert pred == expected\n+\n+    normalizer = (\n+        Normalizer(input_case='cased', lang='es', cache_dir=CACHE_DIR, overwrite_cache=False)\n+        if PYNINI_AVAILABLE\n+        else None\n+    )\n+    normalizer_with_audio = (\n+        NormalizerWithAudio(input_case='cased', lang='es', cache_dir=CACHE_DIR, overwrite_cache=False)\n+        if PYNINI_AVAILABLE and CACHE_DIR\n+        else None\n+    )\n+\n+    @parameterized.expand(parse_test_case_file('es/data_text_normalization/test_cases_ordinal.txt'))\n+    @pytest.mark.skipif(\n+        not PYNINI_AVAILABLE, reason=\"`pynini` not installed, please install via nemo_text_processing/setup.sh\"\n+    )\n+    @pytest.mark.run_only_on('CPU')\n+    @pytest.mark.unit\n+    def test_norm(self, test_input, expected):\n+        pred = self.normalizer.normalize(test_input, verbose=False)\n+        assert pred == expected\n+\n+        if self.normalizer_with_audio:\n+            pred_non_deterministic = self.normalizer_with_audio.normalize(\n+                test_input, n_tagged=30, punct_post_process=False,\n+            )\n+            assert expected in pred_non_deterministic\ndiff --git a/tests/nemo_text_processing/es/test_sparrowhawk_normalization.sh b/tests/nemo_text_processing/es/test_sparrowhawk_normalization.sh\nnew file mode 100644\n--- /dev/null\n+++ b/tests/nemo_text_processing/es/test_sparrowhawk_normalization.sh\n@@ -0,0 +1,84 @@\n+#! /bin/sh\n+\n+PROJECT_DIR=/workspace/tests\n+\n+runtest () {\n+  input=$1\n+  cd /workspace/sparrowhawk/documentation/grammars\n+\n+  # read test file\n+  while read testcase; do\n+    IFS='~' read written spoken <<< $testcase\n+    denorm_pred=$(echo $written | normalizer_main --config=sparrowhawk_configuration.ascii_proto 2>&1 | tail -n 1)\n+\n+    # trim white space\n+    spoken=\"$(echo -e \"${spoken}\" | sed -e 's/^[[:space:]]*//' -e 's/[[:space:]]*$//')\"\n+    denorm_pred=\"$(echo -e \"${denorm_pred}\" | sed -e 's/^[[:space:]]*//' -e 's/[[:space:]]*$//')\"\n+\n+    # input expected actual\n+    assertEquals \"$written\" \"$spoken\" \"$denorm_pred\"\n+  done < \"$input\"\n+}\n+\n+testTNCardinal() {\n+  input=$PROJECT_DIR/es/data_text_normalization/test_cases_cardinal.txt\n+  runtest $input\n+}\n+\n+testTNDate() {\n+  input=$PROJECT_DIR/es/data_text_normalization/test_cases_date.txt\n+  runtest $input\n+}\n+\n+testTNDecimal() {\n+  input=$PROJECT_DIR/es/data_text_normalization/test_cases_decimal.txt\n+  runtest $input\n+}\n+\n+testTNElectronic() {\n+  input=$PROJECT_DIR/es/data_text_normalization/test_cases_electronic.txt\n+  runtest $input\n+}\n+\n+testTNFraction() {\n+  input=$PROJECT_DIR/es/data_text_normalization/test_cases_fraction.txt\n+  runtest $input\n+}\n+\n+testTNMoney() {\n+  input=$PROJECT_DIR/es/data_text_normalization/test_cases_money.txt\n+  runtest $input\n+}\n+\n+testTNOrdinal() {\n+  input=$PROJECT_DIR/es/data_text_normalization/test_cases_ordinal.txt\n+  runtest $input\n+}\n+\n+testTNTelephone() {\n+  input=$PROJECT_DIR/es/data_text_normalization/test_cases_ordinal.txt\n+  runtest $input\n+}\n+\n+testTNTime() {\n+  input=$PROJECT_DIR/es/data_text_normalization/test_cases_time.txt\n+  runtest $input\n+}\n+\n+testTNMeasure() {\n+  input=$PROJECT_DIR/es/data_text_normalization/test_cases_measure.txt\n+  runtest $input\n+}\n+\n+testTNWhitelist() {\n+  input=$PROJECT_DIR/es/data_text_normalization/test_cases_whitelist.txt\n+  runtest $input\n+}\n+\n+testTNWord() {\n+  input=$PROJECT_DIR/es/data_text_normalization/test_cases_word.txt\n+  runtest $input\n+}\n+\n+# Load shUnit2\n+. $PROJECT_DIR/../shunit2/shunit2\ndiff --git a/tests/nemo_text_processing/es/test_telephone.py b/tests/nemo_text_processing/es/test_telephone.py\n--- a/tests/nemo_text_processing/es/test_telephone.py\n+++ b/tests/nemo_text_processing/es/test_telephone.py\n@@ -36,3 +36,31 @@ class TestTelephone:\n     def test_denorm_es(self, test_input, expected):\n         pred = self.inverse_normalizer_es.inverse_normalize(test_input, verbose=False)\n         assert pred == expected\n+\n+    normalizer = (\n+        Normalizer(input_case='cased', lang='es', cache_dir=CACHE_DIR, overwrite_cache=False)\n+        if PYNINI_AVAILABLE\n+        else None\n+    )\n+\n+    normalizer_with_audio = (\n+        NormalizerWithAudio(input_case='cased', lang='es', cache_dir=CACHE_DIR, overwrite_cache=False)\n+        if PYNINI_AVAILABLE and CACHE_DIR\n+        else None\n+    )\n+\n+    @parameterized.expand(parse_test_case_file('es/data_text_normalization/test_cases_telephone.txt'))\n+    @pytest.mark.skipif(\n+        not PYNINI_AVAILABLE, reason=\"`pynini` not installed, please install via nemo_text_processing/setup.sh\"\n+    )\n+    @pytest.mark.run_only_on('CPU')\n+    @pytest.mark.unit\n+    def test_norm(self, test_input, expected):\n+        pred = self.normalizer.normalize(test_input, verbose=False)\n+        assert pred == expected\n+\n+        if self.normalizer_with_audio:\n+            pred_non_deterministic = self.normalizer_with_audio.normalize(\n+                test_input, n_tagged=1000, punct_post_process=False\n+            )\n+            assert expected in pred_non_deterministic\ndiff --git a/tests/nemo_text_processing/es/test_time.py b/tests/nemo_text_processing/es/test_time.py\n--- a/tests/nemo_text_processing/es/test_time.py\n+++ b/tests/nemo_text_processing/es/test_time.py\n@@ -35,3 +35,31 @@ class TestTime:\n     def test_denorm_es(self, test_input, expected):\n         pred = self.inverse_normalizer_es.inverse_normalize(test_input, verbose=False)\n         assert pred == expected\n+\n+    normalizer = (\n+        Normalizer(input_case='cased', lang='es', cache_dir=CACHE_DIR, overwrite_cache=False)\n+        if PYNINI_AVAILABLE\n+        else None\n+    )\n+\n+    normalizer_with_audio = (\n+        NormalizerWithAudio(input_case='cased', lang='es', cache_dir=CACHE_DIR, overwrite_cache=False)\n+        if PYNINI_AVAILABLE and CACHE_DIR\n+        else None\n+    )\n+\n+    @parameterized.expand(parse_test_case_file('es/data_text_normalization/test_cases_time.txt'))\n+    @pytest.mark.skipif(\n+        not PYNINI_AVAILABLE, reason=\"`pynini` not installed, please install via nemo_text_processing/setup.sh\"\n+    )\n+    @pytest.mark.run_only_on('CPU')\n+    @pytest.mark.unit\n+    def test_norm(self, test_input, expected):\n+        pred = self.normalizer.normalize(test_input, verbose=False)\n+        assert pred == expected\n+\n+        if self.normalizer_with_audio:\n+            pred_non_deterministic = self.normalizer_with_audio.normalize(\n+                test_input, n_tagged=1000, punct_post_process=False\n+            )\n+            assert expected in pred_non_deterministic\ndiff --git a/tests/nemo_text_processing/es/test_whitelist.py b/tests/nemo_text_processing/es/test_whitelist.py\n--- a/tests/nemo_text_processing/es/test_whitelist.py\n+++ b/tests/nemo_text_processing/es/test_whitelist.py\n@@ -35,3 +35,30 @@ class TestWhitelist:\n     def test_denorm_es(self, test_input, expected):\n         pred = self.inverse_normalizer_es.inverse_normalize(test_input, verbose=False)\n         assert pred == expected\n+\n+    normalizer = (\n+        Normalizer(input_case='cased', lang='es', cache_dir=CACHE_DIR, overwrite_cache=False)\n+        if PYNINI_AVAILABLE\n+        else None\n+    )\n+    normalizer_with_audio = (\n+        NormalizerWithAudio(input_case='cased', lang='es', cache_dir=CACHE_DIR, overwrite_cache=False)\n+        if PYNINI_AVAILABLE and CACHE_DIR\n+        else None\n+    )\n+\n+    @parameterized.expand(parse_test_case_file('es/data_text_normalization/test_cases_whitelist.txt'))\n+    @pytest.mark.skipif(\n+        not PYNINI_AVAILABLE, reason=\"`pynini` not installed, please install via nemo_text_processing/setup.sh\"\n+    )\n+    @pytest.mark.run_only_on('CPU')\n+    @pytest.mark.unit\n+    def test_norm(self, test_input, expected):\n+        pred = self.normalizer.normalize(test_input, verbose=False)\n+        assert pred == expected\n+\n+        if self.normalizer_with_audio:\n+            pred_non_deterministic = self.normalizer_with_audio.normalize(\n+                test_input, n_tagged=10, punct_post_process=False\n+            )\n+            assert expected in pred_non_deterministic\ndiff --git a/tests/nemo_text_processing/es/test_word.py b/tests/nemo_text_processing/es/test_word.py\n--- a/tests/nemo_text_processing/es/test_word.py\n+++ b/tests/nemo_text_processing/es/test_word.py\n@@ -35,3 +35,30 @@ class TestWord:\n     def test_denorm_es(self, test_input, expected):\n         pred = self.inverse_normalizer_es.inverse_normalize(test_input, verbose=False)\n         assert pred == expected\n+\n+    normalizer_es = (\n+        Normalizer(input_case='cased', lang='es', cache_dir=CACHE_DIR, overwrite_cache=False)\n+        if PYNINI_AVAILABLE\n+        else None\n+    )\n+    normalizer_with_audio_es = (\n+        NormalizerWithAudio(input_case='cased', lang='es', cache_dir=CACHE_DIR, overwrite_cache=False)\n+        if PYNINI_AVAILABLE and CACHE_DIR\n+        else None\n+    )\n+\n+    @parameterized.expand(parse_test_case_file('es/data_text_normalization/test_cases_word.txt'))\n+    @pytest.mark.skipif(\n+        not PYNINI_AVAILABLE, reason=\"`pynini` not installed, please install via nemo_text_processing/setup.sh\"\n+    )\n+    @pytest.mark.run_only_on('CPU')\n+    @pytest.mark.unit\n+    def test_norm(self, test_input, expected):\n+        pred = self.normalizer_es.normalize(test_input, verbose=False)\n+        assert pred == expected, f\"input: {test_input}\"\n+\n+        if self.normalizer_with_audio_es:\n+            pred_non_deterministic = self.normalizer_with_audio_es.normalize(\n+                test_input, n_tagged=150, punct_post_process=False\n+            )\n+            assert expected in pred_non_deterministic, f\"input: {test_input}\"\n",
-    "problem_statement": "./reinstall.sh crashes due to not being able to uninstall llvmlite\nStarting off of `nemo:1.5.1` container, cloning the NeMo repo to a folder inside of it and calling `./reinstall.sh` fails with\r\n```\r\nERROR: Cannot uninstall 'llvmlite'. It is a distutils installed project and thus we cannot accurately determine which files belong to it which would lead to only a partial uninstall.\r\n```\r\n`pip install -e` on the other hand succeeds installing `nemo:1.7.0rc0` and `numpy:1.22.2`, the rest of the packages remain untouched.\r\n\r\nIt seems that `./reinstall.sh` which used to work fine, a week or so ago when following the same procedure to upgrade to `nemo:1.6.0rc` redeveloped issue #841. The solution remains the same, first call\r\n```\r\npip install --ignore-installed llvmlite\r\n```\r\nfollowed by `./reinstall.sh`. In this case, apart `llvml`, the following packages are updated\r\n```\r\nftfy-6.0.3 nemo-toolkit-1.7.0rc0 numba-0.55.1 pytorch-lightning-1.5.9 sacrebleu-2.0.0 setuptools-59.5.0\r\n```\r\nInterestingly `numpy` in this case is left at `1.21.5`.\n",
-    "hints_text": "",
-    "created_at": "2022-02-09T05:12:31Z",
-    "version": "1.0"
-  },
-  {
-    "repo": "NVIDIA/NeMo",
-    "pull_number": 7582,
-    "instance_id": "NVIDIA__NeMo-7582",
-    "issue_numbers": [
-      "7166"
-    ],
-    "base_commit": "8a892b86186dbdf61803d75570cb5c58471e9dda",
-    "patch": "diff --git a/examples/asr/experimental/k2/align_speech_parallel.py b/examples/asr/experimental/k2/align_speech_parallel.py\n--- a/examples/asr/experimental/k2/align_speech_parallel.py\n+++ b/examples/asr/experimental/k2/align_speech_parallel.py\n@@ -74,7 +74,7 @@\n \n \n import os\n-from dataclasses import dataclass, is_dataclass\n+from dataclasses import dataclass, field, is_dataclass\n from typing import Optional\n \n import pytorch_lightning as ptl\n@@ -94,12 +94,14 @@\n @dataclass\n class ParallelAlignmentConfig:\n     model: Optional[str] = None  # name\n-    predict_ds: ASRDatasetConfig = ASRDatasetConfig(return_sample_id=True, num_workers=4)\n-    aligner_args: K2AlignerWrapperModelConfig = K2AlignerWrapperModelConfig()\n+    predict_ds: ASRDatasetConfig = field(\n+        default_factory=lambda: ASRDatasetConfig(return_sample_id=True, num_workers=4)\n+    )\n+    aligner_args: K2AlignerWrapperModelConfig = field(default_factory=lambda: K2AlignerWrapperModelConfig())\n     output_path: str = MISSING\n     model_stride: int = 8\n \n-    trainer: TrainerConfig = TrainerConfig(gpus=-1, accelerator=\"ddp\")\n+    trainer: TrainerConfig = field(default_factory=lambda: TrainerConfig(gpus=-1, accelerator=\"ddp\"))\n \n     # there arguments will be ignored\n     return_predictions: bool = False\ndiff --git a/nemo/collections/asr/metrics/rnnt_wer.py b/nemo/collections/asr/metrics/rnnt_wer.py\n--- a/nemo/collections/asr/metrics/rnnt_wer.py\n+++ b/nemo/collections/asr/metrics/rnnt_wer.py\n@@ -15,7 +15,7 @@\n import copy\n import re\n from abc import abstractmethod\n-from dataclasses import dataclass, is_dataclass\n+from dataclasses import dataclass, field, is_dataclass\n from typing import Callable, Dict, List, Optional, Tuple, Union\n \n import editdistance\n@@ -1299,7 +1299,7 @@ class RNNTDecodingConfig:\n     preserve_alignments: Optional[bool] = None\n \n     #  confidence config\n-    confidence_cfg: ConfidenceConfig = ConfidenceConfig()\n+    confidence_cfg: ConfidenceConfig = field(default_factory=lambda: ConfidenceConfig())\n \n     # RNNT Joint fused batch size\n     fused_batch_size: Optional[int] = None\n@@ -1317,10 +1317,10 @@ class RNNTDecodingConfig:\n     rnnt_timestamp_type: str = \"all\"  # can be char, word or all for both\n \n     # greedy decoding config\n-    greedy: greedy_decode.GreedyRNNTInferConfig = greedy_decode.GreedyRNNTInferConfig()\n+    greedy: greedy_decode.GreedyRNNTInferConfig = field(default_factory=lambda: greedy_decode.GreedyRNNTInferConfig())\n \n     # beam decoding config\n-    beam: beam_decode.BeamRNNTInferConfig = beam_decode.BeamRNNTInferConfig(beam_size=4)\n+    beam: beam_decode.BeamRNNTInferConfig = field(default_factory=lambda: beam_decode.BeamRNNTInferConfig(beam_size=4))\n \n     # can be used to change temperature for decoding\n     temperature: float = 1.0\ndiff --git a/nemo/collections/asr/metrics/wer.py b/nemo/collections/asr/metrics/wer.py\n--- a/nemo/collections/asr/metrics/wer.py\n+++ b/nemo/collections/asr/metrics/wer.py\n@@ -14,7 +14,7 @@\n \n import re\n from abc import abstractmethod\n-from dataclasses import dataclass, is_dataclass\n+from dataclasses import dataclass, field, is_dataclass\n from typing import Callable, Dict, List, Optional, Tuple, Union\n \n import editdistance\n@@ -1297,13 +1297,17 @@ class CTCDecodingConfig:\n     batch_dim_index: int = 0\n \n     # greedy decoding config\n-    greedy: ctc_greedy_decoding.GreedyCTCInferConfig = ctc_greedy_decoding.GreedyCTCInferConfig()\n+    greedy: ctc_greedy_decoding.GreedyCTCInferConfig = field(\n+        default_factory=lambda: ctc_greedy_decoding.GreedyCTCInferConfig()\n+    )\n \n     # beam decoding config\n-    beam: ctc_beam_decoding.BeamCTCInferConfig = ctc_beam_decoding.BeamCTCInferConfig(beam_size=4)\n+    beam: ctc_beam_decoding.BeamCTCInferConfig = field(\n+        default_factory=lambda: ctc_beam_decoding.BeamCTCInferConfig(beam_size=4)\n+    )\n \n     # confidence config\n-    confidence_cfg: ConfidenceConfig = ConfidenceConfig()\n+    confidence_cfg: ConfidenceConfig = field(default_factory=lambda: ConfidenceConfig())\n \n     # can be used to change temperature for decoding\n     temperature: float = 1.0\ndiff --git a/nemo/collections/asr/models/configs/aligner_config.py b/nemo/collections/asr/models/configs/aligner_config.py\n--- a/nemo/collections/asr/models/configs/aligner_config.py\n+++ b/nemo/collections/asr/models/configs/aligner_config.py\n@@ -12,7 +12,7 @@\n # See the License for the specific language governing permissions and\n # limitations under the License.\n \n-from dataclasses import dataclass\n+from dataclasses import dataclass, field\n \n from nemo.collections.asr.parts.k2.classes import GraphModuleConfig\n \n@@ -35,10 +35,10 @@ class AlignerWrapperModelConfig:\n     word_output: bool = True\n     cpu_decoding: bool = False\n     decode_batch_size: int = 0\n-    ctc_cfg: AlignerCTCConfig = AlignerCTCConfig()\n-    rnnt_cfg: AlignerRNNTConfig = AlignerRNNTConfig()\n+    ctc_cfg: AlignerCTCConfig = field(default_factory=lambda: AlignerCTCConfig())\n+    rnnt_cfg: AlignerRNNTConfig = field(default_factory=lambda: AlignerRNNTConfig())\n \n \n @dataclass\n class K2AlignerWrapperModelConfig(AlignerWrapperModelConfig):\n-    decoder_module_cfg: GraphModuleConfig = GraphModuleConfig()\n+    decoder_module_cfg: GraphModuleConfig = field(default_factory=lambda: GraphModuleConfig())\ndiff --git a/nemo/collections/asr/models/configs/asr_models_config.py b/nemo/collections/asr/models/configs/asr_models_config.py\n--- a/nemo/collections/asr/models/configs/asr_models_config.py\n+++ b/nemo/collections/asr/models/configs/asr_models_config.py\n@@ -12,7 +12,7 @@\n # See the License for the specific language governing permissions and\n # limitations under the License.\n \n-from dataclasses import dataclass\n+from dataclasses import dataclass, field\n from typing import Any, Dict, List, Optional\n \n from omegaconf import MISSING\n@@ -74,24 +74,32 @@ class EncDecCTCConfig(model_cfg.ModelConfig):\n     labels: List[str] = MISSING\n \n     # Dataset configs\n-    train_ds: ASRDatasetConfig = ASRDatasetConfig(manifest_filepath=None, shuffle=True)\n-    validation_ds: ASRDatasetConfig = ASRDatasetConfig(manifest_filepath=None, shuffle=False)\n-    test_ds: ASRDatasetConfig = ASRDatasetConfig(manifest_filepath=None, shuffle=False)\n+    train_ds: ASRDatasetConfig = field(default_factory=lambda: ASRDatasetConfig(manifest_filepath=None, shuffle=True))\n+    validation_ds: ASRDatasetConfig = field(\n+        default_factory=lambda: ASRDatasetConfig(manifest_filepath=None, shuffle=False)\n+    )\n+    test_ds: ASRDatasetConfig = field(default_factory=lambda: ASRDatasetConfig(manifest_filepath=None, shuffle=False))\n \n     # Optimizer / Scheduler config\n-    optim: Optional[model_cfg.OptimConfig] = model_cfg.OptimConfig(sched=model_cfg.SchedConfig())\n+    optim: Optional[model_cfg.OptimConfig] = field(\n+        default_factory=lambda: model_cfg.OptimConfig(sched=model_cfg.SchedConfig())\n+    )\n \n     # Model component configs\n-    preprocessor: AudioToMelSpectrogramPreprocessorConfig = AudioToMelSpectrogramPreprocessorConfig()\n-    spec_augment: Optional[SpectrogramAugmentationConfig] = SpectrogramAugmentationConfig()\n-    encoder: ConvASREncoderConfig = ConvASREncoderConfig()\n-    decoder: ConvASRDecoderConfig = ConvASRDecoderConfig()\n-    decoding: CTCDecodingConfig = CTCDecodingConfig()\n+    preprocessor: AudioToMelSpectrogramPreprocessorConfig = field(\n+        default_factory=lambda: AudioToMelSpectrogramPreprocessorConfig()\n+    )\n+    spec_augment: Optional[SpectrogramAugmentationConfig] = field(\n+        default_factory=lambda: SpectrogramAugmentationConfig()\n+    )\n+    encoder: ConvASREncoderConfig = field(default_factory=lambda: ConvASREncoderConfig())\n+    decoder: ConvASRDecoderConfig = field(default_factory=lambda: ConvASRDecoderConfig())\n+    decoding: CTCDecodingConfig = field(default_factory=lambda: CTCDecodingConfig())\n \n \n @dataclass\n class EncDecCTCModelConfig(model_cfg.NemoConfig):\n-    model: EncDecCTCConfig = EncDecCTCConfig()\n+    model: EncDecCTCConfig = field(default_factory=lambda: EncDecCTCConfig())\n \n \n @dataclass\ndiff --git a/nemo/collections/asr/models/configs/classification_models_config.py b/nemo/collections/asr/models/configs/classification_models_config.py\n--- a/nemo/collections/asr/models/configs/classification_models_config.py\n+++ b/nemo/collections/asr/models/configs/classification_models_config.py\n@@ -12,7 +12,7 @@\n # See the License for the specific language governing permissions and\n # limitations under the License.\n \n-from dataclasses import dataclass\n+from dataclasses import dataclass, field\n from typing import Any, Dict, List, Optional\n \n from omegaconf import MISSING\n@@ -72,30 +72,40 @@ class EncDecClassificationConfig(model_cfg.ModelConfig):\n     timesteps: int = MISSING\n \n     # Dataset configs\n-    train_ds: EncDecClassificationDatasetConfig = EncDecClassificationDatasetConfig(\n-        manifest_filepath=None, shuffle=True, trim_silence=False\n+    train_ds: EncDecClassificationDatasetConfig = field(\n+        default_factory=lambda: EncDecClassificationDatasetConfig(\n+            manifest_filepath=None, shuffle=True, trim_silence=False\n+        )\n     )\n-    validation_ds: EncDecClassificationDatasetConfig = EncDecClassificationDatasetConfig(\n-        manifest_filepath=None, shuffle=False\n+    validation_ds: EncDecClassificationDatasetConfig = field(\n+        default_factory=lambda: EncDecClassificationDatasetConfig(manifest_filepath=None, shuffle=False)\n     )\n-    test_ds: EncDecClassificationDatasetConfig = EncDecClassificationDatasetConfig(\n-        manifest_filepath=None, shuffle=False\n+    test_ds: EncDecClassificationDatasetConfig = field(\n+        default_factory=lambda: EncDecClassificationDatasetConfig(manifest_filepath=None, shuffle=False)\n     )\n \n     # Optimizer / Scheduler config\n-    optim: Optional[model_cfg.OptimConfig] = model_cfg.OptimConfig(sched=model_cfg.SchedConfig())\n+    optim: Optional[model_cfg.OptimConfig] = field(\n+        default_factory=lambda: model_cfg.OptimConfig(sched=model_cfg.SchedConfig())\n+    )\n \n     # Model component configs\n-    preprocessor: AudioToMFCCPreprocessorConfig = AudioToMFCCPreprocessorConfig()\n-    spec_augment: Optional[SpectrogramAugmentationConfig] = SpectrogramAugmentationConfig()\n-    crop_or_pad_augment: Optional[CropOrPadSpectrogramAugmentationConfig] = CropOrPadSpectrogramAugmentationConfig(\n-        audio_length=timesteps\n+    preprocessor: AudioToMFCCPreprocessorConfig = field(default_factory=lambda: AudioToMFCCPreprocessorConfig())\n+    spec_augment: Optional[SpectrogramAugmentationConfig] = field(\n+        default_factory=lambda: SpectrogramAugmentationConfig()\n+    )\n+    crop_or_pad_augment: Optional[CropOrPadSpectrogramAugmentationConfig] = field(\n+        default_factory=lambda: CropOrPadSpectrogramAugmentationConfig(audio_length=-1)\n     )\n \n-    encoder: ConvASREncoderConfig = ConvASREncoderConfig()\n-    decoder: ConvASRDecoderClassificationConfig = ConvASRDecoderClassificationConfig()\n+    encoder: ConvASREncoderConfig = field(default_factory=lambda: ConvASREncoderConfig())\n+    decoder: ConvASRDecoderClassificationConfig = field(default_factory=lambda: ConvASRDecoderClassificationConfig())\n+\n+    def __post_init__(self):\n+        if self.crop_or_pad_augment is not None:\n+            self.crop_or_pad_augment.audio_length = self.timesteps\n \n \n @dataclass\n class EncDecClassificationModelConfig(model_cfg.NemoConfig):\n-    model: EncDecClassificationConfig = EncDecClassificationConfig()\n+    model: EncDecClassificationConfig = field(default_factory=lambda: EncDecClassificationConfig())\ndiff --git a/nemo/collections/asr/models/configs/diarizer_config.py b/nemo/collections/asr/models/configs/diarizer_config.py\n--- a/nemo/collections/asr/models/configs/diarizer_config.py\n+++ b/nemo/collections/asr/models/configs/diarizer_config.py\n@@ -12,7 +12,7 @@\n # See the License for the specific language governing permissions and\n # limitations under the License.\n \n-from dataclasses import asdict, dataclass\n+from dataclasses import asdict, dataclass, field\n from typing import Any, Dict, Optional, Tuple, Union\n \n \n@@ -78,9 +78,9 @@ class ASRDiarizerParams(DiarizerComponentConfig):\n @dataclass\n class ASRDiarizerConfig(DiarizerComponentConfig):\n     model_path: Optional[str] = \"stt_en_conformer_ctc_large\"\n-    parameters: ASRDiarizerParams = ASRDiarizerParams()\n-    ctc_decoder_parameters: ASRDiarizerCTCDecoderParams = ASRDiarizerCTCDecoderParams()\n-    realigning_lm_parameters: ASRRealigningLMParams = ASRRealigningLMParams()\n+    parameters: ASRDiarizerParams = field(default_factory=lambda: ASRDiarizerParams())\n+    ctc_decoder_parameters: ASRDiarizerCTCDecoderParams = field(default_factory=lambda: ASRDiarizerCTCDecoderParams())\n+    realigning_lm_parameters: ASRRealigningLMParams = field(default_factory=lambda: ASRRealigningLMParams())\n \n \n @dataclass\n@@ -102,7 +102,7 @@ class VADParams(DiarizerComponentConfig):\n class VADConfig(DiarizerComponentConfig):\n     model_path: str = \"vad_multilingual_marblenet\"  # .nemo local model path or pretrained VAD model name\n     external_vad_manifest: Optional[str] = None\n-    parameters: VADParams = VADParams()\n+    parameters: VADParams = field(default_factory=lambda: VADParams())\n \n \n @dataclass\n@@ -121,7 +121,7 @@ class SpeakerEmbeddingsParams(DiarizerComponentConfig):\n class SpeakerEmbeddingsConfig(DiarizerComponentConfig):\n     # .nemo local model path or pretrained model name (titanet_large, ecapa_tdnn or speakerverification_speakernet)\n     model_path: Optional[str] = None\n-    parameters: SpeakerEmbeddingsParams = SpeakerEmbeddingsParams()\n+    parameters: SpeakerEmbeddingsParams = field(default_factory=lambda: SpeakerEmbeddingsParams())\n \n \n @dataclass\n@@ -142,7 +142,7 @@ class ClusteringParams(DiarizerComponentConfig):\n \n @dataclass\n class ClusteringConfig(DiarizerComponentConfig):\n-    parameters: ClusteringParams = ClusteringParams()\n+    parameters: ClusteringParams = field(default_factory=lambda: ClusteringParams())\n \n \n @dataclass\n@@ -166,7 +166,7 @@ class MSDDParams(DiarizerComponentConfig):\n @dataclass\n class MSDDConfig(DiarizerComponentConfig):\n     model_path: Optional[str] = \"diar_msdd_telephonic\"\n-    parameters: MSDDParams = MSDDParams()\n+    parameters: MSDDParams = field(default_factory=lambda: MSDDParams())\n \n \n @dataclass\n@@ -176,16 +176,16 @@ class DiarizerConfig(DiarizerComponentConfig):\n     oracle_vad: bool = False  # If True, uses RTTM files provided in the manifest file to get VAD timestamps\n     collar: float = 0.25  # Collar value for scoring\n     ignore_overlap: bool = True  # Consider or ignore overlap segments while scoring\n-    vad: VADConfig = VADConfig()\n-    speaker_embeddings: SpeakerEmbeddingsConfig = SpeakerEmbeddingsConfig()\n-    clustering: ClusteringConfig = ClusteringConfig()\n-    msdd_model: MSDDConfig = MSDDConfig()\n-    asr: ASRDiarizerConfig = ASRDiarizerConfig()\n+    vad: VADConfig = field(default_factory=lambda: VADConfig())\n+    speaker_embeddings: SpeakerEmbeddingsConfig = field(default_factory=lambda: SpeakerEmbeddingsConfig())\n+    clustering: ClusteringConfig = field(default_factory=lambda: ClusteringConfig())\n+    msdd_model: MSDDConfig = field(default_factory=lambda: MSDDConfig())\n+    asr: ASRDiarizerConfig = field(default_factory=lambda: ASRDiarizerConfig())\n \n \n @dataclass\n class NeuralDiarizerInferenceConfig(DiarizerComponentConfig):\n-    diarizer: DiarizerConfig = DiarizerConfig()\n+    diarizer: DiarizerConfig = field(default_factory=lambda: DiarizerConfig())\n     device: str = \"cpu\"\n     verbose: bool = False\n     batch_size: int = 64\ndiff --git a/nemo/collections/asr/models/configs/k2_sequence_models_config.py b/nemo/collections/asr/models/configs/k2_sequence_models_config.py\n--- a/nemo/collections/asr/models/configs/k2_sequence_models_config.py\n+++ b/nemo/collections/asr/models/configs/k2_sequence_models_config.py\n@@ -12,7 +12,7 @@\n # See the License for the specific language governing permissions and\n # limitations under the License.\n \n-from dataclasses import dataclass\n+from dataclasses import dataclass, field\n \n from nemo.collections.asr.models.configs.asr_models_config import EncDecCTCConfig\n from nemo.collections.asr.parts.k2.classes import GraphModuleConfig as BackendConfig\n@@ -26,14 +26,14 @@ class GraphModuleConfig:\n     split_batch_size: int = 0\n     dec_type: str = \"topo\"\n     transcribe_training: bool = True\n-    backend_cfg: BackendConfig = BackendConfig()\n+    backend_cfg: BackendConfig = field(default_factory=lambda: BackendConfig())\n \n \n @dataclass\n class EncDecK2SeqConfig(EncDecCTCConfig):\n-    graph_module_cfg: GraphModuleConfig = GraphModuleConfig()\n+    graph_module_cfg: GraphModuleConfig = field(default_factory=lambda: GraphModuleConfig())\n \n \n @dataclass\n class EncDecK2SeqModelConfig(NemoConfig):\n-    model: EncDecK2SeqConfig = EncDecK2SeqConfig()\n+    model: EncDecK2SeqConfig = field(default_factory=lambda: EncDecK2SeqConfig())\ndiff --git a/nemo/collections/asr/models/configs/matchboxnet_config.py b/nemo/collections/asr/models/configs/matchboxnet_config.py\n--- a/nemo/collections/asr/models/configs/matchboxnet_config.py\n+++ b/nemo/collections/asr/models/configs/matchboxnet_config.py\n@@ -107,30 +107,38 @@ class MatchboxNetModelConfig(clf_cfg.EncDecClassificationConfig):\n     labels: List[str] = MISSING\n \n     # Dataset configs\n-    train_ds: clf_cfg.EncDecClassificationDatasetConfig = clf_cfg.EncDecClassificationDatasetConfig(\n-        manifest_filepath=None, shuffle=True, trim_silence=False\n+    train_ds: clf_cfg.EncDecClassificationDatasetConfig = field(\n+        default_factory=lambda: clf_cfg.EncDecClassificationDatasetConfig(\n+            manifest_filepath=None, shuffle=True, trim_silence=False\n+        )\n     )\n-    validation_ds: clf_cfg.EncDecClassificationDatasetConfig = clf_cfg.EncDecClassificationDatasetConfig(\n-        manifest_filepath=None, shuffle=False\n+    validation_ds: clf_cfg.EncDecClassificationDatasetConfig = field(\n+        default_factory=lambda: clf_cfg.EncDecClassificationDatasetConfig(manifest_filepath=None, shuffle=False)\n     )\n-    test_ds: clf_cfg.EncDecClassificationDatasetConfig = clf_cfg.EncDecClassificationDatasetConfig(\n-        manifest_filepath=None, shuffle=False\n+    test_ds: clf_cfg.EncDecClassificationDatasetConfig = field(\n+        default_factory=lambda: clf_cfg.EncDecClassificationDatasetConfig(manifest_filepath=None, shuffle=False)\n     )\n \n     # Optimizer / Scheduler config\n-    optim: Optional[model_cfg.OptimConfig] = model_cfg.OptimConfig(sched=model_cfg.SchedConfig())\n+    optim: Optional[model_cfg.OptimConfig] = field(\n+        default_factory=lambda: model_cfg.OptimConfig(sched=model_cfg.SchedConfig())\n+    )\n \n     # Model general component configs\n-    preprocessor: AudioToMFCCPreprocessorConfig = AudioToMFCCPreprocessorConfig(window_size=0.025)\n-    spec_augment: Optional[SpectrogramAugmentationConfig] = SpectrogramAugmentationConfig(\n-        freq_masks=2, time_masks=2, freq_width=15, time_width=25, rect_masks=5, rect_time=25, rect_freq=15\n+    preprocessor: AudioToMFCCPreprocessorConfig = field(\n+        default_factory=lambda: AudioToMFCCPreprocessorConfig(window_size=0.025)\n+    )\n+    spec_augment: Optional[SpectrogramAugmentationConfig] = field(\n+        default_factory=lambda: SpectrogramAugmentationConfig(\n+            freq_masks=2, time_masks=2, freq_width=15, time_width=25, rect_masks=5, rect_time=25, rect_freq=15\n+        )\n     )\n-    crop_or_pad_augment: Optional[CropOrPadSpectrogramAugmentationConfig] = CropOrPadSpectrogramAugmentationConfig(\n-        audio_length=128\n+    crop_or_pad_augment: Optional[CropOrPadSpectrogramAugmentationConfig] = field(\n+        default_factory=lambda: CropOrPadSpectrogramAugmentationConfig(audio_length=128)\n     )\n \n-    encoder: ConvASREncoderConfig = ConvASREncoderConfig(activation=\"relu\")\n-    decoder: ConvASRDecoderClassificationConfig = ConvASRDecoderClassificationConfig()\n+    encoder: ConvASREncoderConfig = field(default_factory=lambda: ConvASREncoderConfig(activation=\"relu\"))\n+    decoder: ConvASRDecoderClassificationConfig = field(default_factory=lambda: ConvASRDecoderClassificationConfig())\n \n \n @dataclass\ndiff --git a/nemo/collections/asr/models/configs/quartznet_config.py b/nemo/collections/asr/models/configs/quartznet_config.py\n--- a/nemo/collections/asr/models/configs/quartznet_config.py\n+++ b/nemo/collections/asr/models/configs/quartznet_config.py\n@@ -12,7 +12,7 @@\n # See the License for the specific language governing permissions and\n # limitations under the License.\n \n-from dataclasses import dataclass\n+from dataclasses import dataclass, field\n from typing import Any, Callable, List, Optional\n \n from omegaconf import MISSING\n@@ -174,20 +174,30 @@ class JasperModelConfig(ctc_cfg.EncDecCTCConfig):\n     labels: List[str] = MISSING\n \n     # Dataset configs\n-    train_ds: ctc_cfg.ASRDatasetConfig = ctc_cfg.ASRDatasetConfig(\n-        manifest_filepath=None, shuffle=True, trim_silence=True\n+    train_ds: ctc_cfg.ASRDatasetConfig = field(\n+        default_factory=lambda: ctc_cfg.ASRDatasetConfig(manifest_filepath=None, shuffle=True, trim_silence=True)\n+    )\n+    validation_ds: ctc_cfg.ASRDatasetConfig = field(\n+        default_factory=lambda: ctc_cfg.ASRDatasetConfig(manifest_filepath=None, shuffle=False)\n+    )\n+    test_ds: ctc_cfg.ASRDatasetConfig = field(\n+        default_factory=lambda: ctc_cfg.ASRDatasetConfig(manifest_filepath=None, shuffle=False)\n     )\n-    validation_ds: ctc_cfg.ASRDatasetConfig = ctc_cfg.ASRDatasetConfig(manifest_filepath=None, shuffle=False)\n-    test_ds: ctc_cfg.ASRDatasetConfig = ctc_cfg.ASRDatasetConfig(manifest_filepath=None, shuffle=False)\n \n     # Optimizer / Scheduler config\n-    optim: Optional[model_cfg.OptimConfig] = model_cfg.OptimConfig(sched=model_cfg.SchedConfig())\n+    optim: Optional[model_cfg.OptimConfig] = field(\n+        default_factory=lambda: model_cfg.OptimConfig(sched=model_cfg.SchedConfig())\n+    )\n \n     # Model general component configs\n-    preprocessor: AudioToMelSpectrogramPreprocessorConfig = AudioToMelSpectrogramPreprocessorConfig()\n-    spec_augment: Optional[SpectrogramAugmentationConfig] = SpectrogramAugmentationConfig()\n-    encoder: ConvASREncoderConfig = ConvASREncoderConfig(activation=\"relu\")\n-    decoder: ConvASRDecoderConfig = ConvASRDecoderConfig()\n+    preprocessor: AudioToMelSpectrogramPreprocessorConfig = field(\n+        default_factory=lambda: AudioToMelSpectrogramPreprocessorConfig()\n+    )\n+    spec_augment: Optional[SpectrogramAugmentationConfig] = field(\n+        default_factory=lambda: SpectrogramAugmentationConfig()\n+    )\n+    encoder: ConvASREncoderConfig = field(default_factory=lambda: ConvASREncoderConfig(activation=\"relu\"))\n+    decoder: ConvASRDecoderConfig = field(default_factory=lambda: ConvASRDecoderConfig())\n \n \n @dataclass\ndiff --git a/nemo/collections/asr/modules/audio_preprocessing.py b/nemo/collections/asr/modules/audio_preprocessing.py\n--- a/nemo/collections/asr/modules/audio_preprocessing.py\n+++ b/nemo/collections/asr/modules/audio_preprocessing.py\n@@ -634,6 +634,12 @@ def __init__(self, audio_length):\n         super(CropOrPadSpectrogramAugmentation, self).__init__()\n         self.audio_length = audio_length\n \n+        if self.audio_length < 0:\n+            raise ValueError(\n+                'audio_length must be non-negative. If using a dataclass with OmegaConf, '\n+                'please call OmegaConf.to_object(cfg) to call appropriate __post_init__ methods.'\n+            )\n+\n     @typecheck()\n     @torch.no_grad()\n     def forward(self, input_signal, length):\ndiff --git a/nemo/collections/asr/parts/k2/classes.py b/nemo/collections/asr/parts/k2/classes.py\n--- a/nemo/collections/asr/parts/k2/classes.py\n+++ b/nemo/collections/asr/parts/k2/classes.py\n@@ -13,7 +13,7 @@\n # limitations under the License.\n \n from abc import ABC\n-from dataclasses import dataclass\n+from dataclasses import dataclass, field\n from typing import Any, Optional, Tuple\n \n import torch\n@@ -43,7 +43,7 @@ class GraphModuleConfig:\n     topo_with_self_loops: bool = True\n     token_lm: Optional[Any] = None\n     intersect_pruned: bool = False\n-    intersect_conf: GraphIntersectDenseConfig = GraphIntersectDenseConfig()\n+    intersect_conf: GraphIntersectDenseConfig = field(default_factory=lambda: GraphIntersectDenseConfig())\n     boost_coeff: float = 0.0\n     predictor_window_size: int = 0\n     predictor_step_size: int = 1\ndiff --git a/nemo/collections/asr/parts/submodules/adapters/multi_head_attention_adapter_module.py b/nemo/collections/asr/parts/submodules/adapters/multi_head_attention_adapter_module.py\n--- a/nemo/collections/asr/parts/submodules/adapters/multi_head_attention_adapter_module.py\n+++ b/nemo/collections/asr/parts/submodules/adapters/multi_head_attention_adapter_module.py\n@@ -13,7 +13,7 @@\n # limitations under the License.\n \n import math\n-from dataclasses import dataclass\n+from dataclasses import dataclass, field\n from typing import Any, Optional\n \n import torch\n@@ -183,7 +183,7 @@ class MultiHeadAttentionAdapterConfig:\n     n_feat: int\n     dropout_rate: float = 0.0\n     proj_dim: Optional[int] = None\n-    adapter_strategy: Optional[Any] = MHAResidualAddAdapterStrategyConfig()\n+    adapter_strategy: Optional[Any] = field(default_factory=lambda: MHAResidualAddAdapterStrategyConfig())\n     _target_: str = \"{0}.{1}\".format(MultiHeadAttentionAdapter.__module__, MultiHeadAttentionAdapter.__name__)\n \n \n@@ -287,7 +287,7 @@ class RelPositionMultiHeadAttentionAdapterConfig:\n     n_feat: int\n     dropout_rate: float = 0.0\n     proj_dim: Optional[int] = None\n-    adapter_strategy: Optional[Any] = MHAResidualAddAdapterStrategyConfig()\n+    adapter_strategy: Optional[Any] = field(default_factory=lambda: MHAResidualAddAdapterStrategyConfig())\n     _target_: str = \"{0}.{1}\".format(\n         RelPositionMultiHeadAttentionAdapter.__module__, RelPositionMultiHeadAttentionAdapter.__name__\n     )\n@@ -336,7 +336,9 @@ class PositionalEncodingAdapterConfig:\n     d_model: int\n     max_len: int = 5000\n     xscale: float = 1.0\n-    adapter_strategy: Optional[Any] = adapter_mixin_strategies.ResidualAddAdapterStrategyConfig()\n+    adapter_strategy: Optional[Any] = field(\n+        default_factory=lambda: adapter_mixin_strategies.ResidualAddAdapterStrategyConfig()\n+    )\n     _target_: str = \"{0}.{1}\".format(PositionalEncodingAdapter.__module__, PositionalEncodingAdapter.__name__)\n \n \n@@ -378,5 +380,7 @@ class RelPositionalEncodingAdapterConfig:\n     d_model: int\n     max_len: int = 5000\n     xscale: float = 1.0\n-    adapter_strategy: Optional[Any] = adapter_mixin_strategies.ResidualAddAdapterStrategyConfig()\n+    adapter_strategy: Optional[Any] = field(\n+        default_factory=lambda: adapter_mixin_strategies.ResidualAddAdapterStrategyConfig()\n+    )\n     _target_: str = \"{0}.{1}\".format(RelPositionalEncodingAdapter.__module__, RelPositionalEncodingAdapter.__name__)\ndiff --git a/nemo/collections/asr/parts/submodules/ctc_beam_decoding.py b/nemo/collections/asr/parts/submodules/ctc_beam_decoding.py\n--- a/nemo/collections/asr/parts/submodules/ctc_beam_decoding.py\n+++ b/nemo/collections/asr/parts/submodules/ctc_beam_decoding.py\n@@ -14,7 +14,7 @@\n \n import math\n import os\n-from dataclasses import dataclass\n+from dataclasses import dataclass, field\n from typing import List, Optional, Tuple, Union\n \n import torch\n@@ -602,5 +602,5 @@ class BeamCTCInferConfig:\n     beam_beta: float = 0.0\n     kenlm_path: Optional[str] = None\n \n-    flashlight_cfg: Optional[FlashlightConfig] = FlashlightConfig()\n-    pyctcdecode_cfg: Optional[PyCTCDecodeConfig] = PyCTCDecodeConfig()\n+    flashlight_cfg: Optional[FlashlightConfig] = field(default_factory=lambda: FlashlightConfig())\n+    pyctcdecode_cfg: Optional[PyCTCDecodeConfig] = field(default_factory=lambda: PyCTCDecodeConfig())\ndiff --git a/nemo/collections/asr/parts/submodules/ctc_greedy_decoding.py b/nemo/collections/asr/parts/submodules/ctc_greedy_decoding.py\n--- a/nemo/collections/asr/parts/submodules/ctc_greedy_decoding.py\n+++ b/nemo/collections/asr/parts/submodules/ctc_greedy_decoding.py\n@@ -12,7 +12,7 @@\n # See the License for the specific language governing permissions and\n # limitations under the License.\n \n-from dataclasses import dataclass\n+from dataclasses import dataclass, field\n from typing import List, Optional\n \n import torch\n@@ -253,7 +253,9 @@ class GreedyCTCInferConfig:\n     preserve_alignments: bool = False\n     compute_timestamps: bool = False\n     preserve_frame_confidence: bool = False\n-    confidence_measure_cfg: Optional[ConfidenceMeasureConfig] = ConfidenceMeasureConfig()\n+    confidence_measure_cfg: Optional[ConfidenceMeasureConfig] = field(\n+        default_factory=lambda: ConfidenceMeasureConfig()\n+    )\n     confidence_method_cfg: str = \"DEPRECATED\"\n \n     def __post_init__(self):\ndiff --git a/nemo/collections/asr/parts/submodules/rnnt_greedy_decoding.py b/nemo/collections/asr/parts/submodules/rnnt_greedy_decoding.py\n--- a/nemo/collections/asr/parts/submodules/rnnt_greedy_decoding.py\n+++ b/nemo/collections/asr/parts/submodules/rnnt_greedy_decoding.py\n@@ -26,7 +26,7 @@\n # See the License for the specific language governing permissions and\n # limitations under the License.\n \n-from dataclasses import dataclass\n+from dataclasses import dataclass, field\n from typing import List, Optional, Tuple, Union\n \n import numpy as np\n@@ -2185,7 +2185,9 @@ class GreedyRNNTInferConfig:\n     max_symbols_per_step: Optional[int] = 10\n     preserve_alignments: bool = False\n     preserve_frame_confidence: bool = False\n-    confidence_measure_cfg: Optional[ConfidenceMeasureConfig] = ConfidenceMeasureConfig()\n+    confidence_measure_cfg: Optional[ConfidenceMeasureConfig] = field(\n+        default_factory=lambda: ConfidenceMeasureConfig()\n+    )\n     confidence_method_cfg: str = \"DEPRECATED\"\n \n     def __post_init__(self):\n@@ -2217,7 +2219,9 @@ class GreedyBatchedRNNTInferConfig:\n     max_symbols_per_step: Optional[int] = 10\n     preserve_alignments: bool = False\n     preserve_frame_confidence: bool = False\n-    confidence_measure_cfg: Optional[ConfidenceMeasureConfig] = ConfidenceMeasureConfig()\n+    confidence_measure_cfg: Optional[ConfidenceMeasureConfig] = field(\n+        default_factory=lambda: ConfidenceMeasureConfig()\n+    )\n     confidence_method_cfg: str = \"DEPRECATED\"\n \n     def __post_init__(self):\ndiff --git a/nemo/collections/asr/parts/utils/asr_confidence_utils.py b/nemo/collections/asr/parts/utils/asr_confidence_utils.py\n--- a/nemo/collections/asr/parts/utils/asr_confidence_utils.py\n+++ b/nemo/collections/asr/parts/utils/asr_confidence_utils.py\n@@ -14,7 +14,7 @@\n \n import math\n from abc import ABC, abstractmethod\n-from dataclasses import dataclass\n+from dataclasses import dataclass, field\n from functools import partial\n from typing import List, Optional\n \n@@ -181,7 +181,7 @@ class ConfidenceConfig:\n     preserve_word_confidence: bool = False\n     exclude_blank: bool = True\n     aggregation: str = \"min\"\n-    measure_cfg: ConfidenceMeasureConfig = ConfidenceMeasureConfig()\n+    measure_cfg: ConfidenceMeasureConfig = field(default_factory=lambda: ConfidenceMeasureConfig())\n     method_cfg: str = \"DEPRECATED\"\n \n     def __post_init__(self):\ndiff --git a/nemo/collections/common/parts/adapter_modules.py b/nemo/collections/common/parts/adapter_modules.py\n--- a/nemo/collections/common/parts/adapter_modules.py\n+++ b/nemo/collections/common/parts/adapter_modules.py\n@@ -12,7 +12,7 @@\n # See the License for the specific language governing permissions and\n # limitations under the License.\n \n-from dataclasses import dataclass, is_dataclass\n+from dataclasses import dataclass, field, is_dataclass\n from typing import Any, Optional\n \n from hydra.utils import instantiate\n@@ -160,5 +160,7 @@ class LinearAdapterConfig:\n     activation: str = 'swish'\n     norm_position: str = 'pre'\n     dropout: float = 0.0\n-    adapter_strategy: Optional[Any] = adapter_mixin_strategies.ResidualAddAdapterStrategyConfig()\n+    adapter_strategy: Optional[Any] = field(\n+        default_factory=lambda: adapter_mixin_strategies.ResidualAddAdapterStrategyConfig()\n+    )\n     _target_: str = \"{0}.{1}\".format(LinearAdapter.__module__, LinearAdapter.__name__)\ndiff --git a/nemo/collections/common/tokenizers/en_ja_tokenizers.py b/nemo/collections/common/tokenizers/en_ja_tokenizers.py\n--- a/nemo/collections/common/tokenizers/en_ja_tokenizers.py\n+++ b/nemo/collections/common/tokenizers/en_ja_tokenizers.py\n@@ -14,11 +14,19 @@\n import re\n from typing import List\n \n-import ipadic\n-import MeCab\n from pangu import spacing\n from sacremoses import MosesDetokenizer, MosesPunctNormalizer, MosesTokenizer\n \n+try:\n+    import ipadic\n+    import MeCab\n+\n+    HAVE_MECAB = True\n+    HAVE_IPADIC = True\n+except (ImportError, ModuleNotFoundError):\n+    HAVE_MECAB = False\n+    HAVE_IPADIC = False\n+\n \n class EnJaProcessor:\n     \"\"\"\n@@ -67,6 +75,9 @@ class JaMecabProcessor:\n     \"\"\"\n \n     def __init__(self):\n+        if not HAVE_MECAB or not HAVE_IPADIC:\n+            raise ImportError(\"Please ensure that you have installed `MeCab` and `ipadic` to use JaMecabProcessor\")\n+\n         self.mecab_tokenizer = MeCab.Tagger(ipadic.MECAB_ARGS + \" -Owakati\")\n \n     def detokenize(self, text: List[str]) -> str:\ndiff --git a/nemo/collections/nlp/models/machine_translation/mt_enc_dec_config.py b/nemo/collections/nlp/models/machine_translation/mt_enc_dec_config.py\n--- a/nemo/collections/nlp/models/machine_translation/mt_enc_dec_config.py\n+++ b/nemo/collections/nlp/models/machine_translation/mt_enc_dec_config.py\n@@ -12,7 +12,7 @@\n # See the License for the specific language governing permissions and\n # limitations under the License.\n \n-from dataclasses import dataclass\n+from dataclasses import dataclass, field\n from typing import Any, Optional, Tuple\n \n from omegaconf.omegaconf import MISSING\n@@ -46,7 +46,7 @@ class MTOptimConfig(OptimConfig):\n     lr: float = 1e-3\n     betas: Tuple[float, float] = (0.9, 0.98)\n     weight_decay: float = 0.0\n-    sched: Optional[MTSchedConfig] = MTSchedConfig()\n+    sched: Optional[MTSchedConfig] = field(default_factory=lambda: MTSchedConfig())\n \n \n @dataclass\n@@ -74,70 +74,80 @@ class MTEncDecModelConfig(EncDecNLPModelConfig):\n     decoder_tokenizer: Any = MISSING\n     decoder: Any = MISSING\n \n-    head: TokenClassifierConfig = TokenClassifierConfig(log_softmax=True)\n+    head: TokenClassifierConfig = field(default_factory=lambda: TokenClassifierConfig(log_softmax=True))\n \n     # dataset configurations\n-    train_ds: Optional[TranslationDataConfig] = TranslationDataConfig(\n-        src_file_name=MISSING,\n-        tgt_file_name=MISSING,\n-        tokens_in_batch=512,\n-        clean=True,\n-        shuffle=True,\n-        cache_ids=False,\n-        use_cache=False,\n+    train_ds: Optional[TranslationDataConfig] = field(\n+        default_factory=lambda: TranslationDataConfig(\n+            src_file_name=MISSING,\n+            tgt_file_name=MISSING,\n+            tokens_in_batch=512,\n+            clean=True,\n+            shuffle=True,\n+            cache_ids=False,\n+            use_cache=False,\n+        )\n     )\n-    validation_ds: Optional[TranslationDataConfig] = TranslationDataConfig(\n-        src_file_name=MISSING,\n-        tgt_file_name=MISSING,\n-        tokens_in_batch=512,\n-        clean=False,\n-        shuffle=False,\n-        cache_ids=False,\n-        use_cache=False,\n+    validation_ds: Optional[TranslationDataConfig] = field(\n+        default_factory=lambda: TranslationDataConfig(\n+            src_file_name=MISSING,\n+            tgt_file_name=MISSING,\n+            tokens_in_batch=512,\n+            clean=False,\n+            shuffle=False,\n+            cache_ids=False,\n+            use_cache=False,\n+        )\n     )\n-    test_ds: Optional[TranslationDataConfig] = TranslationDataConfig(\n-        src_file_name=MISSING,\n-        tgt_file_name=MISSING,\n-        tokens_in_batch=512,\n-        clean=False,\n-        shuffle=False,\n-        cache_ids=False,\n-        use_cache=False,\n+    test_ds: Optional[TranslationDataConfig] = field(\n+        default_factory=lambda: TranslationDataConfig(\n+            src_file_name=MISSING,\n+            tgt_file_name=MISSING,\n+            tokens_in_batch=512,\n+            clean=False,\n+            shuffle=False,\n+            cache_ids=False,\n+            use_cache=False,\n+        )\n     )\n-    optim: Optional[OptimConfig] = MTOptimConfig()\n+    optim: Optional[OptimConfig] = field(default_factory=lambda: MTOptimConfig())\n \n \n @dataclass\n class AAYNBaseConfig(MTEncDecModelConfig):\n \n     # Attention is All You Need Base Configuration\n-    encoder_tokenizer: TokenizerConfig = TokenizerConfig(library='yttm')\n-    decoder_tokenizer: TokenizerConfig = TokenizerConfig(library='yttm')\n-\n-    encoder: NeMoTransformerEncoderConfig = NeMoTransformerEncoderConfig(\n-        library='nemo',\n-        model_name=None,\n-        pretrained=False,\n-        hidden_size=512,\n-        inner_size=2048,\n-        num_layers=6,\n-        num_attention_heads=8,\n-        ffn_dropout=0.1,\n-        attn_score_dropout=0.1,\n-        attn_layer_dropout=0.1,\n+    encoder_tokenizer: TokenizerConfig = field(default_factory=lambda: TokenizerConfig(library='yttm'))\n+    decoder_tokenizer: TokenizerConfig = field(default_factory=lambda: TokenizerConfig(library='yttm'))\n+\n+    encoder: NeMoTransformerEncoderConfig = field(\n+        default_factory=lambda: NeMoTransformerEncoderConfig(\n+            library='nemo',\n+            model_name=None,\n+            pretrained=False,\n+            hidden_size=512,\n+            inner_size=2048,\n+            num_layers=6,\n+            num_attention_heads=8,\n+            ffn_dropout=0.1,\n+            attn_score_dropout=0.1,\n+            attn_layer_dropout=0.1,\n+        )\n     )\n \n-    decoder: NeMoTransformerConfig = NeMoTransformerConfig(\n-        library='nemo',\n-        model_name=None,\n-        pretrained=False,\n-        hidden_size=512,\n-        inner_size=2048,\n-        num_layers=6,\n-        num_attention_heads=8,\n-        ffn_dropout=0.1,\n-        attn_score_dropout=0.1,\n-        attn_layer_dropout=0.1,\n+    decoder: NeMoTransformerConfig = field(\n+        default_factory=lambda: NeMoTransformerConfig(\n+            library='nemo',\n+            model_name=None,\n+            pretrained=False,\n+            hidden_size=512,\n+            inner_size=2048,\n+            num_layers=6,\n+            num_attention_heads=8,\n+            ffn_dropout=0.1,\n+            attn_score_dropout=0.1,\n+            attn_layer_dropout=0.1,\n+        )\n     )\n \n \n@@ -150,32 +160,36 @@ class MTBottleneckModelConfig(AAYNBaseConfig):\n     recon_per_token: bool = True\n     log_timing: bool = True\n \n-    encoder: NeMoTransformerBottleneckEncoderConfig = NeMoTransformerBottleneckEncoderConfig(\n-        library='nemo',\n-        model_name=None,\n-        pretrained=False,\n-        hidden_size=512,\n-        inner_size=2048,\n-        num_layers=6,\n-        num_attention_heads=8,\n-        ffn_dropout=0.1,\n-        attn_score_dropout=0.1,\n-        attn_layer_dropout=0.1,\n-        arch='seq2seq',\n-        hidden_steps=32,\n-        hidden_blocks=1,\n-        hidden_init_method='params',\n+    encoder: NeMoTransformerBottleneckEncoderConfig = field(\n+        default_factory=lambda: NeMoTransformerBottleneckEncoderConfig(\n+            library='nemo',\n+            model_name=None,\n+            pretrained=False,\n+            hidden_size=512,\n+            inner_size=2048,\n+            num_layers=6,\n+            num_attention_heads=8,\n+            ffn_dropout=0.1,\n+            attn_score_dropout=0.1,\n+            attn_layer_dropout=0.1,\n+            arch='seq2seq',\n+            hidden_steps=32,\n+            hidden_blocks=1,\n+            hidden_init_method='params',\n+        )\n     )\n \n-    decoder: NeMoTransformerBottleneckDecoderConfig = NeMoTransformerBottleneckDecoderConfig(\n-        library='nemo',\n-        model_name=None,\n-        pretrained=False,\n-        inner_size=2048,\n-        num_layers=6,\n-        num_attention_heads=8,\n-        ffn_dropout=0.1,\n-        attn_score_dropout=0.1,\n-        attn_layer_dropout=0.1,\n-        arch='seq2seq',\n+    decoder: NeMoTransformerBottleneckDecoderConfig = field(\n+        default_factory=lambda: NeMoTransformerBottleneckDecoderConfig(\n+            library='nemo',\n+            model_name=None,\n+            pretrained=False,\n+            inner_size=2048,\n+            num_layers=6,\n+            num_attention_heads=8,\n+            ffn_dropout=0.1,\n+            attn_score_dropout=0.1,\n+            attn_layer_dropout=0.1,\n+            arch='seq2seq',\n+        )\n     )\ndiff --git a/nemo/collections/nlp/models/token_classification/punctuation_capitalization_config.py b/nemo/collections/nlp/models/token_classification/punctuation_capitalization_config.py\n--- a/nemo/collections/nlp/models/token_classification/punctuation_capitalization_config.py\n+++ b/nemo/collections/nlp/models/token_classification/punctuation_capitalization_config.py\n@@ -12,7 +12,7 @@\n # See the License for the specific language governing permissions and\n # limitations under the License.\n \n-from dataclasses import dataclass\n+from dataclasses import dataclass, field\n from typing import Any, Dict, Optional\n \n from omegaconf.omegaconf import MISSING, DictConfig, OmegaConf, open_dict\n@@ -215,13 +215,15 @@ class PunctuationCapitalizationModelConfig:\n     This config is a part of :class:`~PunctuationCapitalizationConfig`.\n     \"\"\"\n \n-    class_labels: ClassLabelsConfig = ClassLabelsConfig()\n+    class_labels: ClassLabelsConfig = field(default_factory=lambda: ClassLabelsConfig())\n     \"\"\"A mandatory parameter containing a dictionary with names of label id files used in .nemo checkpoints.\n     These file names can also be used for passing label vocabularies to the model. If you wish to use ``class_labels``\n     for passing vocabularies, please provide path to vocabulary files in\n     ``model.common_dataset_parameters.label_vocab_dir`` parameter.\"\"\"\n \n-    common_dataset_parameters: Optional[CommonDatasetParametersConfig] = CommonDatasetParametersConfig()\n+    common_dataset_parameters: Optional[CommonDatasetParametersConfig] = field(\n+        default_factory=lambda: CommonDatasetParametersConfig()\n+    )\n     \"\"\"Label ids and loss mask information information.\"\"\"\n \n     train_ds: Optional[PunctuationCapitalizationTrainDataConfig] = None\n@@ -233,16 +235,16 @@ class PunctuationCapitalizationModelConfig:\n     test_ds: Optional[PunctuationCapitalizationEvalDataConfig] = None\n     \"\"\"A configuration for creating test datasets and data loaders.\"\"\"\n \n-    punct_head: HeadConfig = HeadConfig()\n+    punct_head: HeadConfig = field(default_factory=lambda: HeadConfig())\n     \"\"\"A configuration for creating punctuation MLP head that is applied to a language model outputs.\"\"\"\n \n-    capit_head: HeadConfig = HeadConfig()\n+    capit_head: HeadConfig = field(default_factory=lambda: HeadConfig())\n     \"\"\"A configuration for creating capitalization MLP head that is applied to a language model outputs.\"\"\"\n \n-    tokenizer: Any = TokenizerConfig()\n+    tokenizer: Any = field(default_factory=lambda: TokenizerConfig())\n     \"\"\"A configuration for source text tokenizer.\"\"\"\n \n-    language_model: LanguageModelConfig = LanguageModelConfig()\n+    language_model: LanguageModelConfig = field(default_factory=lambda: LanguageModelConfig())\n     \"\"\"A configuration of a BERT-like language model which serves as a model body.\"\"\"\n \n     optim: Optional[Any] = None\n@@ -311,22 +313,30 @@ class PunctuationCapitalizationConfig(NemoConfig):\n     do_testing: bool = False\n     \"\"\"Whether ot perform testing of the model.\"\"\"\n \n-    model: PunctuationCapitalizationModelConfig = PunctuationCapitalizationModelConfig()\n+    model: PunctuationCapitalizationModelConfig = field(default_factory=lambda: PunctuationCapitalizationModelConfig())\n     \"\"\"A configuration for the\n     :class:`~nemo.collections.nlp.models.token_classification.punctuation_capitalization_model.PunctuationCapitalizationModel`\n     model.\"\"\"\n \n-    trainer: Optional[TrainerConfig] = TrainerConfig()\n+    trainer: Optional[TrainerConfig] = field(default_factory=lambda: TrainerConfig())\n     \"\"\"Contains ``Trainer`` Lightning class constructor parameters.\"\"\"\n \n-    exp_manager: Optional[ExpManagerConfig] = ExpManagerConfig(name=name, files_to_copy=[])\n+    exp_manager: Optional[ExpManagerConfig] = field(\n+        default_factory=lambda: ExpManagerConfig(name=None, files_to_copy=[])\n+    )\n     \"\"\"A configuration with various NeMo training options such as output directories, resuming from checkpoint,\n     tensorboard and W&B logging, and so on. For possible options see :ref:`exp-manager-label`.\"\"\"\n \n+    def __post_init__(self):\n+        if self.exp_manager is not None:\n+            self.exp_manager.name = self.name\n+\n \n @dataclass\n class PunctuationCapitalizationLexicalAudioConfig(PunctuationCapitalizationConfig):\n-    model: PunctuationCapitalizationLexicalAudioModelConfig = PunctuationCapitalizationLexicalAudioModelConfig()\n+    model: PunctuationCapitalizationLexicalAudioModelConfig = field(\n+        default_factory=lambda: PunctuationCapitalizationLexicalAudioModelConfig()\n+    )\n \n \n def is_legacy_model_config(model_cfg: DictConfig) -> bool:\ndiff --git a/nemo/collections/nlp/modules/common/megatron/megatron_encoders.py b/nemo/collections/nlp/modules/common/megatron/megatron_encoders.py\n--- a/nemo/collections/nlp/modules/common/megatron/megatron_encoders.py\n+++ b/nemo/collections/nlp/modules/common/megatron/megatron_encoders.py\n@@ -13,7 +13,6 @@\n # limitations under the License.\n \n \"\"\"Transformer based language model.\"\"\"\n-from MeCab import Model\n from nemo.collections.nlp.modules.common.megatron.megatron_perceiver_encoders import MegatronPerceiverEncoderModule\n from nemo.collections.nlp.modules.common.megatron.megatron_transformer_encoder import MegatronTransformerEncoderModule\n from nemo.collections.nlp.modules.common.megatron.retrieval_transformer import (\n@@ -25,6 +24,13 @@\n     scaled_init_method_normal,\n )\n \n+try:\n+    from MeCab import Model\n+\n+    HAVE_MECAB = True\n+except (ImportError, ModuleNotFoundError):\n+    HAVE_MECAB = False\n+\n try:\n     from apex.transformer.enums import AttnMaskType, ModelType\n \ndiff --git a/nemo/collections/tts/models/fastpitch.py b/nemo/collections/tts/models/fastpitch.py\n--- a/nemo/collections/tts/models/fastpitch.py\n+++ b/nemo/collections/tts/models/fastpitch.py\n@@ -12,7 +12,7 @@\n # See the License for the specific language governing permissions and\n # limitations under the License.\n import contextlib\n-from dataclasses import dataclass\n+from dataclasses import dataclass, field\n from pathlib import Path\n from typing import List, Optional\n \n@@ -70,12 +70,12 @@ class TextTokenizer:\n     apostrophe: bool = True\n     pad_with_space: bool = True\n     add_blank_at: bool = True\n-    g2p: G2PConfig = G2PConfig()\n+    g2p: G2PConfig = field(default_factory=lambda: G2PConfig())\n \n \n @dataclass\n class TextTokenizerConfig:\n-    text_tokenizer: TextTokenizer = TextTokenizer()\n+    text_tokenizer: TextTokenizer = field(default_factory=lambda: TextTokenizer())\n \n \n class FastPitchModel(SpectrogramGenerator, Exportable, FastPitchAdapterModelMixin):\ndiff --git a/nemo/collections/tts/models/tacotron2.py b/nemo/collections/tts/models/tacotron2.py\n--- a/nemo/collections/tts/models/tacotron2.py\n+++ b/nemo/collections/tts/models/tacotron2.py\n@@ -13,7 +13,7 @@\n # limitations under the License.\n \n import contextlib\n-from dataclasses import dataclass\n+from dataclasses import dataclass, field\n from typing import Any, Dict, List, Optional\n \n import torch\n@@ -53,7 +53,7 @@ class Preprocessor:\n \n @dataclass\n class Tacotron2Config:\n-    preprocessor: Preprocessor = Preprocessor()\n+    preprocessor: Preprocessor = field(default_factory=lambda: Preprocessor())\n     encoder: Dict[Any, Any] = MISSING\n     decoder: Dict[Any, Any] = MISSING\n     postnet: Dict[Any, Any] = MISSING\ndiff --git a/nemo/core/config/modelPT.py b/nemo/core/config/modelPT.py\n--- a/nemo/core/config/modelPT.py\n+++ b/nemo/core/config/modelPT.py\n@@ -58,11 +58,13 @@ class HydraConfig:\n class NemoConfig:\n     name: str = MISSING\n     model: ModelConfig = MISSING\n-    trainer: config.TrainerConfig = config.TrainerConfig(\n-        strategy=\"ddp\", enable_checkpointing=False, logger=False, log_every_n_steps=1, accelerator='gpu'\n+    trainer: config.TrainerConfig = field(\n+        default_factory=lambda: config.TrainerConfig(\n+            strategy=\"ddp\", enable_checkpointing=False, logger=False, log_every_n_steps=1, accelerator='gpu'\n+        )\n     )\n-    exp_manager: Optional[Any] = exp_manager.ExpManagerConfig()\n-    hydra: HydraConfig = HydraConfig()\n+    exp_manager: Optional[Any] = field(default_factory=lambda: exp_manager.ExpManagerConfig())\n+    hydra: HydraConfig = field(default_factory=lambda: HydraConfig())\n \n \n class ModelConfigBuilder:\ndiff --git a/nemo/utils/exp_manager.py b/nemo/utils/exp_manager.py\n--- a/nemo/utils/exp_manager.py\n+++ b/nemo/utils/exp_manager.py\n@@ -18,7 +18,7 @@\n import sys\n import time\n import warnings\n-from dataclasses import dataclass\n+from dataclasses import dataclass, field\n from datetime import timedelta\n from pathlib import Path\n from shutil import copy, move\n@@ -146,28 +146,30 @@ class ExpManagerConfig:\n     create_wandb_logger: Optional[bool] = False\n     wandb_logger_kwargs: Optional[Dict[Any, Any]] = None\n     create_mlflow_logger: Optional[bool] = False\n-    mlflow_logger_kwargs: Optional[MLFlowParams] = MLFlowParams()\n+    mlflow_logger_kwargs: Optional[MLFlowParams] = field(default_factory=lambda: MLFlowParams())\n     create_dllogger_logger: Optional[bool] = False\n-    dllogger_logger_kwargs: Optional[DLLoggerParams] = DLLoggerParams()\n+    dllogger_logger_kwargs: Optional[DLLoggerParams] = field(default_factory=lambda: DLLoggerParams())\n     create_clearml_logger: Optional[bool] = False\n-    clearml_logger_kwargs: Optional[ClearMLParams] = ClearMLParams()\n+    clearml_logger_kwargs: Optional[ClearMLParams] = field(default_factory=lambda: ClearMLParams())\n     # Checkpointing parameters\n     create_checkpoint_callback: Optional[bool] = True\n-    checkpoint_callback_params: Optional[CallbackParams] = CallbackParams()\n+    checkpoint_callback_params: Optional[CallbackParams] = field(default_factory=lambda: CallbackParams())\n     create_early_stopping_callback: Optional[bool] = False\n-    early_stopping_callback_params: Optional[EarlyStoppingParams] = EarlyStoppingParams()\n+    early_stopping_callback_params: Optional[EarlyStoppingParams] = field(\n+        default_factory=lambda: EarlyStoppingParams()\n+    )\n     create_preemption_callback: Optional[bool] = True\n     # Additional exp_manager arguments\n     files_to_copy: Optional[List[str]] = None\n     # logs timing of train/val/test steps\n     log_step_timing: Optional[bool] = True\n-    step_timing_kwargs: Optional[StepTimingParams] = StepTimingParams()\n+    step_timing_kwargs: Optional[StepTimingParams] = field(default_factory=lambda: StepTimingParams())\n     # Configures creation of log files for different ranks\n     log_local_rank_0_only: Optional[bool] = False\n     log_global_rank_0_only: Optional[bool] = False\n     # disable initial validation when resuming from a checkpoint saved during validation\n     disable_validation_on_resume: Optional[bool] = True\n-    ema: Optional[EMAParams] = EMAParams()\n+    ema: Optional[EMAParams] = field(default_factory=lambda: EMAParams())\n     # Wall clock time limit\n     max_time_per_run: Optional[str] = None\n     # time to sleep non 0 ranks during initialization\ndiff --git a/scripts/asr_language_modeling/ngram_lm/eval_beamsearch_ngram.py b/scripts/asr_language_modeling/ngram_lm/eval_beamsearch_ngram.py\n--- a/scripts/asr_language_modeling/ngram_lm/eval_beamsearch_ngram.py\n+++ b/scripts/asr_language_modeling/ngram_lm/eval_beamsearch_ngram.py\n@@ -112,14 +112,14 @@ class EvalBeamSearchNGramConfig:\n     beam_beta: List[float] = field(default_factory=lambda: [0.0])  # The beta parameter or list of the betas for the beam search decoding\n \n     decoding_strategy: str = \"beam\"\n-    decoding: ctc_beam_decoding.BeamCTCInferConfig = ctc_beam_decoding.BeamCTCInferConfig(beam_size=128)\n+    decoding: ctc_beam_decoding.BeamCTCInferConfig = field(default_factory=lambda: ctc_beam_decoding.BeamCTCInferConfig(beam_size=128))\n     \n-    text_processing: Optional[TextProcessingConfig] = TextProcessingConfig(\n+    text_processing: Optional[TextProcessingConfig] = field(default_factory=lambda: TextProcessingConfig(\n         punctuation_marks = \".,?\",\n         separate_punctuation = False,\n         do_lowercase = False,\n         rm_punctuation = False,\n-    )\n+    ))\n # fmt: on\n \n \ndiff --git a/scripts/asr_language_modeling/ngram_lm/eval_beamsearch_ngram_transducer.py b/scripts/asr_language_modeling/ngram_lm/eval_beamsearch_ngram_transducer.py\n--- a/scripts/asr_language_modeling/ngram_lm/eval_beamsearch_ngram_transducer.py\n+++ b/scripts/asr_language_modeling/ngram_lm/eval_beamsearch_ngram_transducer.py\n@@ -115,7 +115,7 @@ class EvalBeamSearchNGramConfig:\n     hat_subtract_ilm: bool = False\n     hat_ilm_weight: List[float] = field(default_factory=lambda: [0.0])\n \n-    decoding: rnnt_beam_decoding.BeamRNNTInferConfig = rnnt_beam_decoding.BeamRNNTInferConfig(beam_size=128)\n+    decoding: rnnt_beam_decoding.BeamRNNTInferConfig = field(default_factory=lambda: rnnt_beam_decoding.BeamRNNTInferConfig(beam_size=128))\n \n \n # fmt: on\ndiff --git a/scripts/confidence_ensembles/build_ensemble.py b/scripts/confidence_ensembles/build_ensemble.py\n--- a/scripts/confidence_ensembles/build_ensemble.py\n+++ b/scripts/confidence_ensembles/build_ensemble.py\n@@ -75,7 +75,7 @@\n import sys\n import tempfile\n from copy import deepcopy\n-from dataclasses import dataclass\n+from dataclasses import dataclass, field\n from pathlib import Path\n from typing import Dict, List, Optional, Tuple\n \n@@ -209,19 +209,23 @@ class BuildEnsembleConfig:\n     random_seed: int = 0  # for reproducibility\n \n     # default confidence, can override\n-    confidence: ConfidenceConfig = ConfidenceConfig(\n-        # we keep frame confidences and apply aggregation manually to get full-utterance confidence\n-        preserve_frame_confidence=True,\n-        exclude_blank=True,\n-        aggregation=\"mean\",\n-        measure_cfg=ConfidenceMeasureConfig(name=\"entropy\", entropy_type=\"renyi\", alpha=0.25, entropy_norm=\"lin\",),\n+    confidence: ConfidenceConfig = field(\n+        default_factory=lambda: ConfidenceConfig(\n+            # we keep frame confidences and apply aggregation manually to get full-utterance confidence\n+            preserve_frame_confidence=True,\n+            exclude_blank=True,\n+            aggregation=\"mean\",\n+            measure_cfg=ConfidenceMeasureConfig(name=\"entropy\", entropy_type=\"renyi\", alpha=0.25, entropy_norm=\"lin\",),\n+        )\n     )\n     temperature: float = 1.0\n \n     # this is optional, but can be used to change any aspect of the transcription\n     # config, such as batch size or amp usage. Note that model, data and confidence\n     # will be overriden by this script\n-    transcription: transcribe_speech.TranscriptionConfig = transcribe_speech.TranscriptionConfig()\n+    transcription: transcribe_speech.TranscriptionConfig = field(\n+        default_factory=lambda: transcribe_speech.TranscriptionConfig()\n+    )\n \n     # set to True to tune the confidence.\n     # requires dev manifests to be specified for each model\n@@ -229,12 +233,14 @@ class BuildEnsembleConfig:\n     # used to specify what to tune over. By default runs tuning over some\n     # reasonalbe grid, so that it does not take forever.\n     # Can be changed as needed\n-    tune_confidence_config: TuneConfidenceConfig = TuneConfidenceConfig()\n+    tune_confidence_config: TuneConfidenceConfig = field(default_factory=lambda: TuneConfidenceConfig())\n \n     # very fast to tune and can be important in case of imbalanced datasets\n     # will automatically set to False if dev data is not available\n     tune_logistic_regression: bool = True\n-    tune_logistic_regression_config: TuneLogisticRegressionConfig = TuneLogisticRegressionConfig()\n+    tune_logistic_regression_config: TuneLogisticRegressionConfig = field(\n+        default_factory=lambda: TuneLogisticRegressionConfig()\n+    )\n \n     def __post_init__(self):\n         \"\"\"Checking that if any dev data is provided, all are provided.\ndiff --git a/scripts/speech_recognition/confidence/benchmark_asr_confidence.py b/scripts/speech_recognition/confidence/benchmark_asr_confidence.py\n--- a/scripts/speech_recognition/confidence/benchmark_asr_confidence.py\n+++ b/scripts/speech_recognition/confidence/benchmark_asr_confidence.py\n@@ -14,7 +14,7 @@\n \n import json\n import os\n-from dataclasses import dataclass, is_dataclass\n+from dataclasses import dataclass, field, is_dataclass\n from pathlib import Path\n from typing import Optional\n \n@@ -124,7 +124,9 @@ class ConfidenceBenchmarkingConfig:\n \n     # Confidence configs\n     target_level: str = \"auto\"  # Choices: \"word\", \"token\", \"auto\" (for both word- and token-level confidence)\n-    confidence_cfg: ConfidenceConfig = ConfidenceConfig(preserve_word_confidence=True, preserve_token_confidence=True)\n+    confidence_cfg: ConfidenceConfig = field(\n+        default_factory=lambda: ConfidenceConfig(preserve_word_confidence=True, preserve_token_confidence=True)\n+    )\n     grid_params: Optional[str] = None  # a dictionary with lists of parameters to iteratively benchmark on\n \n \ndiff --git a/scripts/speech_recognition/convert_to_tarred_audio_dataset.py b/scripts/speech_recognition/convert_to_tarred_audio_dataset.py\n--- a/scripts/speech_recognition/convert_to_tarred_audio_dataset.py\n+++ b/scripts/speech_recognition/convert_to_tarred_audio_dataset.py\n@@ -202,7 +202,7 @@ class ASRTarredDatasetMetadata:\n     num_samples_per_shard: Optional[int] = None\n     is_concatenated_manifest: bool = False\n \n-    dataset_config: Optional[ASRTarredDatasetConfig] = ASRTarredDatasetConfig()\n+    dataset_config: Optional[ASRTarredDatasetConfig] = field(default_factory=lambda: ASRTarredDatasetConfig())\n     history: Optional[List[Any]] = field(default_factory=lambda: [])\n \n     def __post_init__(self):\ndiff --git a/tools/nemo_forced_aligner/align.py b/tools/nemo_forced_aligner/align.py\n--- a/tools/nemo_forced_aligner/align.py\n+++ b/tools/nemo_forced_aligner/align.py\n@@ -149,8 +149,8 @@ class AlignmentConfig:\n \n     # Output file configs\n     save_output_file_formats: List[str] = field(default_factory=lambda: [\"ctm\", \"ass\"])\n-    ctm_file_config: CTMFileConfig = CTMFileConfig()\n-    ass_file_config: ASSFileConfig = ASSFileConfig()\n+    ctm_file_config: CTMFileConfig = field(default_factory=lambda: CTMFileConfig())\n+    ass_file_config: ASSFileConfig = field(default_factory=lambda: ASSFileConfig())\n \n \n @hydra_runner(config_name=\"AlignmentConfig\", schema=AlignmentConfig)\n",
-    "test_patch": "diff --git a/tests/collections/asr/test_text_to_text_dataset.py b/tests/collections/asr/test_text_to_text_dataset.py\n--- a/tests/collections/asr/test_text_to_text_dataset.py\n+++ b/tests/collections/asr/test_text_to_text_dataset.py\n@@ -15,7 +15,7 @@\n import json\n import multiprocessing\n import os\n-from dataclasses import dataclass\n+from dataclasses import dataclass, field\n from pathlib import Path\n \n import pytest\n@@ -110,7 +110,7 @@ class TextTokenizerCfg:\n         apostrophe: bool = True\n         pad_with_space: bool = True\n         add_blank_at: bool = True\n-        g2p: G2PConfig = G2PConfig()\n+        g2p: G2PConfig = field(default_factory=lambda: G2PConfig())\n \n     config = OmegaConf.create(OmegaConf.to_yaml(TextTokenizerCfg()))\n     return instantiate(config)\n",
-    "problem_statement": "Ubuntu 22.04 Python 3.11 [asr]: multiple errors `dataclasses ValueError: mutable default * for field * is not allowed: use default_factory`\n**Describe the bug**\r\n\r\nAfter installing latest stable `1.19.1` from pipI, or the latest current commit with `[asr]` extras, I'm getting this error from `hydra-core==1.2.0` when trying to import `nemo.collections.asr`:\r\n```\r\nTraceback (most recent call last):\r\n  File \"/home/alex/T7/src/speaker_verification/nemo_speaker/test.py\", line 5, in <module>\r\n    import nemo.collections.asr as nemo_asr\r\n  File \"/home/alex/T7/src/speaker_verification/nemo_speaker/venv11/lib/python3.11/site-packages/nemo/collections/asr/__init__.py\", line 15, in <module>\r\n    from nemo.collections.asr import data, losses, models, modules\r\n  File \"/home/alex/T7/src/speaker_verification/nemo_speaker/venv11/lib/python3.11/site-packages/nemo/collections/asr/losses/__init__.py\", line 15, in <module>\r\n    from nemo.collections.asr.losses.angularloss import AngularSoftmaxLoss\r\n  File \"/home/alex/T7/src/speaker_verification/nemo_speaker/venv11/lib/python3.11/site-packages/nemo/collections/asr/losses/angularloss.py\", line 18, in <module>\r\n    from nemo.core.classes import Loss, Typing, typecheck\r\n  File \"/home/alex/T7/src/speaker_verification/nemo_speaker/venv11/lib/python3.11/site-packages/nemo/core/__init__.py\", line 16, in <module>\r\n    from nemo.core.classes import *\r\n  File \"/home/alex/T7/src/speaker_verification/nemo_speaker/venv11/lib/python3.11/site-packages/nemo/core/classes/__init__.py\", line 16, in <module>\r\n    import hydra\r\n  File \"/home/alex/T7/src/speaker_verification/nemo_speaker/venv11/lib/python3.11/site-packages/hydra/__init__.py\", line 5, in <module>\r\n    from hydra import utils\r\n  File \"/home/alex/T7/src/speaker_verification/nemo_speaker/venv11/lib/python3.11/site-packages/hydra/utils.py\", line 8, in <module>\r\n    import hydra._internal.instantiate._instantiate2\r\n  File \"/home/alex/T7/src/speaker_verification/nemo_speaker/venv11/lib/python3.11/site-packages/hydra/_internal/instantiate/_instantiate2.py\", line 12, in <module>\r\n    from hydra._internal.utils import _locate\r\n  File \"/home/alex/T7/src/speaker_verification/nemo_speaker/venv11/lib/python3.11/site-packages/hydra/_internal/utils.py\", line 18, in <module>\r\n    from hydra.core.utils import get_valid_filename, validate_config_path\r\n  File \"/home/alex/T7/src/speaker_verification/nemo_speaker/venv11/lib/python3.11/site-packages/hydra/core/utils.py\", line 20, in <module>\r\n    from hydra.core.hydra_config import HydraConfig\r\n  File \"/home/alex/T7/src/speaker_verification/nemo_speaker/venv11/lib/python3.11/site-packages/hydra/core/hydra_config.py\", line 6, in <module>\r\n    from hydra.conf import HydraConf\r\n  File \"/home/alex/T7/src/speaker_verification/nemo_speaker/venv11/lib/python3.11/site-packages/hydra/conf/__init__.py\", line 46, in <module>\r\n    class JobConf:\r\n  File \"/home/alex/T7/src/speaker_verification/nemo_speaker/venv11/lib/python3.11/site-packages/hydra/conf/__init__.py\", line 75, in JobConf\r\n    @dataclass\r\n     ^^^^^^^^^\r\n  File \"/usr/lib/python3.11/dataclasses.py\", line 1230, in dataclass\r\n    return wrap(cls)\r\n           ^^^^^^^^^\r\n  File \"/usr/lib/python3.11/dataclasses.py\", line 1220, in wrap\r\n    return _process_class(cls, init, repr, eq, order, unsafe_hash,\r\n           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\r\n  File \"/usr/lib/python3.11/dataclasses.py\", line 958, in _process_class\r\n    cls_fields.append(_get_field(cls, name, type, kw_only))\r\n                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\r\n  File \"/usr/lib/python3.11/dataclasses.py\", line 815, in _get_field\r\n    raise ValueError(f'mutable default {type(f.default)} for field '\r\nValueError: mutable default <class 'hydra.conf.JobConf.JobConfig.OverrideDirname'> for field override_dirname is not allowed: use default_factory\r\n```\r\nIf I then manually upgrade `hydra-core` to the current latest (`1.3.2`), I get similar errors from `nemo-toolkit`:\r\n```\r\nTraceback (most recent call last):\r\n  File \"/home/alex/T7/src/speaker_verification/nemo_speaker/test.py\", line 5, in <module>\r\n    import nemo.collections.asr as nemo_asr\r\n  File \"/home/alex/T7/src/speaker_verification/nemo_speaker/venv11/lib/python3.11/site-packages/nemo/collections/asr/__init__.py\", line 15, in <module>\r\n    from nemo.collections.asr import data, losses, models, modules\r\n  File \"/home/alex/T7/src/speaker_verification/nemo_speaker/venv11/lib/python3.11/site-packages/nemo/collections/asr/losses/__init__.py\", line 16, in <module>\r\n    from nemo.collections.asr.losses.audio_losses import SDRLoss\r\n  File \"/home/alex/T7/src/speaker_verification/nemo_speaker/venv11/lib/python3.11/site-packages/nemo/collections/asr/losses/audio_losses.py\", line 21, in <module>\r\n    from nemo.collections.asr.parts.preprocessing.features import make_seq_mask_like\r\n  File \"/home/alex/T7/src/speaker_verification/nemo_speaker/venv11/lib/python3.11/site-packages/nemo/collections/asr/parts/preprocessing/__init__.py\", line 16, in <module>\r\n    from nemo.collections.asr.parts.preprocessing.features import FeaturizerFactory, FilterbankFeatures, WaveformFeaturizer\r\n  File \"/home/alex/T7/src/speaker_verification/nemo_speaker/venv11/lib/python3.11/site-packages/nemo/collections/asr/parts/preprocessing/features.py\", line 44, in <module>\r\n    from nemo.collections.asr.parts.preprocessing.perturb import AudioAugmentor\r\n  File \"/home/alex/T7/src/speaker_verification/nemo_speaker/venv11/lib/python3.11/site-packages/nemo/collections/asr/parts/preprocessing/perturb.py\", line 50, in <module>\r\n    from nemo.collections.common.parts.preprocessing import collections, parsers\r\n  File \"/home/alex/T7/src/speaker_verification/nemo_speaker/venv11/lib/python3.11/site-packages/nemo/collections/common/__init__.py\", line 16, in <module>\r\n    from nemo.collections.common import data, losses, parts, tokenizers\r\n  File \"/home/alex/T7/src/speaker_verification/nemo_speaker/venv11/lib/python3.11/site-packages/nemo/collections/common/parts/__init__.py\", line 15, in <module>\r\n    from nemo.collections.common.parts.adapter_modules import LinearAdapter, LinearAdapterConfig\r\n  File \"/home/alex/T7/src/speaker_verification/nemo_speaker/venv11/lib/python3.11/site-packages/nemo/collections/common/parts/adapter_modules.py\", line 147, in <module>\r\n    @dataclass\r\n     ^^^^^^^^^\r\n  File \"/usr/lib/python3.11/dataclasses.py\", line 1230, in dataclass\r\n    return wrap(cls)\r\n           ^^^^^^^^^\r\n  File \"/usr/lib/python3.11/dataclasses.py\", line 1220, in wrap\r\n    return _process_class(cls, init, repr, eq, order, unsafe_hash,\r\n           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\r\n  File \"/usr/lib/python3.11/dataclasses.py\", line 958, in _process_class\r\n    cls_fields.append(_get_field(cls, name, type, kw_only))\r\n                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\r\n  File \"/usr/lib/python3.11/dataclasses.py\", line 815, in _get_field\r\n    raise ValueError(f'mutable default {type(f.default)} for field '\r\nValueError: mutable default <class 'nemo.core.classes.mixins.adapter_mixin_strategies.ResidualAddAdapterStrategyConfig'> for field adapter_strategy is not allowed: use default_factory\r\n```\r\nIt's easy to fix with a patch like this:\r\n```\r\n--- nemo/collections/common/parts/adapter_modules.py.orig       2023-08-04 13:55:53.464534800 +0200\r\n+++ nemo/collections/common/parts/adapter_modules.py    2023-08-04 14:05:45.579537700 +0200\r\n@@ -12,7 +12,7 @@\r\n # See the License for the specific language governing permissions and\r\n # limitations under the License.\r\n \r\n-from dataclasses import dataclass, is_dataclass\r\n+from dataclasses import dataclass, is_dataclass, field\r\n from typing import Any, Optional\r\n \r\n from hydra.utils import instantiate\r\n@@ -151,5 +151,5 @@ class LinearAdapterConfig:\r\n     activation: str = 'swish'\r\n     norm_position: str = 'pre'\r\n     dropout: float = 0.0\r\n-    adapter_strategy: Optional[Any] = adapter_mixin_strategies.ResidualAddAdapterStrategyConfig()\r\n+    adapter_strategy: Optional[Any] = field(default_factory=adapter_mixin_strategies.ResidualAddAdapterStrategyConfig)\r\n     _target_: str = \"{0}.{1}\".format(LinearAdapter.__module__, LinearAdapter.__name__)\r\n```\r\nHowever then another error of the kind comes up:\r\n```\r\nTraceback (most recent call last):\r\n  File \"/home/alex/T7/src/speaker_verification/nemo_speaker/test.py\", line 5, in <module>\r\n    import nemo.collections.asr as nemo_asr\r\n  File \"/home/alex/T7/src/speaker_verification/nemo_speaker/venv11/lib/python3.11/site-packages/nemo/collections/asr/__init__.py\", line 15, in <module>\r\n    from nemo.collections.asr import data, losses, models, modules\r\n  File \"/home/alex/T7/src/speaker_verification/nemo_speaker/venv11/lib/python3.11/site-packages/nemo/collections/asr/models/__init__.py\", line 18, in <module>\r\n    from nemo.collections.asr.models.clustering_diarizer import ClusteringDiarizer\r\n  File \"/home/alex/T7/src/speaker_verification/nemo_speaker/venv11/lib/python3.11/site-packages/nemo/collections/asr/models/clustering_diarizer.py\", line 29, in <module>\r\n    from nemo.collections.asr.metrics.der import score_labels\r\n  File \"/home/alex/T7/src/speaker_verification/nemo_speaker/venv11/lib/python3.11/site-packages/nemo/collections/asr/metrics/der.py\", line 24, in <module>\r\n    from nemo.collections.asr.metrics.wer import word_error_rate\r\n  File \"/home/alex/T7/src/speaker_verification/nemo_speaker/venv11/lib/python3.11/site-packages/nemo/collections/asr/metrics/wer.py\", line 27, in <module>\r\n    from nemo.collections.asr.parts.submodules import ctc_beam_decoding, ctc_greedy_decoding\r\n  File \"/home/alex/T7/src/speaker_verification/nemo_speaker/venv11/lib/python3.11/site-packages/nemo/collections/asr/parts/submodules/ctc_beam_decoding.py\", line 593, in <module>\r\n    @dataclass\r\n     ^^^^^^^^^\r\n  File \"/usr/lib/python3.11/dataclasses.py\", line 1230, in dataclass\r\n    return wrap(cls)\r\n           ^^^^^^^^^\r\n  File \"/usr/lib/python3.11/dataclasses.py\", line 1220, in wrap\r\n    return _process_class(cls, init, repr, eq, order, unsafe_hash,\r\n           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\r\n  File \"/usr/lib/python3.11/dataclasses.py\", line 958, in _process_class\r\n    cls_fields.append(_get_field(cls, name, type, kw_only))\r\n                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\r\n  File \"/usr/lib/python3.11/dataclasses.py\", line 815, in _get_field\r\n    raise ValueError(f'mutable default {type(f.default)} for field '\r\nValueError: mutable default <class 'nemo.collections.asr.parts.submodules.ctc_beam_decoding.FlashlightConfig'> for field flashlight_cfg is not allowed: use default_factory\r\n```\r\nThis can also be fixed accordingly, but all in all, it appears these issues are pretty common in the code base.\r\n\r\nLooks like NeMo isn't Python 3.11 ready, at least the `asr` collection.\r\n\r\n**Steps/Code to reproduce bug**\r\n\r\nWith Python 3.11\r\n1. Install `nemo-toolkit[asr]` either 1.19.1, or current HEAD from git (b498d438fc4c35ebf364a9a1c5cd3e29a2c0fe50)\r\n2. Run `import nemo.collections.asr`\r\n\r\n\r\n**Expected behavior**\r\n\r\nImport `nemo.collections.asr` without errors\r\n\r\n**Environment overview (please complete the following information)**\r\n\r\n - Environment location: Bare-metal\r\n - Method of NeMo install: `pip install nemo-toolkit==1.19.1` or from source b498d438fc4c35ebf364a9a1c5cd3e29a2c0fe50\r\n\r\n**Environment details**\r\n- Ubuntu 22.04\r\n- PyTorch version 2.0.1\r\n- Python version 3.11.4\r\n\r\n\n",
-    "hints_text": "Seems to be a similar to #7002\nInteresting. The fix is easy but needs to be applied to basically every single place that has this constructor for our adapter configs. Let me see if I can update it. But no guarantees on how soon fixes will come in main. \nLooking forward to it @titu1994 ! Thanks \ud83d\ude03 \n@titu1994 I was looking to use NeMo speaker diarization with Python 3.11 and hit this dataclass issue. I patched everything involved in the specific code paths I needed: https://github.com/lmnt-com/NeMo/commit/d89acf9f0152e43dee29d7d1c4667ee34c26ffd7\r\n\r\nI was using the neural diarizer as described in https://github.com/NVIDIA/NeMo/tree/main/examples/speaker_tasks/diarization\r\n\r\nI'd be happy to upstream this if it's helpful.\r\n\r\nI haven't checked whether this is backwards compatible for earlier python/dataclass versions, do you know?\r\n\r\nFor reference, what led me to this issue, though it's duplicative to the above discussion:\r\n\r\n[Similar error](https://github.com/huggingface/datasets/issues/5230)\r\n[StackOverflow solution](https://stackoverflow.com/questions/53632152/why-cant-dataclasses-have-mutable-defaults-in-their-class-attributes-declaratio)\n@shaper Thanks for sharing. For brevity, you don't really need a `lambda` when you don't pass any init parameters, like this:\r\n```\r\nfield(default_factory=lambda: ConfidenceConfig())\r\n```\r\nYou can just do\r\n```\r\nfield(default_factory=ConfidenceConfig)\r\n```\r\nIt's only needed when you do pass parameter(s), like\r\n```\r\nfield(default_factory=lambda: beam_decode.BeamRNNTInferConfig(beam_size=4))\r\n```\nThis issue is stale because it has been open for 30 days with no activity. Remove stale label or comment or this will be closed in 7 days.\nI have the same issue. @tango4j suggested using one of the models from https://huggingface.co/spaces/hf-audio/open_asr_leaderboard, but I cannot import nemo.collections.asr:\r\n\r\n```\r\n    Traceback (most recent call last):\r\n      File \"/opt/pycharm-2022.3.3/plugins/python/helpers/pycharm/docrunner.py\", line 138, in __run\r\n        exec(compile(example.source, filename, \"single\",\r\n      File \"<doctest NeMoASR[2]>\", line 1, in <module>\r\n        NeMoASR().apply_asr(file)\r\n        ^^^^^^^^^\r\n      File \"/home/cbj/python/cbj/cbj/transcribe/pretrained.py\", line 504, in __init__\r\n        import nemo.collections.asr as nemo_asr\r\n      File \"/opt/py/2023/lib/python3.11/site-packages/nemo/collections/asr/__init__.py\", line 15, in <module>\r\n        from nemo.collections.asr import data, losses, models, modules\r\n      File \"/opt/py/2023/lib/python3.11/site-packages/nemo/collections/asr/losses/__init__.py\", line 16, in <module>\r\n        from nemo.collections.asr.losses.audio_losses import SDRLoss\r\n      File \"/opt/py/2023/lib/python3.11/site-packages/nemo/collections/asr/losses/audio_losses.py\", line 21, in <module>\r\n        from nemo.collections.asr.parts.preprocessing.features import make_seq_mask_like\r\n      File \"/opt/py/2023/lib/python3.11/site-packages/nemo/collections/asr/parts/preprocessing/__init__.py\", line 16, in <module>\r\n        from nemo.collections.asr.parts.preprocessing.features import FeaturizerFactory, FilterbankFeatures, WaveformFeaturizer\r\n      File \"/opt/py/2023/lib/python3.11/site-packages/nemo/collections/asr/parts/preprocessing/features.py\", line 44, in <module>\r\n        from nemo.collections.asr.parts.preprocessing.perturb import AudioAugmentor\r\n      File \"/opt/py/2023/lib/python3.11/site-packages/nemo/collections/asr/parts/preprocessing/perturb.py\", line 50, in <module>\r\n        from nemo.collections.common.parts.preprocessing import collections, parsers\r\n      File \"/opt/py/2023/lib/python3.11/site-packages/nemo/collections/common/__init__.py\", line 16, in <module>\r\n        from nemo.collections.common import data, losses, parts, tokenizers\r\n      File \"/opt/py/2023/lib/python3.11/site-packages/nemo/collections/common/parts/__init__.py\", line 15, in <module>\r\n        from nemo.collections.common.parts.adapter_modules import LinearAdapter, LinearAdapterConfig\r\n      File \"/opt/py/2023/lib/python3.11/site-packages/nemo/collections/common/parts/adapter_modules.py\", line 147, in <module>\r\n        @dataclass\r\n         ^^^^^^^^^\r\n      File \"/opt/py/2023/lib/python3.11/dataclasses.py\", line 1230, in dataclass\r\n        return wrap(cls)\r\n               ^^^^^^^^^\r\n      File \"/opt/py/2023/lib/python3.11/dataclasses.py\", line 1220, in wrap\r\n        return _process_class(cls, init, repr, eq, order, unsafe_hash,\r\n               ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\r\n      File \"/opt/py/2023/lib/python3.11/dataclasses.py\", line 958, in _process_class\r\n        cls_fields.append(_get_field(cls, name, type, kw_only))\r\n                          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\r\n      File \"/opt/py/2023/lib/python3.11/dataclasses.py\", line 815, in _get_field\r\n        raise ValueError(f'mutable default {type(f.default)} for field '\r\n    ValueError: mutable default <class 'nemo.core.classes.mixins.adapter_mixin_strategies.ResidualAddAdapterStrategyConfig'> for field adapter_strategy is not allowed: use default_factory\r\n```\r\n\r\nFor documentation (I had to search in the provided links):\r\nMutable defaults were never allowed in dataclasses (by convention), but in python 3.11 they improved the check: Instead of checking some types (dict, list, set) they now use hashable as indicator for mutable.\r\n\r\nAn alternative to default_factory would be to use frozen dataclasses, but I don't know whether in this code base the configs are used as mutable objects or not.\nYou need to update to NeMo 1.20, omegaconf did a fix that should resolve this \nI have NeMo 1.20.0. \r\nWith `pip install nemo_toolkit` and `pip install pytorch_lightning` I installed yesterday nemo.\r\nSo it should be the newest PyPI version.\r\n\r\n```\r\n$ pip show nemo_toolkit\r\nName: nemo-toolkit\r\nVersion: 1.20.0\r\nSummary: NeMo - a toolkit for Conversational AI\r\nHome-page: https://github.com/nvidia/nemo\r\nAuthor: NVIDIA\r\nAuthor-email: nemo-toolkit@nvidia.com\r\nLicense: Apache2\r\nLocation: /opt/py/2023/lib/python3.11/site-packages\r\nRequires: huggingface-hub, numba, numpy, onnx, python-dateutil, ruamel.yaml, scikit-learn, setuptools, tensorboard, text-unidecode, torch, tqdm, wget, wrapt\r\nRequired-by: \r\n\r\n$ pip show omegaconf\r\nName: omegaconf\r\nVersion: 2.3.0\r\nSummary: A flexible configuration library\r\nHome-page: https://github.com/omry/omegaconf\r\nAuthor: Omry Yadan\r\nAuthor-email: omry@yadan.net\r\nLicense: \r\nLocation: /home/cbj/.local/lib/python3.11/site-packages\r\nRequires: antlr4-python3-runtime, PyYAML\r\nRequired-by: hydra-core\r\n\r\n$ python -c \"import nemo.collections.asr as nemo_asr\"\r\nTraceback (most recent call last):\r\n  File \"<string>\", line 1, in <module>\r\n  File \"/opt/py/2023/lib/python3.11/site-packages/nemo/collections/asr/__init__.py\", line 15, in <module>\r\n    from nemo.collections.asr import data, losses, models, modules\r\n  File \"/opt/py/2023/lib/python3.11/site-packages/nemo/collections/asr/losses/__init__.py\", line 16, in <module>\r\n    from nemo.collections.asr.losses.audio_losses import SDRLoss\r\n  File \"/opt/py/2023/lib/python3.11/site-packages/nemo/collections/asr/losses/audio_losses.py\", line 21, in <module>\r\n    from nemo.collections.asr.parts.preprocessing.features import make_seq_mask_like\r\n  File \"/opt/py/2023/lib/python3.11/site-packages/nemo/collections/asr/parts/preprocessing/__init__.py\", line 16, in <module>\r\n    from nemo.collections.asr.parts.preprocessing.features import FeaturizerFactory, FilterbankFeatures, WaveformFeaturizer\r\n  File \"/opt/py/2023/lib/python3.11/site-packages/nemo/collections/asr/parts/preprocessing/features.py\", line 44, in <module>\r\n    from nemo.collections.asr.parts.preprocessing.perturb import AudioAugmentor\r\n  File \"/opt/py/2023/lib/python3.11/site-packages/nemo/collections/asr/parts/preprocessing/perturb.py\", line 50, in <module>\r\n    from nemo.collections.common.parts.preprocessing import collections, parsers\r\n  File \"/opt/py/2023/lib/python3.11/site-packages/nemo/collections/common/__init__.py\", line 16, in <module>\r\n    from nemo.collections.common import data, losses, parts, tokenizers\r\n  File \"/opt/py/2023/lib/python3.11/site-packages/nemo/collections/common/parts/__init__.py\", line 15, in <module>\r\n    from nemo.collections.common.parts.adapter_modules import LinearAdapter, LinearAdapterConfig\r\n  File \"/opt/py/2023/lib/python3.11/site-packages/nemo/collections/common/parts/adapter_modules.py\", line 147, in <module>\r\n    @dataclass\r\n     ^^^^^^^^^\r\n  File \"/opt/py/2023/lib/python3.11/dataclasses.py\", line 1230, in dataclass\r\n    return wrap(cls)\r\n           ^^^^^^^^^\r\n  File \"/opt/py/2023/lib/python3.11/dataclasses.py\", line 1220, in wrap\r\n    return _process_class(cls, init, repr, eq, order, unsafe_hash,\r\n           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\r\n  File \"/opt/py/2023/lib/python3.11/dataclasses.py\", line 958, in _process_class\r\n    cls_fields.append(_get_field(cls, name, type, kw_only))\r\n                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\r\n  File \"/opt/py/2023/lib/python3.11/dataclasses.py\", line 815, in _get_field\r\n    raise ValueError(f'mutable default {type(f.default)} for field '\r\nValueError: mutable default <class 'nemo.core.classes.mixins.adapter_mixin_strategies.ResidualAddAdapterStrategyConfig'> for field adapter_strategy is not allowed: use default_factory\r\n\r\n```\nHmm ok I'll take a look ",
-    "created_at": "2023-09-30T01:26:50Z",
-    "version": "1.0"
-  },
-  {
-    "repo": "NVIDIA/NeMo",
-    "pull_number": 7616,
-    "instance_id": "NVIDIA__NeMo-7616",
-    "issue_numbers": [
-      "7166"
-    ],
-    "base_commit": "15db83ec4a65e649d83b61d7a4a58d911586e853",
-    "patch": "diff --git a/examples/asr/experimental/k2/align_speech_parallel.py b/examples/asr/experimental/k2/align_speech_parallel.py\n--- a/examples/asr/experimental/k2/align_speech_parallel.py\n+++ b/examples/asr/experimental/k2/align_speech_parallel.py\n@@ -74,7 +74,7 @@\n \n \n import os\n-from dataclasses import dataclass, is_dataclass\n+from dataclasses import dataclass, field, is_dataclass\n from typing import Optional\n \n import pytorch_lightning as ptl\n@@ -94,12 +94,14 @@\n @dataclass\n class ParallelAlignmentConfig:\n     model: Optional[str] = None  # name\n-    predict_ds: ASRDatasetConfig = ASRDatasetConfig(return_sample_id=True, num_workers=4)\n-    aligner_args: K2AlignerWrapperModelConfig = K2AlignerWrapperModelConfig()\n+    predict_ds: ASRDatasetConfig = field(\n+        default_factory=lambda: ASRDatasetConfig(return_sample_id=True, num_workers=4)\n+    )\n+    aligner_args: K2AlignerWrapperModelConfig = field(default_factory=lambda: K2AlignerWrapperModelConfig())\n     output_path: str = MISSING\n     model_stride: int = 8\n \n-    trainer: TrainerConfig = TrainerConfig(gpus=-1, accelerator=\"ddp\")\n+    trainer: TrainerConfig = field(default_factory=lambda: TrainerConfig(gpus=-1, accelerator=\"ddp\"))\n \n     # there arguments will be ignored\n     return_predictions: bool = False\ndiff --git a/nemo/collections/asr/metrics/rnnt_wer.py b/nemo/collections/asr/metrics/rnnt_wer.py\n--- a/nemo/collections/asr/metrics/rnnt_wer.py\n+++ b/nemo/collections/asr/metrics/rnnt_wer.py\n@@ -15,7 +15,7 @@\n import copy\n import re\n from abc import abstractmethod\n-from dataclasses import dataclass, is_dataclass\n+from dataclasses import dataclass, field, is_dataclass\n from typing import Callable, Dict, List, Optional, Tuple, Union\n \n import editdistance\n@@ -1299,7 +1299,7 @@ class RNNTDecodingConfig:\n     preserve_alignments: Optional[bool] = None\n \n     #  confidence config\n-    confidence_cfg: ConfidenceConfig = ConfidenceConfig()\n+    confidence_cfg: ConfidenceConfig = field(default_factory=lambda: ConfidenceConfig())\n \n     # RNNT Joint fused batch size\n     fused_batch_size: Optional[int] = None\n@@ -1317,10 +1317,10 @@ class RNNTDecodingConfig:\n     rnnt_timestamp_type: str = \"all\"  # can be char, word or all for both\n \n     # greedy decoding config\n-    greedy: greedy_decode.GreedyRNNTInferConfig = greedy_decode.GreedyRNNTInferConfig()\n+    greedy: greedy_decode.GreedyRNNTInferConfig = field(default_factory=lambda: greedy_decode.GreedyRNNTInferConfig())\n \n     # beam decoding config\n-    beam: beam_decode.BeamRNNTInferConfig = beam_decode.BeamRNNTInferConfig(beam_size=4)\n+    beam: beam_decode.BeamRNNTInferConfig = field(default_factory=lambda: beam_decode.BeamRNNTInferConfig(beam_size=4))\n \n     # can be used to change temperature for decoding\n     temperature: float = 1.0\ndiff --git a/nemo/collections/asr/metrics/wer.py b/nemo/collections/asr/metrics/wer.py\n--- a/nemo/collections/asr/metrics/wer.py\n+++ b/nemo/collections/asr/metrics/wer.py\n@@ -14,7 +14,7 @@\n \n import re\n from abc import abstractmethod\n-from dataclasses import dataclass, is_dataclass\n+from dataclasses import dataclass, field, is_dataclass\n from typing import Callable, Dict, List, Optional, Tuple, Union\n \n import editdistance\n@@ -1297,13 +1297,17 @@ class CTCDecodingConfig:\n     batch_dim_index: int = 0\n \n     # greedy decoding config\n-    greedy: ctc_greedy_decoding.GreedyCTCInferConfig = ctc_greedy_decoding.GreedyCTCInferConfig()\n+    greedy: ctc_greedy_decoding.GreedyCTCInferConfig = field(\n+        default_factory=lambda: ctc_greedy_decoding.GreedyCTCInferConfig()\n+    )\n \n     # beam decoding config\n-    beam: ctc_beam_decoding.BeamCTCInferConfig = ctc_beam_decoding.BeamCTCInferConfig(beam_size=4)\n+    beam: ctc_beam_decoding.BeamCTCInferConfig = field(\n+        default_factory=lambda: ctc_beam_decoding.BeamCTCInferConfig(beam_size=4)\n+    )\n \n     # confidence config\n-    confidence_cfg: ConfidenceConfig = ConfidenceConfig()\n+    confidence_cfg: ConfidenceConfig = field(default_factory=lambda: ConfidenceConfig())\n \n     # can be used to change temperature for decoding\n     temperature: float = 1.0\ndiff --git a/nemo/collections/asr/models/configs/aligner_config.py b/nemo/collections/asr/models/configs/aligner_config.py\n--- a/nemo/collections/asr/models/configs/aligner_config.py\n+++ b/nemo/collections/asr/models/configs/aligner_config.py\n@@ -12,7 +12,7 @@\n # See the License for the specific language governing permissions and\n # limitations under the License.\n \n-from dataclasses import dataclass\n+from dataclasses import dataclass, field\n \n from nemo.collections.asr.parts.k2.classes import GraphModuleConfig\n \n@@ -35,10 +35,10 @@ class AlignerWrapperModelConfig:\n     word_output: bool = True\n     cpu_decoding: bool = False\n     decode_batch_size: int = 0\n-    ctc_cfg: AlignerCTCConfig = AlignerCTCConfig()\n-    rnnt_cfg: AlignerRNNTConfig = AlignerRNNTConfig()\n+    ctc_cfg: AlignerCTCConfig = field(default_factory=lambda: AlignerCTCConfig())\n+    rnnt_cfg: AlignerRNNTConfig = field(default_factory=lambda: AlignerRNNTConfig())\n \n \n @dataclass\n class K2AlignerWrapperModelConfig(AlignerWrapperModelConfig):\n-    decoder_module_cfg: GraphModuleConfig = GraphModuleConfig()\n+    decoder_module_cfg: GraphModuleConfig = field(default_factory=lambda: GraphModuleConfig())\ndiff --git a/nemo/collections/asr/models/configs/asr_models_config.py b/nemo/collections/asr/models/configs/asr_models_config.py\n--- a/nemo/collections/asr/models/configs/asr_models_config.py\n+++ b/nemo/collections/asr/models/configs/asr_models_config.py\n@@ -12,7 +12,7 @@\n # See the License for the specific language governing permissions and\n # limitations under the License.\n \n-from dataclasses import dataclass\n+from dataclasses import dataclass, field\n from typing import Any, Dict, List, Optional\n \n from omegaconf import MISSING\n@@ -74,24 +74,32 @@ class EncDecCTCConfig(model_cfg.ModelConfig):\n     labels: List[str] = MISSING\n \n     # Dataset configs\n-    train_ds: ASRDatasetConfig = ASRDatasetConfig(manifest_filepath=None, shuffle=True)\n-    validation_ds: ASRDatasetConfig = ASRDatasetConfig(manifest_filepath=None, shuffle=False)\n-    test_ds: ASRDatasetConfig = ASRDatasetConfig(manifest_filepath=None, shuffle=False)\n+    train_ds: ASRDatasetConfig = field(default_factory=lambda: ASRDatasetConfig(manifest_filepath=None, shuffle=True))\n+    validation_ds: ASRDatasetConfig = field(\n+        default_factory=lambda: ASRDatasetConfig(manifest_filepath=None, shuffle=False)\n+    )\n+    test_ds: ASRDatasetConfig = field(default_factory=lambda: ASRDatasetConfig(manifest_filepath=None, shuffle=False))\n \n     # Optimizer / Scheduler config\n-    optim: Optional[model_cfg.OptimConfig] = model_cfg.OptimConfig(sched=model_cfg.SchedConfig())\n+    optim: Optional[model_cfg.OptimConfig] = field(\n+        default_factory=lambda: model_cfg.OptimConfig(sched=model_cfg.SchedConfig())\n+    )\n \n     # Model component configs\n-    preprocessor: AudioToMelSpectrogramPreprocessorConfig = AudioToMelSpectrogramPreprocessorConfig()\n-    spec_augment: Optional[SpectrogramAugmentationConfig] = SpectrogramAugmentationConfig()\n-    encoder: ConvASREncoderConfig = ConvASREncoderConfig()\n-    decoder: ConvASRDecoderConfig = ConvASRDecoderConfig()\n-    decoding: CTCDecodingConfig = CTCDecodingConfig()\n+    preprocessor: AudioToMelSpectrogramPreprocessorConfig = field(\n+        default_factory=lambda: AudioToMelSpectrogramPreprocessorConfig()\n+    )\n+    spec_augment: Optional[SpectrogramAugmentationConfig] = field(\n+        default_factory=lambda: SpectrogramAugmentationConfig()\n+    )\n+    encoder: ConvASREncoderConfig = field(default_factory=lambda: ConvASREncoderConfig())\n+    decoder: ConvASRDecoderConfig = field(default_factory=lambda: ConvASRDecoderConfig())\n+    decoding: CTCDecodingConfig = field(default_factory=lambda: CTCDecodingConfig())\n \n \n @dataclass\n class EncDecCTCModelConfig(model_cfg.NemoConfig):\n-    model: EncDecCTCConfig = EncDecCTCConfig()\n+    model: EncDecCTCConfig = field(default_factory=lambda: EncDecCTCConfig())\n \n \n @dataclass\ndiff --git a/nemo/collections/asr/models/configs/classification_models_config.py b/nemo/collections/asr/models/configs/classification_models_config.py\n--- a/nemo/collections/asr/models/configs/classification_models_config.py\n+++ b/nemo/collections/asr/models/configs/classification_models_config.py\n@@ -12,7 +12,7 @@\n # See the License for the specific language governing permissions and\n # limitations under the License.\n \n-from dataclasses import dataclass\n+from dataclasses import dataclass, field\n from typing import Any, Dict, List, Optional\n \n from omegaconf import MISSING\n@@ -72,30 +72,40 @@ class EncDecClassificationConfig(model_cfg.ModelConfig):\n     timesteps: int = MISSING\n \n     # Dataset configs\n-    train_ds: EncDecClassificationDatasetConfig = EncDecClassificationDatasetConfig(\n-        manifest_filepath=None, shuffle=True, trim_silence=False\n+    train_ds: EncDecClassificationDatasetConfig = field(\n+        default_factory=lambda: EncDecClassificationDatasetConfig(\n+            manifest_filepath=None, shuffle=True, trim_silence=False\n+        )\n     )\n-    validation_ds: EncDecClassificationDatasetConfig = EncDecClassificationDatasetConfig(\n-        manifest_filepath=None, shuffle=False\n+    validation_ds: EncDecClassificationDatasetConfig = field(\n+        default_factory=lambda: EncDecClassificationDatasetConfig(manifest_filepath=None, shuffle=False)\n     )\n-    test_ds: EncDecClassificationDatasetConfig = EncDecClassificationDatasetConfig(\n-        manifest_filepath=None, shuffle=False\n+    test_ds: EncDecClassificationDatasetConfig = field(\n+        default_factory=lambda: EncDecClassificationDatasetConfig(manifest_filepath=None, shuffle=False)\n     )\n \n     # Optimizer / Scheduler config\n-    optim: Optional[model_cfg.OptimConfig] = model_cfg.OptimConfig(sched=model_cfg.SchedConfig())\n+    optim: Optional[model_cfg.OptimConfig] = field(\n+        default_factory=lambda: model_cfg.OptimConfig(sched=model_cfg.SchedConfig())\n+    )\n \n     # Model component configs\n-    preprocessor: AudioToMFCCPreprocessorConfig = AudioToMFCCPreprocessorConfig()\n-    spec_augment: Optional[SpectrogramAugmentationConfig] = SpectrogramAugmentationConfig()\n-    crop_or_pad_augment: Optional[CropOrPadSpectrogramAugmentationConfig] = CropOrPadSpectrogramAugmentationConfig(\n-        audio_length=timesteps\n+    preprocessor: AudioToMFCCPreprocessorConfig = field(default_factory=lambda: AudioToMFCCPreprocessorConfig())\n+    spec_augment: Optional[SpectrogramAugmentationConfig] = field(\n+        default_factory=lambda: SpectrogramAugmentationConfig()\n+    )\n+    crop_or_pad_augment: Optional[CropOrPadSpectrogramAugmentationConfig] = field(\n+        default_factory=lambda: CropOrPadSpectrogramAugmentationConfig(audio_length=-1)\n     )\n \n-    encoder: ConvASREncoderConfig = ConvASREncoderConfig()\n-    decoder: ConvASRDecoderClassificationConfig = ConvASRDecoderClassificationConfig()\n+    encoder: ConvASREncoderConfig = field(default_factory=lambda: ConvASREncoderConfig())\n+    decoder: ConvASRDecoderClassificationConfig = field(default_factory=lambda: ConvASRDecoderClassificationConfig())\n+\n+    def __post_init__(self):\n+        if self.crop_or_pad_augment is not None:\n+            self.crop_or_pad_augment.audio_length = self.timesteps\n \n \n @dataclass\n class EncDecClassificationModelConfig(model_cfg.NemoConfig):\n-    model: EncDecClassificationConfig = EncDecClassificationConfig()\n+    model: EncDecClassificationConfig = field(default_factory=lambda: EncDecClassificationConfig())\ndiff --git a/nemo/collections/asr/models/configs/diarizer_config.py b/nemo/collections/asr/models/configs/diarizer_config.py\n--- a/nemo/collections/asr/models/configs/diarizer_config.py\n+++ b/nemo/collections/asr/models/configs/diarizer_config.py\n@@ -12,7 +12,7 @@\n # See the License for the specific language governing permissions and\n # limitations under the License.\n \n-from dataclasses import asdict, dataclass\n+from dataclasses import asdict, dataclass, field\n from typing import Any, Dict, Optional, Tuple, Union\n \n \n@@ -78,9 +78,9 @@ class ASRDiarizerParams(DiarizerComponentConfig):\n @dataclass\n class ASRDiarizerConfig(DiarizerComponentConfig):\n     model_path: Optional[str] = \"stt_en_conformer_ctc_large\"\n-    parameters: ASRDiarizerParams = ASRDiarizerParams()\n-    ctc_decoder_parameters: ASRDiarizerCTCDecoderParams = ASRDiarizerCTCDecoderParams()\n-    realigning_lm_parameters: ASRRealigningLMParams = ASRRealigningLMParams()\n+    parameters: ASRDiarizerParams = field(default_factory=lambda: ASRDiarizerParams())\n+    ctc_decoder_parameters: ASRDiarizerCTCDecoderParams = field(default_factory=lambda: ASRDiarizerCTCDecoderParams())\n+    realigning_lm_parameters: ASRRealigningLMParams = field(default_factory=lambda: ASRRealigningLMParams())\n \n \n @dataclass\n@@ -102,7 +102,7 @@ class VADParams(DiarizerComponentConfig):\n class VADConfig(DiarizerComponentConfig):\n     model_path: str = \"vad_multilingual_marblenet\"  # .nemo local model path or pretrained VAD model name\n     external_vad_manifest: Optional[str] = None\n-    parameters: VADParams = VADParams()\n+    parameters: VADParams = field(default_factory=lambda: VADParams())\n \n \n @dataclass\n@@ -121,7 +121,7 @@ class SpeakerEmbeddingsParams(DiarizerComponentConfig):\n class SpeakerEmbeddingsConfig(DiarizerComponentConfig):\n     # .nemo local model path or pretrained model name (titanet_large, ecapa_tdnn or speakerverification_speakernet)\n     model_path: Optional[str] = None\n-    parameters: SpeakerEmbeddingsParams = SpeakerEmbeddingsParams()\n+    parameters: SpeakerEmbeddingsParams = field(default_factory=lambda: SpeakerEmbeddingsParams())\n \n \n @dataclass\n@@ -142,7 +142,7 @@ class ClusteringParams(DiarizerComponentConfig):\n \n @dataclass\n class ClusteringConfig(DiarizerComponentConfig):\n-    parameters: ClusteringParams = ClusteringParams()\n+    parameters: ClusteringParams = field(default_factory=lambda: ClusteringParams())\n \n \n @dataclass\n@@ -166,7 +166,7 @@ class MSDDParams(DiarizerComponentConfig):\n @dataclass\n class MSDDConfig(DiarizerComponentConfig):\n     model_path: Optional[str] = \"diar_msdd_telephonic\"\n-    parameters: MSDDParams = MSDDParams()\n+    parameters: MSDDParams = field(default_factory=lambda: MSDDParams())\n \n \n @dataclass\n@@ -176,16 +176,16 @@ class DiarizerConfig(DiarizerComponentConfig):\n     oracle_vad: bool = False  # If True, uses RTTM files provided in the manifest file to get VAD timestamps\n     collar: float = 0.25  # Collar value for scoring\n     ignore_overlap: bool = True  # Consider or ignore overlap segments while scoring\n-    vad: VADConfig = VADConfig()\n-    speaker_embeddings: SpeakerEmbeddingsConfig = SpeakerEmbeddingsConfig()\n-    clustering: ClusteringConfig = ClusteringConfig()\n-    msdd_model: MSDDConfig = MSDDConfig()\n-    asr: ASRDiarizerConfig = ASRDiarizerConfig()\n+    vad: VADConfig = field(default_factory=lambda: VADConfig())\n+    speaker_embeddings: SpeakerEmbeddingsConfig = field(default_factory=lambda: SpeakerEmbeddingsConfig())\n+    clustering: ClusteringConfig = field(default_factory=lambda: ClusteringConfig())\n+    msdd_model: MSDDConfig = field(default_factory=lambda: MSDDConfig())\n+    asr: ASRDiarizerConfig = field(default_factory=lambda: ASRDiarizerConfig())\n \n \n @dataclass\n class NeuralDiarizerInferenceConfig(DiarizerComponentConfig):\n-    diarizer: DiarizerConfig = DiarizerConfig()\n+    diarizer: DiarizerConfig = field(default_factory=lambda: DiarizerConfig())\n     device: str = \"cpu\"\n     verbose: bool = False\n     batch_size: int = 64\ndiff --git a/nemo/collections/asr/models/configs/k2_sequence_models_config.py b/nemo/collections/asr/models/configs/k2_sequence_models_config.py\n--- a/nemo/collections/asr/models/configs/k2_sequence_models_config.py\n+++ b/nemo/collections/asr/models/configs/k2_sequence_models_config.py\n@@ -12,7 +12,7 @@\n # See the License for the specific language governing permissions and\n # limitations under the License.\n \n-from dataclasses import dataclass\n+from dataclasses import dataclass, field\n \n from nemo.collections.asr.models.configs.asr_models_config import EncDecCTCConfig\n from nemo.collections.asr.parts.k2.classes import GraphModuleConfig as BackendConfig\n@@ -26,14 +26,14 @@ class GraphModuleConfig:\n     split_batch_size: int = 0\n     dec_type: str = \"topo\"\n     transcribe_training: bool = True\n-    backend_cfg: BackendConfig = BackendConfig()\n+    backend_cfg: BackendConfig = field(default_factory=lambda: BackendConfig())\n \n \n @dataclass\n class EncDecK2SeqConfig(EncDecCTCConfig):\n-    graph_module_cfg: GraphModuleConfig = GraphModuleConfig()\n+    graph_module_cfg: GraphModuleConfig = field(default_factory=lambda: GraphModuleConfig())\n \n \n @dataclass\n class EncDecK2SeqModelConfig(NemoConfig):\n-    model: EncDecK2SeqConfig = EncDecK2SeqConfig()\n+    model: EncDecK2SeqConfig = field(default_factory=lambda: EncDecK2SeqConfig())\ndiff --git a/nemo/collections/asr/models/configs/matchboxnet_config.py b/nemo/collections/asr/models/configs/matchboxnet_config.py\n--- a/nemo/collections/asr/models/configs/matchboxnet_config.py\n+++ b/nemo/collections/asr/models/configs/matchboxnet_config.py\n@@ -107,30 +107,38 @@ class MatchboxNetModelConfig(clf_cfg.EncDecClassificationConfig):\n     labels: List[str] = MISSING\n \n     # Dataset configs\n-    train_ds: clf_cfg.EncDecClassificationDatasetConfig = clf_cfg.EncDecClassificationDatasetConfig(\n-        manifest_filepath=None, shuffle=True, trim_silence=False\n+    train_ds: clf_cfg.EncDecClassificationDatasetConfig = field(\n+        default_factory=lambda: clf_cfg.EncDecClassificationDatasetConfig(\n+            manifest_filepath=None, shuffle=True, trim_silence=False\n+        )\n     )\n-    validation_ds: clf_cfg.EncDecClassificationDatasetConfig = clf_cfg.EncDecClassificationDatasetConfig(\n-        manifest_filepath=None, shuffle=False\n+    validation_ds: clf_cfg.EncDecClassificationDatasetConfig = field(\n+        default_factory=lambda: clf_cfg.EncDecClassificationDatasetConfig(manifest_filepath=None, shuffle=False)\n     )\n-    test_ds: clf_cfg.EncDecClassificationDatasetConfig = clf_cfg.EncDecClassificationDatasetConfig(\n-        manifest_filepath=None, shuffle=False\n+    test_ds: clf_cfg.EncDecClassificationDatasetConfig = field(\n+        default_factory=lambda: clf_cfg.EncDecClassificationDatasetConfig(manifest_filepath=None, shuffle=False)\n     )\n \n     # Optimizer / Scheduler config\n-    optim: Optional[model_cfg.OptimConfig] = model_cfg.OptimConfig(sched=model_cfg.SchedConfig())\n+    optim: Optional[model_cfg.OptimConfig] = field(\n+        default_factory=lambda: model_cfg.OptimConfig(sched=model_cfg.SchedConfig())\n+    )\n \n     # Model general component configs\n-    preprocessor: AudioToMFCCPreprocessorConfig = AudioToMFCCPreprocessorConfig(window_size=0.025)\n-    spec_augment: Optional[SpectrogramAugmentationConfig] = SpectrogramAugmentationConfig(\n-        freq_masks=2, time_masks=2, freq_width=15, time_width=25, rect_masks=5, rect_time=25, rect_freq=15\n+    preprocessor: AudioToMFCCPreprocessorConfig = field(\n+        default_factory=lambda: AudioToMFCCPreprocessorConfig(window_size=0.025)\n+    )\n+    spec_augment: Optional[SpectrogramAugmentationConfig] = field(\n+        default_factory=lambda: SpectrogramAugmentationConfig(\n+            freq_masks=2, time_masks=2, freq_width=15, time_width=25, rect_masks=5, rect_time=25, rect_freq=15\n+        )\n     )\n-    crop_or_pad_augment: Optional[CropOrPadSpectrogramAugmentationConfig] = CropOrPadSpectrogramAugmentationConfig(\n-        audio_length=128\n+    crop_or_pad_augment: Optional[CropOrPadSpectrogramAugmentationConfig] = field(\n+        default_factory=lambda: CropOrPadSpectrogramAugmentationConfig(audio_length=128)\n     )\n \n-    encoder: ConvASREncoderConfig = ConvASREncoderConfig(activation=\"relu\")\n-    decoder: ConvASRDecoderClassificationConfig = ConvASRDecoderClassificationConfig()\n+    encoder: ConvASREncoderConfig = field(default_factory=lambda: ConvASREncoderConfig(activation=\"relu\"))\n+    decoder: ConvASRDecoderClassificationConfig = field(default_factory=lambda: ConvASRDecoderClassificationConfig())\n \n \n @dataclass\ndiff --git a/nemo/collections/asr/models/configs/quartznet_config.py b/nemo/collections/asr/models/configs/quartznet_config.py\n--- a/nemo/collections/asr/models/configs/quartznet_config.py\n+++ b/nemo/collections/asr/models/configs/quartznet_config.py\n@@ -12,7 +12,7 @@\n # See the License for the specific language governing permissions and\n # limitations under the License.\n \n-from dataclasses import dataclass\n+from dataclasses import dataclass, field\n from typing import Any, Callable, List, Optional\n \n from omegaconf import MISSING\n@@ -174,20 +174,30 @@ class JasperModelConfig(ctc_cfg.EncDecCTCConfig):\n     labels: List[str] = MISSING\n \n     # Dataset configs\n-    train_ds: ctc_cfg.ASRDatasetConfig = ctc_cfg.ASRDatasetConfig(\n-        manifest_filepath=None, shuffle=True, trim_silence=True\n+    train_ds: ctc_cfg.ASRDatasetConfig = field(\n+        default_factory=lambda: ctc_cfg.ASRDatasetConfig(manifest_filepath=None, shuffle=True, trim_silence=True)\n+    )\n+    validation_ds: ctc_cfg.ASRDatasetConfig = field(\n+        default_factory=lambda: ctc_cfg.ASRDatasetConfig(manifest_filepath=None, shuffle=False)\n+    )\n+    test_ds: ctc_cfg.ASRDatasetConfig = field(\n+        default_factory=lambda: ctc_cfg.ASRDatasetConfig(manifest_filepath=None, shuffle=False)\n     )\n-    validation_ds: ctc_cfg.ASRDatasetConfig = ctc_cfg.ASRDatasetConfig(manifest_filepath=None, shuffle=False)\n-    test_ds: ctc_cfg.ASRDatasetConfig = ctc_cfg.ASRDatasetConfig(manifest_filepath=None, shuffle=False)\n \n     # Optimizer / Scheduler config\n-    optim: Optional[model_cfg.OptimConfig] = model_cfg.OptimConfig(sched=model_cfg.SchedConfig())\n+    optim: Optional[model_cfg.OptimConfig] = field(\n+        default_factory=lambda: model_cfg.OptimConfig(sched=model_cfg.SchedConfig())\n+    )\n \n     # Model general component configs\n-    preprocessor: AudioToMelSpectrogramPreprocessorConfig = AudioToMelSpectrogramPreprocessorConfig()\n-    spec_augment: Optional[SpectrogramAugmentationConfig] = SpectrogramAugmentationConfig()\n-    encoder: ConvASREncoderConfig = ConvASREncoderConfig(activation=\"relu\")\n-    decoder: ConvASRDecoderConfig = ConvASRDecoderConfig()\n+    preprocessor: AudioToMelSpectrogramPreprocessorConfig = field(\n+        default_factory=lambda: AudioToMelSpectrogramPreprocessorConfig()\n+    )\n+    spec_augment: Optional[SpectrogramAugmentationConfig] = field(\n+        default_factory=lambda: SpectrogramAugmentationConfig()\n+    )\n+    encoder: ConvASREncoderConfig = field(default_factory=lambda: ConvASREncoderConfig(activation=\"relu\"))\n+    decoder: ConvASRDecoderConfig = field(default_factory=lambda: ConvASRDecoderConfig())\n \n \n @dataclass\ndiff --git a/nemo/collections/asr/modules/audio_preprocessing.py b/nemo/collections/asr/modules/audio_preprocessing.py\n--- a/nemo/collections/asr/modules/audio_preprocessing.py\n+++ b/nemo/collections/asr/modules/audio_preprocessing.py\n@@ -634,6 +634,12 @@ def __init__(self, audio_length):\n         super(CropOrPadSpectrogramAugmentation, self).__init__()\n         self.audio_length = audio_length\n \n+        if self.audio_length < 0:\n+            raise ValueError(\n+                'audio_length must be non-negative. If using a dataclass with OmegaConf, '\n+                'please call OmegaConf.to_object(cfg) to call appropriate __post_init__ methods.'\n+            )\n+\n     @typecheck()\n     @torch.no_grad()\n     def forward(self, input_signal, length):\ndiff --git a/nemo/collections/asr/parts/k2/classes.py b/nemo/collections/asr/parts/k2/classes.py\n--- a/nemo/collections/asr/parts/k2/classes.py\n+++ b/nemo/collections/asr/parts/k2/classes.py\n@@ -13,7 +13,7 @@\n # limitations under the License.\n \n from abc import ABC\n-from dataclasses import dataclass\n+from dataclasses import dataclass, field\n from typing import Any, Optional, Tuple\n \n import torch\n@@ -43,7 +43,7 @@ class GraphModuleConfig:\n     topo_with_self_loops: bool = True\n     token_lm: Optional[Any] = None\n     intersect_pruned: bool = False\n-    intersect_conf: GraphIntersectDenseConfig = GraphIntersectDenseConfig()\n+    intersect_conf: GraphIntersectDenseConfig = field(default_factory=lambda: GraphIntersectDenseConfig())\n     boost_coeff: float = 0.0\n     predictor_window_size: int = 0\n     predictor_step_size: int = 1\ndiff --git a/nemo/collections/asr/parts/submodules/adapters/multi_head_attention_adapter_module.py b/nemo/collections/asr/parts/submodules/adapters/multi_head_attention_adapter_module.py\n--- a/nemo/collections/asr/parts/submodules/adapters/multi_head_attention_adapter_module.py\n+++ b/nemo/collections/asr/parts/submodules/adapters/multi_head_attention_adapter_module.py\n@@ -13,7 +13,7 @@\n # limitations under the License.\n \n import math\n-from dataclasses import dataclass\n+from dataclasses import dataclass, field\n from typing import Any, Optional\n \n import torch\n@@ -183,7 +183,7 @@ class MultiHeadAttentionAdapterConfig:\n     n_feat: int\n     dropout_rate: float = 0.0\n     proj_dim: Optional[int] = None\n-    adapter_strategy: Optional[Any] = MHAResidualAddAdapterStrategyConfig()\n+    adapter_strategy: Optional[Any] = field(default_factory=lambda: MHAResidualAddAdapterStrategyConfig())\n     _target_: str = \"{0}.{1}\".format(MultiHeadAttentionAdapter.__module__, MultiHeadAttentionAdapter.__name__)\n \n \n@@ -287,7 +287,7 @@ class RelPositionMultiHeadAttentionAdapterConfig:\n     n_feat: int\n     dropout_rate: float = 0.0\n     proj_dim: Optional[int] = None\n-    adapter_strategy: Optional[Any] = MHAResidualAddAdapterStrategyConfig()\n+    adapter_strategy: Optional[Any] = field(default_factory=lambda: MHAResidualAddAdapterStrategyConfig())\n     _target_: str = \"{0}.{1}\".format(\n         RelPositionMultiHeadAttentionAdapter.__module__, RelPositionMultiHeadAttentionAdapter.__name__\n     )\n@@ -336,7 +336,9 @@ class PositionalEncodingAdapterConfig:\n     d_model: int\n     max_len: int = 5000\n     xscale: float = 1.0\n-    adapter_strategy: Optional[Any] = adapter_mixin_strategies.ResidualAddAdapterStrategyConfig()\n+    adapter_strategy: Optional[Any] = field(\n+        default_factory=lambda: adapter_mixin_strategies.ResidualAddAdapterStrategyConfig()\n+    )\n     _target_: str = \"{0}.{1}\".format(PositionalEncodingAdapter.__module__, PositionalEncodingAdapter.__name__)\n \n \n@@ -378,5 +380,7 @@ class RelPositionalEncodingAdapterConfig:\n     d_model: int\n     max_len: int = 5000\n     xscale: float = 1.0\n-    adapter_strategy: Optional[Any] = adapter_mixin_strategies.ResidualAddAdapterStrategyConfig()\n+    adapter_strategy: Optional[Any] = field(\n+        default_factory=lambda: adapter_mixin_strategies.ResidualAddAdapterStrategyConfig()\n+    )\n     _target_: str = \"{0}.{1}\".format(RelPositionalEncodingAdapter.__module__, RelPositionalEncodingAdapter.__name__)\ndiff --git a/nemo/collections/asr/parts/submodules/ctc_beam_decoding.py b/nemo/collections/asr/parts/submodules/ctc_beam_decoding.py\n--- a/nemo/collections/asr/parts/submodules/ctc_beam_decoding.py\n+++ b/nemo/collections/asr/parts/submodules/ctc_beam_decoding.py\n@@ -14,7 +14,7 @@\n \n import math\n import os\n-from dataclasses import dataclass\n+from dataclasses import dataclass, field\n from typing import List, Optional, Tuple, Union\n \n import torch\n@@ -602,5 +602,5 @@ class BeamCTCInferConfig:\n     beam_beta: float = 0.0\n     kenlm_path: Optional[str] = None\n \n-    flashlight_cfg: Optional[FlashlightConfig] = FlashlightConfig()\n-    pyctcdecode_cfg: Optional[PyCTCDecodeConfig] = PyCTCDecodeConfig()\n+    flashlight_cfg: Optional[FlashlightConfig] = field(default_factory=lambda: FlashlightConfig())\n+    pyctcdecode_cfg: Optional[PyCTCDecodeConfig] = field(default_factory=lambda: PyCTCDecodeConfig())\ndiff --git a/nemo/collections/asr/parts/submodules/ctc_greedy_decoding.py b/nemo/collections/asr/parts/submodules/ctc_greedy_decoding.py\n--- a/nemo/collections/asr/parts/submodules/ctc_greedy_decoding.py\n+++ b/nemo/collections/asr/parts/submodules/ctc_greedy_decoding.py\n@@ -12,7 +12,7 @@\n # See the License for the specific language governing permissions and\n # limitations under the License.\n \n-from dataclasses import dataclass\n+from dataclasses import dataclass, field\n from typing import List, Optional\n \n import torch\n@@ -253,7 +253,7 @@ class GreedyCTCInferConfig:\n     preserve_alignments: bool = False\n     compute_timestamps: bool = False\n     preserve_frame_confidence: bool = False\n-    confidence_method_cfg: Optional[ConfidenceMethodConfig] = ConfidenceMethodConfig()\n+    confidence_method_cfg: Optional[ConfidenceMethodConfig] = field(default_factory=lambda: ConfidenceMethodConfig())\n \n     def __post_init__(self):\n         # OmegaConf.structured ensures that post_init check is always executed\ndiff --git a/nemo/collections/asr/parts/submodules/rnnt_greedy_decoding.py b/nemo/collections/asr/parts/submodules/rnnt_greedy_decoding.py\n--- a/nemo/collections/asr/parts/submodules/rnnt_greedy_decoding.py\n+++ b/nemo/collections/asr/parts/submodules/rnnt_greedy_decoding.py\n@@ -26,7 +26,7 @@\n # See the License for the specific language governing permissions and\n # limitations under the License.\n \n-from dataclasses import dataclass\n+from dataclasses import dataclass, field\n from typing import List, Optional, Tuple, Union\n \n import numpy as np\n@@ -2185,7 +2185,7 @@ class GreedyRNNTInferConfig:\n     max_symbols_per_step: Optional[int] = 10\n     preserve_alignments: bool = False\n     preserve_frame_confidence: bool = False\n-    confidence_method_cfg: Optional[ConfidenceMethodConfig] = ConfidenceMethodConfig()\n+    confidence_method_cfg: Optional[ConfidenceMethodConfig] = field(default_factory=lambda: ConfidenceMethodConfig())\n \n     def __post_init__(self):\n         # OmegaConf.structured ensures that post_init check is always executed\n@@ -2201,7 +2201,7 @@ class GreedyBatchedRNNTInferConfig:\n     max_symbols_per_step: Optional[int] = 10\n     preserve_alignments: bool = False\n     preserve_frame_confidence: bool = False\n-    confidence_method_cfg: Optional[ConfidenceMethodConfig] = ConfidenceMethodConfig()\n+    confidence_method_cfg: Optional[ConfidenceMethodConfig] = field(default_factory=lambda: ConfidenceMethodConfig())\n \n     def __post_init__(self):\n         # OmegaConf.structured ensures that post_init check is always executed\ndiff --git a/nemo/collections/asr/parts/utils/asr_confidence_utils.py b/nemo/collections/asr/parts/utils/asr_confidence_utils.py\n--- a/nemo/collections/asr/parts/utils/asr_confidence_utils.py\n+++ b/nemo/collections/asr/parts/utils/asr_confidence_utils.py\n@@ -14,7 +14,7 @@\n \n import math\n from abc import ABC, abstractmethod\n-from dataclasses import dataclass\n+from dataclasses import dataclass, field\n from functools import partial\n from typing import List, Optional\n \n@@ -175,7 +175,7 @@ class ConfidenceConfig:\n     preserve_word_confidence: bool = False\n     exclude_blank: bool = True\n     aggregation: str = \"min\"\n-    method_cfg: ConfidenceMethodConfig = ConfidenceMethodConfig()\n+    method_cfg: ConfidenceMethodConfig = field(default_factory=lambda: ConfidenceMethodConfig())\n \n     def __post_init__(self):\n         # OmegaConf.structured ensures that post_init check is always executed\ndiff --git a/nemo/collections/common/parts/adapter_modules.py b/nemo/collections/common/parts/adapter_modules.py\n--- a/nemo/collections/common/parts/adapter_modules.py\n+++ b/nemo/collections/common/parts/adapter_modules.py\n@@ -12,7 +12,7 @@\n # See the License for the specific language governing permissions and\n # limitations under the License.\n \n-from dataclasses import dataclass, is_dataclass\n+from dataclasses import dataclass, field, is_dataclass\n from typing import Any, Optional\n \n from hydra.utils import instantiate\n@@ -160,5 +160,7 @@ class LinearAdapterConfig:\n     activation: str = 'swish'\n     norm_position: str = 'pre'\n     dropout: float = 0.0\n-    adapter_strategy: Optional[Any] = adapter_mixin_strategies.ResidualAddAdapterStrategyConfig()\n+    adapter_strategy: Optional[Any] = field(\n+        default_factory=lambda: adapter_mixin_strategies.ResidualAddAdapterStrategyConfig()\n+    )\n     _target_: str = \"{0}.{1}\".format(LinearAdapter.__module__, LinearAdapter.__name__)\ndiff --git a/nemo/collections/common/tokenizers/en_ja_tokenizers.py b/nemo/collections/common/tokenizers/en_ja_tokenizers.py\n--- a/nemo/collections/common/tokenizers/en_ja_tokenizers.py\n+++ b/nemo/collections/common/tokenizers/en_ja_tokenizers.py\n@@ -14,11 +14,19 @@\n import re\n from typing import List\n \n-import ipadic\n-import MeCab\n from pangu import spacing\n from sacremoses import MosesDetokenizer, MosesPunctNormalizer, MosesTokenizer\n \n+try:\n+    import ipadic\n+    import MeCab\n+\n+    HAVE_MECAB = True\n+    HAVE_IPADIC = True\n+except (ImportError, ModuleNotFoundError):\n+    HAVE_MECAB = False\n+    HAVE_IPADIC = False\n+\n \n class EnJaProcessor:\n     \"\"\"\n@@ -67,6 +75,9 @@ class JaMecabProcessor:\n     \"\"\"\n \n     def __init__(self):\n+        if not HAVE_MECAB or not HAVE_IPADIC:\n+            raise ImportError(\"Please ensure that you have installed `MeCab` and `ipadic` to use JaMecabProcessor\")\n+\n         self.mecab_tokenizer = MeCab.Tagger(ipadic.MECAB_ARGS + \" -Owakati\")\n \n     def detokenize(self, text: List[str]) -> str:\ndiff --git a/nemo/collections/nlp/models/machine_translation/mt_enc_dec_config.py b/nemo/collections/nlp/models/machine_translation/mt_enc_dec_config.py\n--- a/nemo/collections/nlp/models/machine_translation/mt_enc_dec_config.py\n+++ b/nemo/collections/nlp/models/machine_translation/mt_enc_dec_config.py\n@@ -12,7 +12,7 @@\n # See the License for the specific language governing permissions and\n # limitations under the License.\n \n-from dataclasses import dataclass\n+from dataclasses import dataclass, field\n from typing import Any, Optional, Tuple\n \n from omegaconf.omegaconf import MISSING\n@@ -46,7 +46,7 @@ class MTOptimConfig(OptimConfig):\n     lr: float = 1e-3\n     betas: Tuple[float, float] = (0.9, 0.98)\n     weight_decay: float = 0.0\n-    sched: Optional[MTSchedConfig] = MTSchedConfig()\n+    sched: Optional[MTSchedConfig] = field(default_factory=lambda: MTSchedConfig())\n \n \n @dataclass\n@@ -74,70 +74,80 @@ class MTEncDecModelConfig(EncDecNLPModelConfig):\n     decoder_tokenizer: Any = MISSING\n     decoder: Any = MISSING\n \n-    head: TokenClassifierConfig = TokenClassifierConfig(log_softmax=True)\n+    head: TokenClassifierConfig = field(default_factory=lambda: TokenClassifierConfig(log_softmax=True))\n \n     # dataset configurations\n-    train_ds: Optional[TranslationDataConfig] = TranslationDataConfig(\n-        src_file_name=MISSING,\n-        tgt_file_name=MISSING,\n-        tokens_in_batch=512,\n-        clean=True,\n-        shuffle=True,\n-        cache_ids=False,\n-        use_cache=False,\n+    train_ds: Optional[TranslationDataConfig] = field(\n+        default_factory=lambda: TranslationDataConfig(\n+            src_file_name=MISSING,\n+            tgt_file_name=MISSING,\n+            tokens_in_batch=512,\n+            clean=True,\n+            shuffle=True,\n+            cache_ids=False,\n+            use_cache=False,\n+        )\n     )\n-    validation_ds: Optional[TranslationDataConfig] = TranslationDataConfig(\n-        src_file_name=MISSING,\n-        tgt_file_name=MISSING,\n-        tokens_in_batch=512,\n-        clean=False,\n-        shuffle=False,\n-        cache_ids=False,\n-        use_cache=False,\n+    validation_ds: Optional[TranslationDataConfig] = field(\n+        default_factory=lambda: TranslationDataConfig(\n+            src_file_name=MISSING,\n+            tgt_file_name=MISSING,\n+            tokens_in_batch=512,\n+            clean=False,\n+            shuffle=False,\n+            cache_ids=False,\n+            use_cache=False,\n+        )\n     )\n-    test_ds: Optional[TranslationDataConfig] = TranslationDataConfig(\n-        src_file_name=MISSING,\n-        tgt_file_name=MISSING,\n-        tokens_in_batch=512,\n-        clean=False,\n-        shuffle=False,\n-        cache_ids=False,\n-        use_cache=False,\n+    test_ds: Optional[TranslationDataConfig] = field(\n+        default_factory=lambda: TranslationDataConfig(\n+            src_file_name=MISSING,\n+            tgt_file_name=MISSING,\n+            tokens_in_batch=512,\n+            clean=False,\n+            shuffle=False,\n+            cache_ids=False,\n+            use_cache=False,\n+        )\n     )\n-    optim: Optional[OptimConfig] = MTOptimConfig()\n+    optim: Optional[OptimConfig] = field(default_factory=lambda: MTOptimConfig())\n \n \n @dataclass\n class AAYNBaseConfig(MTEncDecModelConfig):\n \n     # Attention is All You Need Base Configuration\n-    encoder_tokenizer: TokenizerConfig = TokenizerConfig(library='yttm')\n-    decoder_tokenizer: TokenizerConfig = TokenizerConfig(library='yttm')\n-\n-    encoder: NeMoTransformerEncoderConfig = NeMoTransformerEncoderConfig(\n-        library='nemo',\n-        model_name=None,\n-        pretrained=False,\n-        hidden_size=512,\n-        inner_size=2048,\n-        num_layers=6,\n-        num_attention_heads=8,\n-        ffn_dropout=0.1,\n-        attn_score_dropout=0.1,\n-        attn_layer_dropout=0.1,\n+    encoder_tokenizer: TokenizerConfig = field(default_factory=lambda: TokenizerConfig(library='yttm'))\n+    decoder_tokenizer: TokenizerConfig = field(default_factory=lambda: TokenizerConfig(library='yttm'))\n+\n+    encoder: NeMoTransformerEncoderConfig = field(\n+        default_factory=lambda: NeMoTransformerEncoderConfig(\n+            library='nemo',\n+            model_name=None,\n+            pretrained=False,\n+            hidden_size=512,\n+            inner_size=2048,\n+            num_layers=6,\n+            num_attention_heads=8,\n+            ffn_dropout=0.1,\n+            attn_score_dropout=0.1,\n+            attn_layer_dropout=0.1,\n+        )\n     )\n \n-    decoder: NeMoTransformerConfig = NeMoTransformerConfig(\n-        library='nemo',\n-        model_name=None,\n-        pretrained=False,\n-        hidden_size=512,\n-        inner_size=2048,\n-        num_layers=6,\n-        num_attention_heads=8,\n-        ffn_dropout=0.1,\n-        attn_score_dropout=0.1,\n-        attn_layer_dropout=0.1,\n+    decoder: NeMoTransformerConfig = field(\n+        default_factory=lambda: NeMoTransformerConfig(\n+            library='nemo',\n+            model_name=None,\n+            pretrained=False,\n+            hidden_size=512,\n+            inner_size=2048,\n+            num_layers=6,\n+            num_attention_heads=8,\n+            ffn_dropout=0.1,\n+            attn_score_dropout=0.1,\n+            attn_layer_dropout=0.1,\n+        )\n     )\n \n \n@@ -150,32 +160,36 @@ class MTBottleneckModelConfig(AAYNBaseConfig):\n     recon_per_token: bool = True\n     log_timing: bool = True\n \n-    encoder: NeMoTransformerBottleneckEncoderConfig = NeMoTransformerBottleneckEncoderConfig(\n-        library='nemo',\n-        model_name=None,\n-        pretrained=False,\n-        hidden_size=512,\n-        inner_size=2048,\n-        num_layers=6,\n-        num_attention_heads=8,\n-        ffn_dropout=0.1,\n-        attn_score_dropout=0.1,\n-        attn_layer_dropout=0.1,\n-        arch='seq2seq',\n-        hidden_steps=32,\n-        hidden_blocks=1,\n-        hidden_init_method='params',\n+    encoder: NeMoTransformerBottleneckEncoderConfig = field(\n+        default_factory=lambda: NeMoTransformerBottleneckEncoderConfig(\n+            library='nemo',\n+            model_name=None,\n+            pretrained=False,\n+            hidden_size=512,\n+            inner_size=2048,\n+            num_layers=6,\n+            num_attention_heads=8,\n+            ffn_dropout=0.1,\n+            attn_score_dropout=0.1,\n+            attn_layer_dropout=0.1,\n+            arch='seq2seq',\n+            hidden_steps=32,\n+            hidden_blocks=1,\n+            hidden_init_method='params',\n+        )\n     )\n \n-    decoder: NeMoTransformerBottleneckDecoderConfig = NeMoTransformerBottleneckDecoderConfig(\n-        library='nemo',\n-        model_name=None,\n-        pretrained=False,\n-        inner_size=2048,\n-        num_layers=6,\n-        num_attention_heads=8,\n-        ffn_dropout=0.1,\n-        attn_score_dropout=0.1,\n-        attn_layer_dropout=0.1,\n-        arch='seq2seq',\n+    decoder: NeMoTransformerBottleneckDecoderConfig = field(\n+        default_factory=lambda: NeMoTransformerBottleneckDecoderConfig(\n+            library='nemo',\n+            model_name=None,\n+            pretrained=False,\n+            inner_size=2048,\n+            num_layers=6,\n+            num_attention_heads=8,\n+            ffn_dropout=0.1,\n+            attn_score_dropout=0.1,\n+            attn_layer_dropout=0.1,\n+            arch='seq2seq',\n+        )\n     )\ndiff --git a/nemo/collections/nlp/models/token_classification/punctuation_capitalization_config.py b/nemo/collections/nlp/models/token_classification/punctuation_capitalization_config.py\n--- a/nemo/collections/nlp/models/token_classification/punctuation_capitalization_config.py\n+++ b/nemo/collections/nlp/models/token_classification/punctuation_capitalization_config.py\n@@ -12,7 +12,7 @@\n # See the License for the specific language governing permissions and\n # limitations under the License.\n \n-from dataclasses import dataclass\n+from dataclasses import dataclass, field\n from typing import Any, Dict, Optional\n \n from omegaconf.omegaconf import MISSING, DictConfig, OmegaConf, open_dict\n@@ -215,13 +215,15 @@ class PunctuationCapitalizationModelConfig:\n     This config is a part of :class:`~PunctuationCapitalizationConfig`.\n     \"\"\"\n \n-    class_labels: ClassLabelsConfig = ClassLabelsConfig()\n+    class_labels: ClassLabelsConfig = field(default_factory=lambda: ClassLabelsConfig())\n     \"\"\"A mandatory parameter containing a dictionary with names of label id files used in .nemo checkpoints.\n     These file names can also be used for passing label vocabularies to the model. If you wish to use ``class_labels``\n     for passing vocabularies, please provide path to vocabulary files in\n     ``model.common_dataset_parameters.label_vocab_dir`` parameter.\"\"\"\n \n-    common_dataset_parameters: Optional[CommonDatasetParametersConfig] = CommonDatasetParametersConfig()\n+    common_dataset_parameters: Optional[CommonDatasetParametersConfig] = field(\n+        default_factory=lambda: CommonDatasetParametersConfig()\n+    )\n     \"\"\"Label ids and loss mask information information.\"\"\"\n \n     train_ds: Optional[PunctuationCapitalizationTrainDataConfig] = None\n@@ -233,16 +235,16 @@ class PunctuationCapitalizationModelConfig:\n     test_ds: Optional[PunctuationCapitalizationEvalDataConfig] = None\n     \"\"\"A configuration for creating test datasets and data loaders.\"\"\"\n \n-    punct_head: HeadConfig = HeadConfig()\n+    punct_head: HeadConfig = field(default_factory=lambda: HeadConfig())\n     \"\"\"A configuration for creating punctuation MLP head that is applied to a language model outputs.\"\"\"\n \n-    capit_head: HeadConfig = HeadConfig()\n+    capit_head: HeadConfig = field(default_factory=lambda: HeadConfig())\n     \"\"\"A configuration for creating capitalization MLP head that is applied to a language model outputs.\"\"\"\n \n-    tokenizer: Any = TokenizerConfig()\n+    tokenizer: Any = field(default_factory=lambda: TokenizerConfig())\n     \"\"\"A configuration for source text tokenizer.\"\"\"\n \n-    language_model: LanguageModelConfig = LanguageModelConfig()\n+    language_model: LanguageModelConfig = field(default_factory=lambda: LanguageModelConfig())\n     \"\"\"A configuration of a BERT-like language model which serves as a model body.\"\"\"\n \n     optim: Optional[Any] = None\n@@ -311,22 +313,30 @@ class PunctuationCapitalizationConfig(NemoConfig):\n     do_testing: bool = False\n     \"\"\"Whether ot perform testing of the model.\"\"\"\n \n-    model: PunctuationCapitalizationModelConfig = PunctuationCapitalizationModelConfig()\n+    model: PunctuationCapitalizationModelConfig = field(default_factory=lambda: PunctuationCapitalizationModelConfig())\n     \"\"\"A configuration for the\n     :class:`~nemo.collections.nlp.models.token_classification.punctuation_capitalization_model.PunctuationCapitalizationModel`\n     model.\"\"\"\n \n-    trainer: Optional[TrainerConfig] = TrainerConfig()\n+    trainer: Optional[TrainerConfig] = field(default_factory=lambda: TrainerConfig())\n     \"\"\"Contains ``Trainer`` Lightning class constructor parameters.\"\"\"\n \n-    exp_manager: Optional[ExpManagerConfig] = ExpManagerConfig(name=name, files_to_copy=[])\n+    exp_manager: Optional[ExpManagerConfig] = field(\n+        default_factory=lambda: ExpManagerConfig(name=None, files_to_copy=[])\n+    )\n     \"\"\"A configuration with various NeMo training options such as output directories, resuming from checkpoint,\n     tensorboard and W&B logging, and so on. For possible options see :ref:`exp-manager-label`.\"\"\"\n \n+    def __post_init__(self):\n+        if self.exp_manager is not None:\n+            self.exp_manager.name = self.name\n+\n \n @dataclass\n class PunctuationCapitalizationLexicalAudioConfig(PunctuationCapitalizationConfig):\n-    model: PunctuationCapitalizationLexicalAudioModelConfig = PunctuationCapitalizationLexicalAudioModelConfig()\n+    model: PunctuationCapitalizationLexicalAudioModelConfig = field(\n+        default_factory=lambda: PunctuationCapitalizationLexicalAudioModelConfig()\n+    )\n \n \n def is_legacy_model_config(model_cfg: DictConfig) -> bool:\ndiff --git a/nemo/collections/nlp/modules/common/megatron/megatron_encoders.py b/nemo/collections/nlp/modules/common/megatron/megatron_encoders.py\n--- a/nemo/collections/nlp/modules/common/megatron/megatron_encoders.py\n+++ b/nemo/collections/nlp/modules/common/megatron/megatron_encoders.py\n@@ -13,7 +13,6 @@\n # limitations under the License.\n \n \"\"\"Transformer based language model.\"\"\"\n-from MeCab import Model\n from nemo.collections.nlp.modules.common.megatron.megatron_perceiver_encoders import MegatronPerceiverEncoderModule\n from nemo.collections.nlp.modules.common.megatron.megatron_transformer_encoder import MegatronTransformerEncoderModule\n from nemo.collections.nlp.modules.common.megatron.retrieval_transformer import (\n@@ -25,6 +24,13 @@\n     scaled_init_method_normal,\n )\n \n+try:\n+    from MeCab import Model\n+\n+    HAVE_MECAB = True\n+except (ImportError, ModuleNotFoundError):\n+    HAVE_MECAB = False\n+\n try:\n     from apex.transformer.enums import AttnMaskType, ModelType\n \ndiff --git a/nemo/collections/tts/models/fastpitch.py b/nemo/collections/tts/models/fastpitch.py\n--- a/nemo/collections/tts/models/fastpitch.py\n+++ b/nemo/collections/tts/models/fastpitch.py\n@@ -12,7 +12,7 @@\n # See the License for the specific language governing permissions and\n # limitations under the License.\n import contextlib\n-from dataclasses import dataclass\n+from dataclasses import dataclass, field\n from pathlib import Path\n from typing import List, Optional\n \n@@ -70,12 +70,12 @@ class TextTokenizer:\n     apostrophe: bool = True\n     pad_with_space: bool = True\n     add_blank_at: bool = True\n-    g2p: G2PConfig = G2PConfig()\n+    g2p: G2PConfig = field(default_factory=lambda: G2PConfig())\n \n \n @dataclass\n class TextTokenizerConfig:\n-    text_tokenizer: TextTokenizer = TextTokenizer()\n+    text_tokenizer: TextTokenizer = field(default_factory=lambda: TextTokenizer())\n \n \n class FastPitchModel(SpectrogramGenerator, Exportable, FastPitchAdapterModelMixin):\ndiff --git a/nemo/collections/tts/models/tacotron2.py b/nemo/collections/tts/models/tacotron2.py\n--- a/nemo/collections/tts/models/tacotron2.py\n+++ b/nemo/collections/tts/models/tacotron2.py\n@@ -13,7 +13,7 @@\n # limitations under the License.\n \n import contextlib\n-from dataclasses import dataclass\n+from dataclasses import dataclass, field\n from typing import Any, Dict, List, Optional\n \n import torch\n@@ -53,7 +53,7 @@ class Preprocessor:\n \n @dataclass\n class Tacotron2Config:\n-    preprocessor: Preprocessor = Preprocessor()\n+    preprocessor: Preprocessor = field(default_factory=lambda: Preprocessor())\n     encoder: Dict[Any, Any] = MISSING\n     decoder: Dict[Any, Any] = MISSING\n     postnet: Dict[Any, Any] = MISSING\ndiff --git a/nemo/core/config/modelPT.py b/nemo/core/config/modelPT.py\n--- a/nemo/core/config/modelPT.py\n+++ b/nemo/core/config/modelPT.py\n@@ -58,11 +58,13 @@ class HydraConfig:\n class NemoConfig:\n     name: str = MISSING\n     model: ModelConfig = MISSING\n-    trainer: config.TrainerConfig = config.TrainerConfig(\n-        strategy=\"ddp\", enable_checkpointing=False, logger=False, log_every_n_steps=1, accelerator='gpu'\n+    trainer: config.TrainerConfig = field(\n+        default_factory=lambda: config.TrainerConfig(\n+            strategy=\"ddp\", enable_checkpointing=False, logger=False, log_every_n_steps=1, accelerator='gpu'\n+        )\n     )\n-    exp_manager: Optional[Any] = exp_manager.ExpManagerConfig()\n-    hydra: HydraConfig = HydraConfig()\n+    exp_manager: Optional[Any] = field(default_factory=lambda: exp_manager.ExpManagerConfig())\n+    hydra: HydraConfig = field(default_factory=lambda: HydraConfig())\n \n \n class ModelConfigBuilder:\ndiff --git a/nemo/utils/exp_manager.py b/nemo/utils/exp_manager.py\n--- a/nemo/utils/exp_manager.py\n+++ b/nemo/utils/exp_manager.py\n@@ -18,7 +18,7 @@\n import sys\n import time\n import warnings\n-from dataclasses import dataclass\n+from dataclasses import dataclass, field\n from datetime import timedelta\n from pathlib import Path\n from shutil import copy, move\n@@ -146,28 +146,30 @@ class ExpManagerConfig:\n     create_wandb_logger: Optional[bool] = False\n     wandb_logger_kwargs: Optional[Dict[Any, Any]] = None\n     create_mlflow_logger: Optional[bool] = False\n-    mlflow_logger_kwargs: Optional[MLFlowParams] = MLFlowParams()\n+    mlflow_logger_kwargs: Optional[MLFlowParams] = field(default_factory=lambda: MLFlowParams())\n     create_dllogger_logger: Optional[bool] = False\n-    dllogger_logger_kwargs: Optional[DLLoggerParams] = DLLoggerParams()\n+    dllogger_logger_kwargs: Optional[DLLoggerParams] = field(default_factory=lambda: DLLoggerParams())\n     create_clearml_logger: Optional[bool] = False\n-    clearml_logger_kwargs: Optional[ClearMLParams] = ClearMLParams()\n+    clearml_logger_kwargs: Optional[ClearMLParams] = field(default_factory=lambda: ClearMLParams())\n     # Checkpointing parameters\n     create_checkpoint_callback: Optional[bool] = True\n-    checkpoint_callback_params: Optional[CallbackParams] = CallbackParams()\n+    checkpoint_callback_params: Optional[CallbackParams] = field(default_factory=lambda: CallbackParams())\n     create_early_stopping_callback: Optional[bool] = False\n-    early_stopping_callback_params: Optional[EarlyStoppingParams] = EarlyStoppingParams()\n+    early_stopping_callback_params: Optional[EarlyStoppingParams] = field(\n+        default_factory=lambda: EarlyStoppingParams()\n+    )\n     create_preemption_callback: Optional[bool] = True\n     # Additional exp_manager arguments\n     files_to_copy: Optional[List[str]] = None\n     # logs timing of train/val/test steps\n     log_step_timing: Optional[bool] = True\n-    step_timing_kwargs: Optional[StepTimingParams] = StepTimingParams()\n+    step_timing_kwargs: Optional[StepTimingParams] = field(default_factory=lambda: StepTimingParams())\n     # Configures creation of log files for different ranks\n     log_local_rank_0_only: Optional[bool] = False\n     log_global_rank_0_only: Optional[bool] = False\n     # disable initial validation when resuming from a checkpoint saved during validation\n     disable_validation_on_resume: Optional[bool] = True\n-    ema: Optional[EMAParams] = EMAParams()\n+    ema: Optional[EMAParams] = field(default_factory=lambda: EMAParams())\n     # Wall clock time limit\n     max_time_per_run: Optional[str] = None\n     # time to sleep non 0 ranks during initialization\ndiff --git a/scripts/asr_language_modeling/ngram_lm/eval_beamsearch_ngram.py b/scripts/asr_language_modeling/ngram_lm/eval_beamsearch_ngram.py\n--- a/scripts/asr_language_modeling/ngram_lm/eval_beamsearch_ngram.py\n+++ b/scripts/asr_language_modeling/ngram_lm/eval_beamsearch_ngram.py\n@@ -112,14 +112,14 @@ class EvalBeamSearchNGramConfig:\n     beam_beta: List[float] = field(default_factory=lambda: [0.0])  # The beta parameter or list of the betas for the beam search decoding\n \n     decoding_strategy: str = \"beam\"\n-    decoding: ctc_beam_decoding.BeamCTCInferConfig = ctc_beam_decoding.BeamCTCInferConfig(beam_size=128)\n+    decoding: ctc_beam_decoding.BeamCTCInferConfig = field(default_factory=lambda: ctc_beam_decoding.BeamCTCInferConfig(beam_size=128))\n     \n-    text_processing: Optional[TextProcessingConfig] = TextProcessingConfig(\n+    text_processing: Optional[TextProcessingConfig] = field(default_factory=lambda: TextProcessingConfig(\n         punctuation_marks = \".,?\",\n         separate_punctuation = False,\n         do_lowercase = False,\n         rm_punctuation = False,\n-    )\n+    ))\n # fmt: on\n \n \ndiff --git a/scripts/asr_language_modeling/ngram_lm/eval_beamsearch_ngram_transducer.py b/scripts/asr_language_modeling/ngram_lm/eval_beamsearch_ngram_transducer.py\n--- a/scripts/asr_language_modeling/ngram_lm/eval_beamsearch_ngram_transducer.py\n+++ b/scripts/asr_language_modeling/ngram_lm/eval_beamsearch_ngram_transducer.py\n@@ -115,7 +115,7 @@ class EvalBeamSearchNGramConfig:\n     hat_subtract_ilm: bool = False\n     hat_ilm_weight: List[float] = field(default_factory=lambda: [0.0])\n \n-    decoding: rnnt_beam_decoding.BeamRNNTInferConfig = rnnt_beam_decoding.BeamRNNTInferConfig(beam_size=128)\n+    decoding: rnnt_beam_decoding.BeamRNNTInferConfig = field(default_factory=lambda: rnnt_beam_decoding.BeamRNNTInferConfig(beam_size=128))\n \n \n # fmt: on\ndiff --git a/scripts/confidence_ensembles/build_ensemble.py b/scripts/confidence_ensembles/build_ensemble.py\n--- a/scripts/confidence_ensembles/build_ensemble.py\n+++ b/scripts/confidence_ensembles/build_ensemble.py\n@@ -75,7 +75,7 @@\n import sys\n import tempfile\n from copy import deepcopy\n-from dataclasses import dataclass\n+from dataclasses import dataclass, field\n from pathlib import Path\n from typing import Dict, List, Optional, Tuple\n \n@@ -209,19 +209,23 @@ class BuildEnsembleConfig:\n     random_seed: int = 0  # for reproducibility\n \n     # default confidence, can override\n-    confidence: ConfidenceConfig = ConfidenceConfig(\n-        # we keep frame confidences and apply aggregation manually to get full-utterance confidence\n-        preserve_frame_confidence=True,\n-        exclude_blank=True,\n-        aggregation=\"mean\",\n-        method_cfg=ConfidenceMethodConfig(name=\"entropy\", entropy_type=\"renyi\", alpha=0.25, entropy_norm=\"lin\",),\n+    confidence: ConfidenceConfig = field(\n+        default_factory=lambda: ConfidenceConfig(\n+            # we keep frame confidences and apply aggregation manually to get full-utterance confidence\n+            preserve_frame_confidence=True,\n+            exclude_blank=True,\n+            aggregation=\"mean\",\n+            measure_cfg=ConfidenceMethodConfig(name=\"entropy\", entropy_type=\"renyi\", alpha=0.25, entropy_norm=\"lin\",),\n+        )\n     )\n     temperature: float = 1.0\n \n     # this is optional, but can be used to change any aspect of the transcription\n     # config, such as batch size or amp usage. Note that model, data and confidence\n     # will be overriden by this script\n-    transcription: transcribe_speech.TranscriptionConfig = transcribe_speech.TranscriptionConfig()\n+    transcription: transcribe_speech.TranscriptionConfig = field(\n+        default_factory=lambda: transcribe_speech.TranscriptionConfig()\n+    )\n \n     # set to True to tune the confidence.\n     # requires dev manifests to be specified for each model\n@@ -229,12 +233,14 @@ class BuildEnsembleConfig:\n     # used to specify what to tune over. By default runs tuning over some\n     # reasonalbe grid, so that it does not take forever.\n     # Can be changed as needed\n-    tune_confidence_config: TuneConfidenceConfig = TuneConfidenceConfig()\n+    tune_confidence_config: TuneConfidenceConfig = field(default_factory=lambda: TuneConfidenceConfig())\n \n     # very fast to tune and can be important in case of imbalanced datasets\n     # will automatically set to False if dev data is not available\n     tune_logistic_regression: bool = True\n-    tune_logistic_regression_config: TuneLogisticRegressionConfig = TuneLogisticRegressionConfig()\n+    tune_logistic_regression_config: TuneLogisticRegressionConfig = field(\n+        default_factory=lambda: TuneLogisticRegressionConfig()\n+    )\n \n     def __post_init__(self):\n         \"\"\"Checking that if any dev data is provided, all are provided.\ndiff --git a/scripts/speech_recognition/confidence/benchmark_asr_confidence.py b/scripts/speech_recognition/confidence/benchmark_asr_confidence.py\n--- a/scripts/speech_recognition/confidence/benchmark_asr_confidence.py\n+++ b/scripts/speech_recognition/confidence/benchmark_asr_confidence.py\n@@ -14,7 +14,7 @@\n \n import json\n import os\n-from dataclasses import dataclass, is_dataclass\n+from dataclasses import dataclass, field, is_dataclass\n from pathlib import Path\n from typing import Optional\n \n@@ -124,7 +124,9 @@ class ConfidenceBenchmarkingConfig:\n \n     # Confidence configs\n     target_level: str = \"auto\"  # Choices: \"word\", \"token\", \"auto\" (for both word- and token-level confidence)\n-    confidence_cfg: ConfidenceConfig = ConfidenceConfig(preserve_word_confidence=True, preserve_token_confidence=True)\n+    confidence_cfg: ConfidenceConfig = field(\n+        default_factory=lambda: ConfidenceConfig(preserve_word_confidence=True, preserve_token_confidence=True)\n+    )\n     grid_params: Optional[str] = None  # a dictionary with lists of parameters to iteratively benchmark on\n \n \ndiff --git a/scripts/speech_recognition/convert_to_tarred_audio_dataset.py b/scripts/speech_recognition/convert_to_tarred_audio_dataset.py\n--- a/scripts/speech_recognition/convert_to_tarred_audio_dataset.py\n+++ b/scripts/speech_recognition/convert_to_tarred_audio_dataset.py\n@@ -202,7 +202,7 @@ class ASRTarredDatasetMetadata:\n     num_samples_per_shard: Optional[int] = None\n     is_concatenated_manifest: bool = False\n \n-    dataset_config: Optional[ASRTarredDatasetConfig] = ASRTarredDatasetConfig()\n+    dataset_config: Optional[ASRTarredDatasetConfig] = field(default_factory=lambda: ASRTarredDatasetConfig())\n     history: Optional[List[Any]] = field(default_factory=lambda: [])\n \n     def __post_init__(self):\ndiff --git a/tools/nemo_forced_aligner/align.py b/tools/nemo_forced_aligner/align.py\n--- a/tools/nemo_forced_aligner/align.py\n+++ b/tools/nemo_forced_aligner/align.py\n@@ -149,8 +149,8 @@ class AlignmentConfig:\n \n     # Output file configs\n     save_output_file_formats: List[str] = field(default_factory=lambda: [\"ctm\", \"ass\"])\n-    ctm_file_config: CTMFileConfig = CTMFileConfig()\n-    ass_file_config: ASSFileConfig = ASSFileConfig()\n+    ctm_file_config: CTMFileConfig = field(default_factory=lambda: CTMFileConfig())\n+    ass_file_config: ASSFileConfig = field(default_factory=lambda: ASSFileConfig())\n \n \n @hydra_runner(config_name=\"AlignmentConfig\", schema=AlignmentConfig)\n",
-    "test_patch": "diff --git a/tests/collections/asr/test_text_to_text_dataset.py b/tests/collections/asr/test_text_to_text_dataset.py\n--- a/tests/collections/asr/test_text_to_text_dataset.py\n+++ b/tests/collections/asr/test_text_to_text_dataset.py\n@@ -15,7 +15,7 @@\n import json\n import multiprocessing\n import os\n-from dataclasses import dataclass\n+from dataclasses import dataclass, field\n from pathlib import Path\n \n import pytest\n@@ -118,7 +118,7 @@ class TextTokenizerCfg:\n         apostrophe: bool = True\n         pad_with_space: bool = True\n         add_blank_at: bool = True\n-        g2p: G2PConfig = G2PConfig()\n+        g2p: G2PConfig = field(default_factory=lambda: G2PConfig())\n \n     config = OmegaConf.create(OmegaConf.to_yaml(TextTokenizerCfg()))\n     return instantiate(config)\n",
-    "problem_statement": "Ubuntu 22.04 Python 3.11 [asr]: multiple errors `dataclasses ValueError: mutable default * for field * is not allowed: use default_factory`\n**Describe the bug**\r\n\r\nAfter installing latest stable `1.19.1` from pipI, or the latest current commit with `[asr]` extras, I'm getting this error from `hydra-core==1.2.0` when trying to import `nemo.collections.asr`:\r\n```\r\nTraceback (most recent call last):\r\n  File \"/home/alex/T7/src/speaker_verification/nemo_speaker/test.py\", line 5, in <module>\r\n    import nemo.collections.asr as nemo_asr\r\n  File \"/home/alex/T7/src/speaker_verification/nemo_speaker/venv11/lib/python3.11/site-packages/nemo/collections/asr/__init__.py\", line 15, in <module>\r\n    from nemo.collections.asr import data, losses, models, modules\r\n  File \"/home/alex/T7/src/speaker_verification/nemo_speaker/venv11/lib/python3.11/site-packages/nemo/collections/asr/losses/__init__.py\", line 15, in <module>\r\n    from nemo.collections.asr.losses.angularloss import AngularSoftmaxLoss\r\n  File \"/home/alex/T7/src/speaker_verification/nemo_speaker/venv11/lib/python3.11/site-packages/nemo/collections/asr/losses/angularloss.py\", line 18, in <module>\r\n    from nemo.core.classes import Loss, Typing, typecheck\r\n  File \"/home/alex/T7/src/speaker_verification/nemo_speaker/venv11/lib/python3.11/site-packages/nemo/core/__init__.py\", line 16, in <module>\r\n    from nemo.core.classes import *\r\n  File \"/home/alex/T7/src/speaker_verification/nemo_speaker/venv11/lib/python3.11/site-packages/nemo/core/classes/__init__.py\", line 16, in <module>\r\n    import hydra\r\n  File \"/home/alex/T7/src/speaker_verification/nemo_speaker/venv11/lib/python3.11/site-packages/hydra/__init__.py\", line 5, in <module>\r\n    from hydra import utils\r\n  File \"/home/alex/T7/src/speaker_verification/nemo_speaker/venv11/lib/python3.11/site-packages/hydra/utils.py\", line 8, in <module>\r\n    import hydra._internal.instantiate._instantiate2\r\n  File \"/home/alex/T7/src/speaker_verification/nemo_speaker/venv11/lib/python3.11/site-packages/hydra/_internal/instantiate/_instantiate2.py\", line 12, in <module>\r\n    from hydra._internal.utils import _locate\r\n  File \"/home/alex/T7/src/speaker_verification/nemo_speaker/venv11/lib/python3.11/site-packages/hydra/_internal/utils.py\", line 18, in <module>\r\n    from hydra.core.utils import get_valid_filename, validate_config_path\r\n  File \"/home/alex/T7/src/speaker_verification/nemo_speaker/venv11/lib/python3.11/site-packages/hydra/core/utils.py\", line 20, in <module>\r\n    from hydra.core.hydra_config import HydraConfig\r\n  File \"/home/alex/T7/src/speaker_verification/nemo_speaker/venv11/lib/python3.11/site-packages/hydra/core/hydra_config.py\", line 6, in <module>\r\n    from hydra.conf import HydraConf\r\n  File \"/home/alex/T7/src/speaker_verification/nemo_speaker/venv11/lib/python3.11/site-packages/hydra/conf/__init__.py\", line 46, in <module>\r\n    class JobConf:\r\n  File \"/home/alex/T7/src/speaker_verification/nemo_speaker/venv11/lib/python3.11/site-packages/hydra/conf/__init__.py\", line 75, in JobConf\r\n    @dataclass\r\n     ^^^^^^^^^\r\n  File \"/usr/lib/python3.11/dataclasses.py\", line 1230, in dataclass\r\n    return wrap(cls)\r\n           ^^^^^^^^^\r\n  File \"/usr/lib/python3.11/dataclasses.py\", line 1220, in wrap\r\n    return _process_class(cls, init, repr, eq, order, unsafe_hash,\r\n           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\r\n  File \"/usr/lib/python3.11/dataclasses.py\", line 958, in _process_class\r\n    cls_fields.append(_get_field(cls, name, type, kw_only))\r\n                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\r\n  File \"/usr/lib/python3.11/dataclasses.py\", line 815, in _get_field\r\n    raise ValueError(f'mutable default {type(f.default)} for field '\r\nValueError: mutable default <class 'hydra.conf.JobConf.JobConfig.OverrideDirname'> for field override_dirname is not allowed: use default_factory\r\n```\r\nIf I then manually upgrade `hydra-core` to the current latest (`1.3.2`), I get similar errors from `nemo-toolkit`:\r\n```\r\nTraceback (most recent call last):\r\n  File \"/home/alex/T7/src/speaker_verification/nemo_speaker/test.py\", line 5, in <module>\r\n    import nemo.collections.asr as nemo_asr\r\n  File \"/home/alex/T7/src/speaker_verification/nemo_speaker/venv11/lib/python3.11/site-packages/nemo/collections/asr/__init__.py\", line 15, in <module>\r\n    from nemo.collections.asr import data, losses, models, modules\r\n  File \"/home/alex/T7/src/speaker_verification/nemo_speaker/venv11/lib/python3.11/site-packages/nemo/collections/asr/losses/__init__.py\", line 16, in <module>\r\n    from nemo.collections.asr.losses.audio_losses import SDRLoss\r\n  File \"/home/alex/T7/src/speaker_verification/nemo_speaker/venv11/lib/python3.11/site-packages/nemo/collections/asr/losses/audio_losses.py\", line 21, in <module>\r\n    from nemo.collections.asr.parts.preprocessing.features import make_seq_mask_like\r\n  File \"/home/alex/T7/src/speaker_verification/nemo_speaker/venv11/lib/python3.11/site-packages/nemo/collections/asr/parts/preprocessing/__init__.py\", line 16, in <module>\r\n    from nemo.collections.asr.parts.preprocessing.features import FeaturizerFactory, FilterbankFeatures, WaveformFeaturizer\r\n  File \"/home/alex/T7/src/speaker_verification/nemo_speaker/venv11/lib/python3.11/site-packages/nemo/collections/asr/parts/preprocessing/features.py\", line 44, in <module>\r\n    from nemo.collections.asr.parts.preprocessing.perturb import AudioAugmentor\r\n  File \"/home/alex/T7/src/speaker_verification/nemo_speaker/venv11/lib/python3.11/site-packages/nemo/collections/asr/parts/preprocessing/perturb.py\", line 50, in <module>\r\n    from nemo.collections.common.parts.preprocessing import collections, parsers\r\n  File \"/home/alex/T7/src/speaker_verification/nemo_speaker/venv11/lib/python3.11/site-packages/nemo/collections/common/__init__.py\", line 16, in <module>\r\n    from nemo.collections.common import data, losses, parts, tokenizers\r\n  File \"/home/alex/T7/src/speaker_verification/nemo_speaker/venv11/lib/python3.11/site-packages/nemo/collections/common/parts/__init__.py\", line 15, in <module>\r\n    from nemo.collections.common.parts.adapter_modules import LinearAdapter, LinearAdapterConfig\r\n  File \"/home/alex/T7/src/speaker_verification/nemo_speaker/venv11/lib/python3.11/site-packages/nemo/collections/common/parts/adapter_modules.py\", line 147, in <module>\r\n    @dataclass\r\n     ^^^^^^^^^\r\n  File \"/usr/lib/python3.11/dataclasses.py\", line 1230, in dataclass\r\n    return wrap(cls)\r\n           ^^^^^^^^^\r\n  File \"/usr/lib/python3.11/dataclasses.py\", line 1220, in wrap\r\n    return _process_class(cls, init, repr, eq, order, unsafe_hash,\r\n           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\r\n  File \"/usr/lib/python3.11/dataclasses.py\", line 958, in _process_class\r\n    cls_fields.append(_get_field(cls, name, type, kw_only))\r\n                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\r\n  File \"/usr/lib/python3.11/dataclasses.py\", line 815, in _get_field\r\n    raise ValueError(f'mutable default {type(f.default)} for field '\r\nValueError: mutable default <class 'nemo.core.classes.mixins.adapter_mixin_strategies.ResidualAddAdapterStrategyConfig'> for field adapter_strategy is not allowed: use default_factory\r\n```\r\nIt's easy to fix with a patch like this:\r\n```\r\n--- nemo/collections/common/parts/adapter_modules.py.orig       2023-08-04 13:55:53.464534800 +0200\r\n+++ nemo/collections/common/parts/adapter_modules.py    2023-08-04 14:05:45.579537700 +0200\r\n@@ -12,7 +12,7 @@\r\n # See the License for the specific language governing permissions and\r\n # limitations under the License.\r\n \r\n-from dataclasses import dataclass, is_dataclass\r\n+from dataclasses import dataclass, is_dataclass, field\r\n from typing import Any, Optional\r\n \r\n from hydra.utils import instantiate\r\n@@ -151,5 +151,5 @@ class LinearAdapterConfig:\r\n     activation: str = 'swish'\r\n     norm_position: str = 'pre'\r\n     dropout: float = 0.0\r\n-    adapter_strategy: Optional[Any] = adapter_mixin_strategies.ResidualAddAdapterStrategyConfig()\r\n+    adapter_strategy: Optional[Any] = field(default_factory=adapter_mixin_strategies.ResidualAddAdapterStrategyConfig)\r\n     _target_: str = \"{0}.{1}\".format(LinearAdapter.__module__, LinearAdapter.__name__)\r\n```\r\nHowever then another error of the kind comes up:\r\n```\r\nTraceback (most recent call last):\r\n  File \"/home/alex/T7/src/speaker_verification/nemo_speaker/test.py\", line 5, in <module>\r\n    import nemo.collections.asr as nemo_asr\r\n  File \"/home/alex/T7/src/speaker_verification/nemo_speaker/venv11/lib/python3.11/site-packages/nemo/collections/asr/__init__.py\", line 15, in <module>\r\n    from nemo.collections.asr import data, losses, models, modules\r\n  File \"/home/alex/T7/src/speaker_verification/nemo_speaker/venv11/lib/python3.11/site-packages/nemo/collections/asr/models/__init__.py\", line 18, in <module>\r\n    from nemo.collections.asr.models.clustering_diarizer import ClusteringDiarizer\r\n  File \"/home/alex/T7/src/speaker_verification/nemo_speaker/venv11/lib/python3.11/site-packages/nemo/collections/asr/models/clustering_diarizer.py\", line 29, in <module>\r\n    from nemo.collections.asr.metrics.der import score_labels\r\n  File \"/home/alex/T7/src/speaker_verification/nemo_speaker/venv11/lib/python3.11/site-packages/nemo/collections/asr/metrics/der.py\", line 24, in <module>\r\n    from nemo.collections.asr.metrics.wer import word_error_rate\r\n  File \"/home/alex/T7/src/speaker_verification/nemo_speaker/venv11/lib/python3.11/site-packages/nemo/collections/asr/metrics/wer.py\", line 27, in <module>\r\n    from nemo.collections.asr.parts.submodules import ctc_beam_decoding, ctc_greedy_decoding\r\n  File \"/home/alex/T7/src/speaker_verification/nemo_speaker/venv11/lib/python3.11/site-packages/nemo/collections/asr/parts/submodules/ctc_beam_decoding.py\", line 593, in <module>\r\n    @dataclass\r\n     ^^^^^^^^^\r\n  File \"/usr/lib/python3.11/dataclasses.py\", line 1230, in dataclass\r\n    return wrap(cls)\r\n           ^^^^^^^^^\r\n  File \"/usr/lib/python3.11/dataclasses.py\", line 1220, in wrap\r\n    return _process_class(cls, init, repr, eq, order, unsafe_hash,\r\n           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\r\n  File \"/usr/lib/python3.11/dataclasses.py\", line 958, in _process_class\r\n    cls_fields.append(_get_field(cls, name, type, kw_only))\r\n                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\r\n  File \"/usr/lib/python3.11/dataclasses.py\", line 815, in _get_field\r\n    raise ValueError(f'mutable default {type(f.default)} for field '\r\nValueError: mutable default <class 'nemo.collections.asr.parts.submodules.ctc_beam_decoding.FlashlightConfig'> for field flashlight_cfg is not allowed: use default_factory\r\n```\r\nThis can also be fixed accordingly, but all in all, it appears these issues are pretty common in the code base.\r\n\r\nLooks like NeMo isn't Python 3.11 ready, at least the `asr` collection.\r\n\r\n**Steps/Code to reproduce bug**\r\n\r\nWith Python 3.11\r\n1. Install `nemo-toolkit[asr]` either 1.19.1, or current HEAD from git (b498d438fc4c35ebf364a9a1c5cd3e29a2c0fe50)\r\n2. Run `import nemo.collections.asr`\r\n\r\n\r\n**Expected behavior**\r\n\r\nImport `nemo.collections.asr` without errors\r\n\r\n**Environment overview (please complete the following information)**\r\n\r\n - Environment location: Bare-metal\r\n - Method of NeMo install: `pip install nemo-toolkit==1.19.1` or from source b498d438fc4c35ebf364a9a1c5cd3e29a2c0fe50\r\n\r\n**Environment details**\r\n- Ubuntu 22.04\r\n- PyTorch version 2.0.1\r\n- Python version 3.11.4\r\n\r\n\n",
-    "hints_text": "Seems to be a similar to #7002\nInteresting. The fix is easy but needs to be applied to basically every single place that has this constructor for our adapter configs. Let me see if I can update it. But no guarantees on how soon fixes will come in main. \nLooking forward to it @titu1994 ! Thanks \ud83d\ude03 \n@titu1994 I was looking to use NeMo speaker diarization with Python 3.11 and hit this dataclass issue. I patched everything involved in the specific code paths I needed: https://github.com/lmnt-com/NeMo/commit/d89acf9f0152e43dee29d7d1c4667ee34c26ffd7\r\n\r\nI was using the neural diarizer as described in https://github.com/NVIDIA/NeMo/tree/main/examples/speaker_tasks/diarization\r\n\r\nI'd be happy to upstream this if it's helpful.\r\n\r\nI haven't checked whether this is backwards compatible for earlier python/dataclass versions, do you know?\r\n\r\nFor reference, what led me to this issue, though it's duplicative to the above discussion:\r\n\r\n[Similar error](https://github.com/huggingface/datasets/issues/5230)\r\n[StackOverflow solution](https://stackoverflow.com/questions/53632152/why-cant-dataclasses-have-mutable-defaults-in-their-class-attributes-declaratio)\n@shaper Thanks for sharing. For brevity, you don't really need a `lambda` when you don't pass any init parameters, like this:\r\n```\r\nfield(default_factory=lambda: ConfidenceConfig())\r\n```\r\nYou can just do\r\n```\r\nfield(default_factory=ConfidenceConfig)\r\n```\r\nIt's only needed when you do pass parameter(s), like\r\n```\r\nfield(default_factory=lambda: beam_decode.BeamRNNTInferConfig(beam_size=4))\r\n```\nThis issue is stale because it has been open for 30 days with no activity. Remove stale label or comment or this will be closed in 7 days.\nI have the same issue. @tango4j suggested using one of the models from https://huggingface.co/spaces/hf-audio/open_asr_leaderboard, but I cannot import nemo.collections.asr:\r\n\r\n```\r\n    Traceback (most recent call last):\r\n      File \"/opt/pycharm-2022.3.3/plugins/python/helpers/pycharm/docrunner.py\", line 138, in __run\r\n        exec(compile(example.source, filename, \"single\",\r\n      File \"<doctest NeMoASR[2]>\", line 1, in <module>\r\n        NeMoASR().apply_asr(file)\r\n        ^^^^^^^^^\r\n      File \"/home/cbj/python/cbj/cbj/transcribe/pretrained.py\", line 504, in __init__\r\n        import nemo.collections.asr as nemo_asr\r\n      File \"/opt/py/2023/lib/python3.11/site-packages/nemo/collections/asr/__init__.py\", line 15, in <module>\r\n        from nemo.collections.asr import data, losses, models, modules\r\n      File \"/opt/py/2023/lib/python3.11/site-packages/nemo/collections/asr/losses/__init__.py\", line 16, in <module>\r\n        from nemo.collections.asr.losses.audio_losses import SDRLoss\r\n      File \"/opt/py/2023/lib/python3.11/site-packages/nemo/collections/asr/losses/audio_losses.py\", line 21, in <module>\r\n        from nemo.collections.asr.parts.preprocessing.features import make_seq_mask_like\r\n      File \"/opt/py/2023/lib/python3.11/site-packages/nemo/collections/asr/parts/preprocessing/__init__.py\", line 16, in <module>\r\n        from nemo.collections.asr.parts.preprocessing.features import FeaturizerFactory, FilterbankFeatures, WaveformFeaturizer\r\n      File \"/opt/py/2023/lib/python3.11/site-packages/nemo/collections/asr/parts/preprocessing/features.py\", line 44, in <module>\r\n        from nemo.collections.asr.parts.preprocessing.perturb import AudioAugmentor\r\n      File \"/opt/py/2023/lib/python3.11/site-packages/nemo/collections/asr/parts/preprocessing/perturb.py\", line 50, in <module>\r\n        from nemo.collections.common.parts.preprocessing import collections, parsers\r\n      File \"/opt/py/2023/lib/python3.11/site-packages/nemo/collections/common/__init__.py\", line 16, in <module>\r\n        from nemo.collections.common import data, losses, parts, tokenizers\r\n      File \"/opt/py/2023/lib/python3.11/site-packages/nemo/collections/common/parts/__init__.py\", line 15, in <module>\r\n        from nemo.collections.common.parts.adapter_modules import LinearAdapter, LinearAdapterConfig\r\n      File \"/opt/py/2023/lib/python3.11/site-packages/nemo/collections/common/parts/adapter_modules.py\", line 147, in <module>\r\n        @dataclass\r\n         ^^^^^^^^^\r\n      File \"/opt/py/2023/lib/python3.11/dataclasses.py\", line 1230, in dataclass\r\n        return wrap(cls)\r\n               ^^^^^^^^^\r\n      File \"/opt/py/2023/lib/python3.11/dataclasses.py\", line 1220, in wrap\r\n        return _process_class(cls, init, repr, eq, order, unsafe_hash,\r\n               ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\r\n      File \"/opt/py/2023/lib/python3.11/dataclasses.py\", line 958, in _process_class\r\n        cls_fields.append(_get_field(cls, name, type, kw_only))\r\n                          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\r\n      File \"/opt/py/2023/lib/python3.11/dataclasses.py\", line 815, in _get_field\r\n        raise ValueError(f'mutable default {type(f.default)} for field '\r\n    ValueError: mutable default <class 'nemo.core.classes.mixins.adapter_mixin_strategies.ResidualAddAdapterStrategyConfig'> for field adapter_strategy is not allowed: use default_factory\r\n```\r\n\r\nFor documentation (I had to search in the provided links):\r\nMutable defaults were never allowed in dataclasses (by convention), but in python 3.11 they improved the check: Instead of checking some types (dict, list, set) they now use hashable as indicator for mutable.\r\n\r\nAn alternative to default_factory would be to use frozen dataclasses, but I don't know whether in this code base the configs are used as mutable objects or not.\nYou need to update to NeMo 1.20, omegaconf did a fix that should resolve this \nI have NeMo 1.20.0. \r\nWith `pip install nemo_toolkit` and `pip install pytorch_lightning` I installed yesterday nemo.\r\nSo it should be the newest PyPI version.\r\n\r\n```\r\n$ pip show nemo_toolkit\r\nName: nemo-toolkit\r\nVersion: 1.20.0\r\nSummary: NeMo - a toolkit for Conversational AI\r\nHome-page: https://github.com/nvidia/nemo\r\nAuthor: NVIDIA\r\nAuthor-email: nemo-toolkit@nvidia.com\r\nLicense: Apache2\r\nLocation: /opt/py/2023/lib/python3.11/site-packages\r\nRequires: huggingface-hub, numba, numpy, onnx, python-dateutil, ruamel.yaml, scikit-learn, setuptools, tensorboard, text-unidecode, torch, tqdm, wget, wrapt\r\nRequired-by: \r\n\r\n$ pip show omegaconf\r\nName: omegaconf\r\nVersion: 2.3.0\r\nSummary: A flexible configuration library\r\nHome-page: https://github.com/omry/omegaconf\r\nAuthor: Omry Yadan\r\nAuthor-email: omry@yadan.net\r\nLicense: \r\nLocation: /home/cbj/.local/lib/python3.11/site-packages\r\nRequires: antlr4-python3-runtime, PyYAML\r\nRequired-by: hydra-core\r\n\r\n$ python -c \"import nemo.collections.asr as nemo_asr\"\r\nTraceback (most recent call last):\r\n  File \"<string>\", line 1, in <module>\r\n  File \"/opt/py/2023/lib/python3.11/site-packages/nemo/collections/asr/__init__.py\", line 15, in <module>\r\n    from nemo.collections.asr import data, losses, models, modules\r\n  File \"/opt/py/2023/lib/python3.11/site-packages/nemo/collections/asr/losses/__init__.py\", line 16, in <module>\r\n    from nemo.collections.asr.losses.audio_losses import SDRLoss\r\n  File \"/opt/py/2023/lib/python3.11/site-packages/nemo/collections/asr/losses/audio_losses.py\", line 21, in <module>\r\n    from nemo.collections.asr.parts.preprocessing.features import make_seq_mask_like\r\n  File \"/opt/py/2023/lib/python3.11/site-packages/nemo/collections/asr/parts/preprocessing/__init__.py\", line 16, in <module>\r\n    from nemo.collections.asr.parts.preprocessing.features import FeaturizerFactory, FilterbankFeatures, WaveformFeaturizer\r\n  File \"/opt/py/2023/lib/python3.11/site-packages/nemo/collections/asr/parts/preprocessing/features.py\", line 44, in <module>\r\n    from nemo.collections.asr.parts.preprocessing.perturb import AudioAugmentor\r\n  File \"/opt/py/2023/lib/python3.11/site-packages/nemo/collections/asr/parts/preprocessing/perturb.py\", line 50, in <module>\r\n    from nemo.collections.common.parts.preprocessing import collections, parsers\r\n  File \"/opt/py/2023/lib/python3.11/site-packages/nemo/collections/common/__init__.py\", line 16, in <module>\r\n    from nemo.collections.common import data, losses, parts, tokenizers\r\n  File \"/opt/py/2023/lib/python3.11/site-packages/nemo/collections/common/parts/__init__.py\", line 15, in <module>\r\n    from nemo.collections.common.parts.adapter_modules import LinearAdapter, LinearAdapterConfig\r\n  File \"/opt/py/2023/lib/python3.11/site-packages/nemo/collections/common/parts/adapter_modules.py\", line 147, in <module>\r\n    @dataclass\r\n     ^^^^^^^^^\r\n  File \"/opt/py/2023/lib/python3.11/dataclasses.py\", line 1230, in dataclass\r\n    return wrap(cls)\r\n           ^^^^^^^^^\r\n  File \"/opt/py/2023/lib/python3.11/dataclasses.py\", line 1220, in wrap\r\n    return _process_class(cls, init, repr, eq, order, unsafe_hash,\r\n           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\r\n  File \"/opt/py/2023/lib/python3.11/dataclasses.py\", line 958, in _process_class\r\n    cls_fields.append(_get_field(cls, name, type, kw_only))\r\n                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\r\n  File \"/opt/py/2023/lib/python3.11/dataclasses.py\", line 815, in _get_field\r\n    raise ValueError(f'mutable default {type(f.default)} for field '\r\nValueError: mutable default <class 'nemo.core.classes.mixins.adapter_mixin_strategies.ResidualAddAdapterStrategyConfig'> for field adapter_strategy is not allowed: use default_factory\r\n\r\n```\nHmm ok I'll take a look ",
-    "created_at": "2023-10-03T19:14:38Z",
-    "version": "1.0"
-  },
-  {
-    "repo": "slackapi/python-slack-events-api",
-    "pull_number": 71,
-    "instance_id": "slackapi__python-slack-events-api-71",
-    "issue_numbers": [
-      "66",
-      "66"
-    ],
-    "base_commit": "0c0ce604b502508622fb14c278a0d64841fa32e3",
-    "patch": "diff --git a/example/current_app/main.py b/example/current_app/main.py\nnew file mode 100644\n--- /dev/null\n+++ b/example/current_app/main.py\n@@ -0,0 +1,49 @@\n+# ------------------\n+# Only for running this script here\n+import sys\n+from os.path import dirname\n+sys.path.insert(1, f\"{dirname(__file__)}/../..\")\n+# ------------------\n+\n+import os\n+from slack import WebClient\n+import logging\n+logging.basicConfig(level=logging.DEBUG)\n+\n+from flask import Flask\n+\n+app = Flask(__name__)\n+\n+with app.app_context():\n+    from test_module.slack_app import slack_events_adapter\n+\n+    slack_bot_token = os.environ[\"SLACK_BOT_TOKEN\"]\n+    slack_client = WebClient(slack_bot_token)\n+\n+\n+    @slack_events_adapter.on(\"message\")\n+    def handle_message(event_data):\n+        message = event_data[\"event\"]\n+        if message.get(\"subtype\") is None and \"hi\" in message.get('text'):\n+            channel = message[\"channel\"]\n+            message = \"Hi <@%s>! :tada:\" % message[\"user\"]\n+            slack_client.chat_postMessage(channel=channel, text=message)\n+\n+\n+    @slack_events_adapter.on(\"error\")\n+    def error_handler(err):\n+        print(\"ERROR: \" + str(err))\n+\n+# (Terminal A)\n+# source env/bin/activate\n+# (env) $ export SLACK_BOT_TOKEN=xoxb-***\n+# (env) $ export SLACK_SIGNING_SECRET=**\n+# (env) $ cd example/current_app\n+# (env) $ FLASK_APP=main.py FLASK_ENV=development flask run --port 3000\n+\n+# (Terminal B)\n+# ngrok http 3000\n+\n+# in Slack\n+# /invite @{your app's bot user}\n+# post a message \"hi\" in the channel\ndiff --git a/slackeventsapi/server.py b/slackeventsapi/server.py\n--- a/slackeventsapi/server.py\n+++ b/slackeventsapi/server.py\n@@ -1,10 +1,13 @@\n-from flask import Flask, request, make_response, Blueprint\n+import hashlib\n+import hmac\n import json\n import platform\n import sys\n-import hmac\n-import hashlib\n from time import time\n+\n+from flask import Flask, request, make_response, Blueprint\n+from werkzeug.local import LocalProxy\n+\n from .version import __version__\n \n \n@@ -18,10 +21,10 @@ def __init__(self, signing_secret, endpoint, emitter, server):\n         # If a server is passed in, bind the event handler routes to it,\n         # otherwise create a new Flask instance.\n         if server:\n-            if isinstance(server, Flask) or isinstance(server, Blueprint):\n+            if isinstance(server, (Flask, Blueprint, LocalProxy)):\n                 self.bind_route(server)\n             else:\n-                raise TypeError(\"Server must be an instance of Flask or Blueprint\")\n+                raise TypeError(\"Server must be an instance of Flask, Blueprint, or LocalProxy\")\n         else:\n             Flask.__init__(self, __name__)\n             self.bind_route(self)\n",
-    "test_patch": "diff --git a/example/current_app/test_module/__init__.py b/example/current_app/test_module/__init__.py\nnew file mode 100644\ndiff --git a/example/current_app/test_module/slack_app.py b/example/current_app/test_module/slack_app.py\nnew file mode 100644\n--- /dev/null\n+++ b/example/current_app/test_module/slack_app.py\n@@ -0,0 +1,16 @@\n+# ------------------\n+# Only for running this script here\n+import logging\n+import sys\n+from os.path import dirname\n+\n+sys.path.insert(1, f\"{dirname(__file__)}/../../..\")\n+logging.basicConfig(level=logging.DEBUG)\n+# ------------------\n+\n+from flask import current_app as app\n+from slackeventsapi import SlackEventAdapter\n+import os\n+\n+slack_signing_secret = os.environ[\"SLACK_SIGNING_SECRET\"]\n+slack_events_adapter = SlackEventAdapter(slack_signing_secret, \"/slack/events\", app)\ndiff --git a/tests/test_server.py b/tests/test_server.py\n--- a/tests/test_server.py\n+++ b/tests/test_server.py\n@@ -18,7 +18,7 @@ def test_server_not_flask():\n     with pytest.raises(TypeError) as e:\n         invalid_flask = \"I am not a Flask\"\n         SlackEventAdapter(\"SIGNING_SECRET\", \"/slack/events\", invalid_flask)\n-    assert e.value.args[0] == 'Server must be an instance of Flask or Blueprint'\n+    assert e.value.args[0] == 'Server must be an instance of Flask, Blueprint, or LocalProxy'\n \n \n def test_blueprint_server():\n",
-    "problem_statement": "Passing Flask app proxy as server\nHi Guys,\r\n\r\nI have an app factory on my setup and the app object usually it is invoked as :\r\n`from flask import current_app as app`\r\n\r\nHowever, the slackeventsapi complains about the app object : \r\n`TypeError(\"Server must be an instance of Flask\")`\r\n\r\nI have fixed adding the following to server.py : \r\n\r\n`from werkzeug.local import LocalProxy # Importing the localproxy class`\r\n\r\nLine 25 \r\n  Changed from : \r\n   ` if isinstance(server, Flask):`\r\n  to :\r\n    `if isinstance(server, Flask) or isinstance(server, LocalProxy):`\r\n\r\nBasically, if a Flask app proxy is passed the api will carry on without complaining since it has the same methods as the Flask app object.\r\n\r\nI hope this help other people and it is considered as a solution if more information is needed I am help to provide. \r\n\r\nThanks for the good work with the API.\r\n\r\n\r\n\r\n### What type of issue is this? (place an `x` in one of the `[ ]`)\r\n- [X] bug ?\r\n- [X] enhancement (feature request)\r\n- [ ] question\r\n- [ ] documentation related\r\n- [ ] testing related\r\n- [ ] discussion\r\n\r\n### Requirements\r\n* [X] I've read and understood the [Contributing guidelines](https://github.com/slackapi/python-slack-events-api/blob/master/.github/contributing.md) and have done my best effort to follow them.\r\n* [X] I've read and agree to the [Code of Conduct](https://slackhq.github.io/code-of-conduct).\r\n* [X] I've searched for any related issues and avoided creating a duplicate issue.\r\n\r\n#### Reproducible in:\r\nslackeventsapi version: slackeventsapi==2.1.0\r\npython version: Python 3.7.3\r\nOS version(s): \r\n\r\n\r\n\nPassing Flask app proxy as server\nHi Guys,\r\n\r\nI have an app factory on my setup and the app object usually it is invoked as :\r\n`from flask import current_app as app`\r\n\r\nHowever, the slackeventsapi complains about the app object : \r\n`TypeError(\"Server must be an instance of Flask\")`\r\n\r\nI have fixed adding the following to server.py : \r\n\r\n`from werkzeug.local import LocalProxy # Importing the localproxy class`\r\n\r\nLine 25 \r\n  Changed from : \r\n   ` if isinstance(server, Flask):`\r\n  to :\r\n    `if isinstance(server, Flask) or isinstance(server, LocalProxy):`\r\n\r\nBasically, if a Flask app proxy is passed the api will carry on without complaining since it has the same methods as the Flask app object.\r\n\r\nI hope this help other people and it is considered as a solution if more information is needed I am help to provide. \r\n\r\nThanks for the good work with the API.\r\n\r\n\r\n\r\n### What type of issue is this? (place an `x` in one of the `[ ]`)\r\n- [X] bug ?\r\n- [X] enhancement (feature request)\r\n- [ ] question\r\n- [ ] documentation related\r\n- [ ] testing related\r\n- [ ] discussion\r\n\r\n### Requirements\r\n* [X] I've read and understood the [Contributing guidelines](https://github.com/slackapi/python-slack-events-api/blob/master/.github/contributing.md) and have done my best effort to follow them.\r\n* [X] I've read and agree to the [Code of Conduct](https://slackhq.github.io/code-of-conduct).\r\n* [X] I've searched for any related issues and avoided creating a duplicate issue.\r\n\r\n#### Reproducible in:\r\nslackeventsapi version: slackeventsapi==2.1.0\r\npython version: Python 3.7.3\r\nOS version(s): \r\n\r\n\r\n\n",
-    "hints_text": "\n",
-    "created_at": "2020-06-12T06:58:10Z",
-    "version": "1.0"
-  },
-  {
-    "repo": "celery/celery",
-    "pull_number": 2598,
-    "instance_id": "celery__celery-2598",
-    "issue_numbers": [
-      "2518"
-    ],
-    "base_commit": "6592ff64b6b024a4b68abcc53b151888fdf0dee3",
-    "patch": "diff --git a/celery/backends/amqp.py b/celery/backends/amqp.py\n--- a/celery/backends/amqp.py\n+++ b/celery/backends/amqp.py\n@@ -195,7 +195,7 @@ def drain_events(self, connection, consumer,\n \n         def callback(meta, message):\n             if meta['status'] in states.READY_STATES:\n-                results[meta['task_id']] = meta\n+                results[meta['task_id']] = self.meta_from_decoded(meta)\n \n         consumer.callbacks[:] = [callback]\n         time_start = now()\n",
-    "test_patch": "diff --git a/celery/tests/backends/test_amqp.py b/celery/tests/backends/test_amqp.py\n--- a/celery/tests/backends/test_amqp.py\n+++ b/celery/tests/backends/test_amqp.py\n@@ -13,6 +13,7 @@\n from celery.backends.amqp import AMQPBackend\n from celery.exceptions import TimeoutError\n from celery.five import Empty, Queue, range\n+from celery.result import AsyncResult\n from celery.utils import uuid\n \n from celery.tests.case import (\n@@ -246,10 +247,20 @@ def test_wait_for(self):\n         with self.assertRaises(TimeoutError):\n             b.wait_for(tid, timeout=0.01, cache=False)\n \n-    def test_drain_events_remaining_timeouts(self):\n+    def test_drain_events_decodes_exceptions_in_meta(self):\n+        tid = uuid()\n+        b = self.create_backend(serializer=\"json\")\n+        b.store_result(tid, RuntimeError(\"aap\"), states.FAILURE)\n+        result = AsyncResult(tid, backend=b)\n \n-        class Connection(object):\n+        with self.assertRaises(Exception) as cm:\n+            result.get()\n \n+        self.assertEqual(cm.exception.__class__.__name__, \"RuntimeError\")\n+        self.assertEqual(str(cm.exception), \"aap\")\n+\n+    def test_drain_events_remaining_timeouts(self):\n+        class Connection(object):\n             def drain_events(self, timeout=None):\n                 pass\n \n",
-    "problem_statement": "CELERY_RESULT_SERIALIZER = 'json' breaks Exception marshaling\nSetting `CELERY_RESULT_SERIALIZER = json` and raising an exception in the worker leads to this:\n\n```\n/path/to/lib/python2.7/site-packages/celery/result.py in get(self, timeout, propagate, interval, no_ack, follow_parents, EXCEPTION_STATES, PROPAGATE_STATES)\n    173             status = meta['status']\n    174             if status in PROPAGATE_STATES and propagate:\n--> 175                 raise meta['result']\n    176             return meta['result']\n    177     wait = get  # deprecated alias to :meth:`get`.\n\nTypeError: exceptions must be old-style classes or derived from BaseException, not dict\n```\n\nwhere the contents of `meta['result']` are (in my case):\n\n```\n{u'exc_message': u'unknown keys: nam', u'exc_type': u'ValueError'}\n```\n\nso it _looks_ like celery could convert the dict to a real exception before raising, but it does not currently.  Changing back to `pickle` works as expected.\n\nbug can be reproduced with the following:\n\n``` python\n# jsonresults.py\nfrom celery.app.base import Celery\n\nCELERY_RESULT_SERIALIZER = 'json'\nCELERY_RESULT_BACKEND = 'amqp'\n\napp = Celery(config_source=__name__)\n\n@app.task\ndef hello():\n    raise ValueError('go away')\n```\n\nworker:\n\n```\n# python -m celery --app=jsonresults:app worker\n```\n\ncaller:\n\n``` python\nimport jsonresults\njsonresults.hello.delay().get()\n```\n\n",
-    "hints_text": "This is biting me as well. Any news?\n",
-    "created_at": "2015-04-29T14:52:17Z",
-    "version": "1.0"
-  }
-]
\ No newline at end of file