Training in progress, step 2500, checkpoint

.gitattributes

@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+checkpoint-2500/tokenizer.json filter=lfs diff=lfs merge=lfs -text

checkpoint-2500/added_tokens.json

@@ -0,0 +1,24 @@
+{
+  "</tool_call>": 151658,
+  "<tool_call>": 151657,
+  "<|box_end|>": 151649,
+  "<|box_start|>": 151648,
+  "<|endoftext|>": 151643,
+  "<|file_sep|>": 151664,
+  "<|fim_middle|>": 151660,
+  "<|fim_pad|>": 151662,
+  "<|fim_prefix|>": 151659,
+  "<|fim_suffix|>": 151661,
+  "<|im_end|>": 151645,
+  "<|im_start|>": 151644,
+  "<|image_pad|>": 151655,
+  "<|object_ref_end|>": 151647,
+  "<|object_ref_start|>": 151646,
+  "<|quad_end|>": 151651,
+  "<|quad_start|>": 151650,
+  "<|repo_name|>": 151663,
+  "<|video_pad|>": 151656,
+  "<|vision_end|>": 151653,
+  "<|vision_pad|>": 151654,
+  "<|vision_start|>": 151652
+}

checkpoint-2500/config.json

@@ -0,0 +1,29 @@
+{
+  "_name_or_path": "Qwen/Qwen2.5-Math-1.5B-Instruct",
+  "architectures": [
+    "Qwen2ForCausalLM"
+  ],
+  "attention_dropout": 0.0,
+  "bos_token_id": 151643,
+  "eos_token_id": 151645,
+  "hidden_act": "silu",
+  "hidden_size": 1536,
+  "initializer_range": 0.02,
+  "intermediate_size": 8960,
+  "max_position_embeddings": 4096,
+  "max_window_layers": 21,
+  "model_type": "qwen2",
+  "num_attention_heads": 12,
+  "num_hidden_layers": 28,
+  "num_key_value_heads": 2,
+  "rms_norm_eps": 1e-06,
+  "rope_scaling": null,
+  "rope_theta": 10000.0,
+  "sliding_window": null,
+  "tie_word_embeddings": true,
+  "torch_dtype": "bfloat16",
+  "transformers_version": "4.46.3",
+  "use_cache": true,
+  "use_sliding_window": false,
+  "vocab_size": 151936
+}

checkpoint-2500/generation_config.json

@@ -0,0 +1,9 @@
+{
+  "bos_token_id": 151643,
+  "eos_token_id": [
+    151645,
+    151643
+  ],
+  "pad_token_id": 151643,
+  "transformers_version": "4.46.3"
+}

checkpoint-2500/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8ca8b87da7d8354af58bbf3797de51ae0e78a6ad1a612a9f4882d57400c3a4e7
+size 3087467144

checkpoint-2500/optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2c63b1540aa7c35b86dbec89d81d302012e3fdf514334565059fd56301c96757
+size 6057291

checkpoint-2500/rng_state.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:16f15a6ba1a4de4b5948e1c3c9c28c11e3fdc55e338eafa4a2edc35e2503e347
+size 14645

checkpoint-2500/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:eb41608c74f0601f51aba42ba070368d58f306fd9096a73c5fc7061ed921c210
+size 1465

checkpoint-2500/special_tokens_map.json

@@ -0,0 +1,31 @@
+{
+  "additional_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "eos_token": {
+    "content": "<|im_end|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

checkpoint-2500/tokenizer.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9c5ae00e602b8860cbd784ba82a8aa14e8feecec692e7076590d014d7b7fdafa
+size 11421896

checkpoint-2500/tokenizer_config.json

@@ -0,0 +1,208 @@
+{
+  "add_bos_token": false,
+  "add_prefix_space": false,
+  "added_tokens_decoder": {
+    "151643": {
+      "content": "<|endoftext|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151644": {
+      "content": "<|im_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151645": {
+      "content": "<|im_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151646": {
+      "content": "<|object_ref_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151647": {
+      "content": "<|object_ref_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151648": {
+      "content": "<|box_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151649": {
+      "content": "<|box_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151650": {
+      "content": "<|quad_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151651": {
+      "content": "<|quad_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151652": {
+      "content": "<|vision_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151653": {
+      "content": "<|vision_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151654": {
+      "content": "<|vision_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151655": {
+      "content": "<|image_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151656": {
+      "content": "<|video_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151657": {
+      "content": "<tool_call>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151658": {
+      "content": "</tool_call>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151659": {
+      "content": "<|fim_prefix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151660": {
+      "content": "<|fim_middle|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151661": {
+      "content": "<|fim_suffix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151662": {
+      "content": "<|fim_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151663": {
+      "content": "<|repo_name|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151664": {
+      "content": "<|file_sep|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    }
+  },
+  "additional_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "bos_token": null,
+  "chat_template": "{%- if tools %}\n    {{- '<|im_start|>system\\n' }}\n    {%- if messages[0]['role'] == 'system' %}\n        {{- messages[0]['content'] }}\n    {%- else %}\n        {{- 'Please reason step by step, and put your final answer within \\\\boxed{}.' }}\n    {%- endif %}\n    {{- \"\\n\\n# Tools\\n\\nYou may call one or more functions to assist with the user query.\\n\\nYou are provided with function signatures within <tools></tools> XML tags:\\n<tools>\" }}\n    {%- for tool in tools %}\n        {{- \"\\n\" }}\n        {{- tool | tojson }}\n    {%- endfor %}\n    {{- \"\\n</tools>\\n\\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\\n<tool_call>\\n{\\\"name\\\": <function-name>, \\\"arguments\\\": <args-json-object>}\\n</tool_call><|im_end|>\\n\" }}\n{%- else %}\n    {%- if messages[0]['role'] == 'system' %}\n        {{- '<|im_start|>system\\n' + messages[0]['content'] + '<|im_end|>\\n' }}\n    {%- else %}\n        {{- '<|im_start|>system\\nPlease reason step by step, and put your final answer within \\\\boxed{}.<|im_end|>\\n' }}\n    {%- endif %}\n{%- endif %}\n{%- for message in messages %}\n    {%- if (message.role == \"user\") or (message.role == \"system\" and not loop.first) or (message.role == \"assistant\" and not message.tool_calls) %}\n        {{- '<|im_start|>' + message.role + '\\n' + message.content + '<|im_end|>' + '\\n' }}\n    {%- elif message.role == \"assistant\" %}\n        {{- '<|im_start|>' + message.role }}\n        {%- if message.content %}\n            {{- '\\n' + message.content }}\n        {%- endif %}\n        {%- for tool_call in message.tool_calls %}\n            {%- if tool_call.function is defined %}\n                {%- set tool_call = tool_call.function %}\n            {%- endif %}\n            {{- '\\n<tool_call>\\n{\"name\": \"' }}\n            {{- tool_call.name }}\n            {{- '\", \"arguments\": ' }}\n            {{- tool_call.arguments | tojson }}\n            {{- '}\\n</tool_call>' }}\n        {%- endfor %}\n        {{- '<|im_end|>\\n' }}\n    {%- elif message.role == \"tool\" %}\n        {%- if (loop.index0 == 0) or (messages[loop.index0 - 1].role != \"tool\") %}\n            {{- '<|im_start|>user' }}\n        {%- endif %}\n        {{- '\\n<tool_response>\\n' }}\n        {{- message.content }}\n        {{- '\\n</tool_response>' }}\n        {%- if loop.last or (messages[loop.index0 + 1].role != \"tool\") %}\n            {{- '<|im_end|>\\n' }}\n        {%- endif %}\n    {%- endif %}\n{%- endfor %}\n{%- if add_generation_prompt %}\n    {{- '<|im_start|>assistant\\n' }}\n{%- endif %}\n",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|im_end|>",
+  "errors": "replace",
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "padding_side": "left",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}

checkpoint-2500/trainer_state.json

@@ -0,0 +1,1783 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.0,
+  "eval_steps": 500,
+  "global_step": 2500,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.004,
+      "grad_norm": 42.0,
+      "learning_rate": 8.000000000000001e-07,
+      "loss": 16.1356,
+      "step": 10
+    },
+    {
+      "epoch": 0.008,
+      "grad_norm": 39.25,
+      "learning_rate": 1.6000000000000001e-06,
+      "loss": 16.1383,
+      "step": 20
+    },
+    {
+      "epoch": 0.012,
+      "grad_norm": 40.25,
+      "learning_rate": 2.4000000000000003e-06,
+      "loss": 15.4331,
+      "step": 30
+    },
+    {
+      "epoch": 0.016,
+      "grad_norm": 34.75,
+      "learning_rate": 3.2000000000000003e-06,
+      "loss": 16.3866,
+      "step": 40
+    },
+    {
+      "epoch": 0.02,
+      "grad_norm": 33.25,
+      "learning_rate": 4.000000000000001e-06,
+      "loss": 16.0455,
+      "step": 50
+    },
+    {
+      "epoch": 0.024,
+      "grad_norm": 26.875,
+      "learning_rate": 4.800000000000001e-06,
+      "loss": 16.2563,
+      "step": 60
+    },
+    {
+      "epoch": 0.028,
+      "grad_norm": 24.25,
+      "learning_rate": 5.600000000000001e-06,
+      "loss": 16.1983,
+      "step": 70
+    },
+    {
+      "epoch": 0.032,
+      "grad_norm": 20.625,
+      "learning_rate": 6.4000000000000006e-06,
+      "loss": 15.3843,
+      "step": 80
+    },
+    {
+      "epoch": 0.036,
+      "grad_norm": 23.25,
+      "learning_rate": 7.2000000000000005e-06,
+      "loss": 15.5479,
+      "step": 90
+    },
+    {
+      "epoch": 0.04,
+      "grad_norm": 20.75,
+      "learning_rate": 8.000000000000001e-06,
+      "loss": 15.0409,
+      "step": 100
+    },
+    {
+      "epoch": 0.044,
+      "grad_norm": 25.5,
+      "learning_rate": 8.8e-06,
+      "loss": 15.0215,
+      "step": 110
+    },
+    {
+      "epoch": 0.048,
+      "grad_norm": 24.625,
+      "learning_rate": 9.600000000000001e-06,
+      "loss": 14.8879,
+      "step": 120
+    },
+    {
+      "epoch": 0.052,
+      "grad_norm": 26.625,
+      "learning_rate": 1.04e-05,
+      "loss": 14.597,
+      "step": 130
+    },
+    {
+      "epoch": 0.056,
+      "grad_norm": 21.25,
+      "learning_rate": 1.1200000000000001e-05,
+      "loss": 15.0546,
+      "step": 140
+    },
+    {
+      "epoch": 0.06,
+      "grad_norm": 25.875,
+      "learning_rate": 1.2e-05,
+      "loss": 14.8386,
+      "step": 150
+    },
+    {
+      "epoch": 0.064,
+      "grad_norm": 25.875,
+      "learning_rate": 1.2800000000000001e-05,
+      "loss": 14.5003,
+      "step": 160
+    },
+    {
+      "epoch": 0.068,
+      "grad_norm": 31.375,
+      "learning_rate": 1.3600000000000002e-05,
+      "loss": 15.1483,
+      "step": 170
+    },
+    {
+      "epoch": 0.072,
+      "grad_norm": 25.25,
+      "learning_rate": 1.4400000000000001e-05,
+      "loss": 14.6315,
+      "step": 180
+    },
+    {
+      "epoch": 0.076,
+      "grad_norm": 35.25,
+      "learning_rate": 1.5200000000000002e-05,
+      "loss": 13.9271,
+      "step": 190
+    },
+    {
+      "epoch": 0.08,
+      "grad_norm": 28.5,
+      "learning_rate": 1.6000000000000003e-05,
+      "loss": 14.139,
+      "step": 200
+    },
+    {
+      "epoch": 0.084,
+      "grad_norm": 22.25,
+      "learning_rate": 1.6800000000000002e-05,
+      "loss": 14.9061,
+      "step": 210
+    },
+    {
+      "epoch": 0.088,
+      "grad_norm": 37.0,
+      "learning_rate": 1.76e-05,
+      "loss": 14.4556,
+      "step": 220
+    },
+    {
+      "epoch": 0.092,
+      "grad_norm": 32.25,
+      "learning_rate": 1.8400000000000003e-05,
+      "loss": 14.2981,
+      "step": 230
+    },
+    {
+      "epoch": 0.096,
+      "grad_norm": 41.25,
+      "learning_rate": 1.9200000000000003e-05,
+      "loss": 14.9684,
+      "step": 240
+    },
+    {
+      "epoch": 0.1,
+      "grad_norm": 33.75,
+      "learning_rate": 2e-05,
+      "loss": 14.6508,
+      "step": 250
+    },
+    {
+      "epoch": 0.104,
+      "grad_norm": 33.25,
+      "learning_rate": 1.9999025240093045e-05,
+      "loss": 14.3454,
+      "step": 260
+    },
+    {
+      "epoch": 0.108,
+      "grad_norm": 28.375,
+      "learning_rate": 1.9996101150403543e-05,
+      "loss": 14.1579,
+      "step": 270
+    },
+    {
+      "epoch": 0.112,
+      "grad_norm": 31.875,
+      "learning_rate": 1.9991228300988586e-05,
+      "loss": 14.1456,
+      "step": 280
+    },
+    {
+      "epoch": 0.116,
+      "grad_norm": 33.0,
+      "learning_rate": 1.9984407641819812e-05,
+      "loss": 14.0944,
+      "step": 290
+    },
+    {
+      "epoch": 0.12,
+      "grad_norm": 31.0,
+      "learning_rate": 1.9975640502598243e-05,
+      "loss": 14.7894,
+      "step": 300
+    },
+    {
+      "epoch": 0.124,
+      "grad_norm": 26.75,
+      "learning_rate": 1.9964928592495046e-05,
+      "loss": 14.2806,
+      "step": 310
+    },
+    {
+      "epoch": 0.128,
+      "grad_norm": 36.0,
+      "learning_rate": 1.9952273999818312e-05,
+      "loss": 14.3856,
+      "step": 320
+    },
+    {
+      "epoch": 0.132,
+      "grad_norm": 24.5,
+      "learning_rate": 1.9937679191605964e-05,
+      "loss": 14.24,
+      "step": 330
+    },
+    {
+      "epoch": 0.136,
+      "grad_norm": 30.625,
+      "learning_rate": 1.9921147013144782e-05,
+      "loss": 13.9409,
+      "step": 340
+    },
+    {
+      "epoch": 0.14,
+      "grad_norm": 28.0,
+      "learning_rate": 1.9902680687415704e-05,
+      "loss": 13.7884,
+      "step": 350
+    },
+    {
+      "epoch": 0.144,
+      "grad_norm": 30.0,
+      "learning_rate": 1.988228381446553e-05,
+      "loss": 13.975,
+      "step": 360
+    },
+    {
+      "epoch": 0.148,
+      "grad_norm": 50.0,
+      "learning_rate": 1.985996037070505e-05,
+      "loss": 13.902,
+      "step": 370
+    },
+    {
+      "epoch": 0.152,
+      "grad_norm": 33.25,
+      "learning_rate": 1.983571470813386e-05,
+      "loss": 14.394,
+      "step": 380
+    },
+    {
+      "epoch": 0.156,
+      "grad_norm": 30.25,
+      "learning_rate": 1.9809551553491918e-05,
+      "loss": 14.2122,
+      "step": 390
+    },
+    {
+      "epoch": 0.16,
+      "grad_norm": 33.0,
+      "learning_rate": 1.9781476007338058e-05,
+      "loss": 14.048,
+      "step": 400
+    },
+    {
+      "epoch": 0.164,
+      "grad_norm": 27.75,
+      "learning_rate": 1.9751493543055634e-05,
+      "loss": 14.3168,
+      "step": 410
+    },
+    {
+      "epoch": 0.168,
+      "grad_norm": 25.0,
+      "learning_rate": 1.9719610005785466e-05,
+      "loss": 14.0735,
+      "step": 420
+    },
+    {
+      "epoch": 0.172,
+      "grad_norm": 25.5,
+      "learning_rate": 1.9685831611286312e-05,
+      "loss": 14.1113,
+      "step": 430
+    },
+    {
+      "epoch": 0.176,
+      "grad_norm": 28.375,
+      "learning_rate": 1.9650164944723116e-05,
+      "loss": 14.3061,
+      "step": 440
+    },
+    {
+      "epoch": 0.18,
+      "grad_norm": 32.75,
+      "learning_rate": 1.961261695938319e-05,
+      "loss": 13.619,
+      "step": 450
+    },
+    {
+      "epoch": 0.184,
+      "grad_norm": 31.25,
+      "learning_rate": 1.9573194975320672e-05,
+      "loss": 14.4487,
+      "step": 460
+    },
+    {
+      "epoch": 0.188,
+      "grad_norm": 29.0,
+      "learning_rate": 1.9531906677929472e-05,
+      "loss": 13.936,
+      "step": 470
+    },
+    {
+      "epoch": 0.192,
+      "grad_norm": 24.0,
+      "learning_rate": 1.9488760116444966e-05,
+      "loss": 14.0659,
+      "step": 480
+    },
+    {
+      "epoch": 0.196,
+      "grad_norm": 26.625,
+      "learning_rate": 1.944376370237481e-05,
+      "loss": 13.7547,
+      "step": 490
+    },
+    {
+      "epoch": 0.2,
+      "grad_norm": 28.0,
+      "learning_rate": 1.9396926207859085e-05,
+      "loss": 13.8495,
+      "step": 500
+    },
+    {
+      "epoch": 0.204,
+      "grad_norm": 26.75,
+      "learning_rate": 1.9348256763960146e-05,
+      "loss": 13.625,
+      "step": 510
+    },
+    {
+      "epoch": 0.208,
+      "grad_norm": 24.875,
+      "learning_rate": 1.9297764858882516e-05,
+      "loss": 13.6376,
+      "step": 520
+    },
+    {
+      "epoch": 0.212,
+      "grad_norm": 25.625,
+      "learning_rate": 1.9245460336123136e-05,
+      "loss": 13.6954,
+      "step": 530
+    },
+    {
+      "epoch": 0.216,
+      "grad_norm": 26.875,
+      "learning_rate": 1.9191353392552346e-05,
+      "loss": 13.8046,
+      "step": 540
+    },
+    {
+      "epoch": 0.22,
+      "grad_norm": 22.5,
+      "learning_rate": 1.913545457642601e-05,
+      "loss": 13.4518,
+      "step": 550
+    },
+    {
+      "epoch": 0.224,
+      "grad_norm": 26.125,
+      "learning_rate": 1.907777478532909e-05,
+      "loss": 13.6853,
+      "step": 560
+    },
+    {
+      "epoch": 0.228,
+      "grad_norm": 28.5,
+      "learning_rate": 1.901832526405114e-05,
+      "loss": 13.5204,
+      "step": 570
+    },
+    {
+      "epoch": 0.232,
+      "grad_norm": 24.375,
+      "learning_rate": 1.895711760239413e-05,
+      "loss": 13.5718,
+      "step": 580
+    },
+    {
+      "epoch": 0.236,
+      "grad_norm": 28.375,
+      "learning_rate": 1.889416373291298e-05,
+      "loss": 13.2873,
+      "step": 590
+    },
+    {
+      "epoch": 0.24,
+      "grad_norm": 26.75,
+      "learning_rate": 1.8829475928589272e-05,
+      "loss": 13.448,
+      "step": 600
+    },
+    {
+      "epoch": 0.244,
+      "grad_norm": 21.375,
+      "learning_rate": 1.8763066800438638e-05,
+      "loss": 12.6493,
+      "step": 610
+    },
+    {
+      "epoch": 0.248,
+      "grad_norm": 26.125,
+      "learning_rate": 1.869494929505219e-05,
+      "loss": 12.9608,
+      "step": 620
+    },
+    {
+      "epoch": 0.252,
+      "grad_norm": 28.5,
+      "learning_rate": 1.8625136692072577e-05,
+      "loss": 12.5389,
+      "step": 630
+    },
+    {
+      "epoch": 0.256,
+      "grad_norm": 25.0,
+      "learning_rate": 1.855364260160507e-05,
+      "loss": 13.3493,
+      "step": 640
+    },
+    {
+      "epoch": 0.26,
+      "grad_norm": 24.25,
+      "learning_rate": 1.848048096156426e-05,
+      "loss": 13.155,
+      "step": 650
+    },
+    {
+      "epoch": 0.264,
+      "grad_norm": 34.5,
+      "learning_rate": 1.8405666034956842e-05,
+      "loss": 13.3605,
+      "step": 660
+    },
+    {
+      "epoch": 0.268,
+      "grad_norm": 25.0,
+      "learning_rate": 1.8329212407100996e-05,
+      "loss": 13.2862,
+      "step": 670
+    },
+    {
+      "epoch": 0.272,
+      "grad_norm": 24.125,
+      "learning_rate": 1.8251134982782952e-05,
+      "loss": 13.1185,
+      "step": 680
+    },
+    {
+      "epoch": 0.276,
+      "grad_norm": 22.625,
+      "learning_rate": 1.8171448983351284e-05,
+      "loss": 12.7952,
+      "step": 690
+    },
+    {
+      "epoch": 0.28,
+      "grad_norm": 35.0,
+      "learning_rate": 1.8090169943749477e-05,
+      "loss": 13.311,
+      "step": 700
+    },
+    {
+      "epoch": 0.284,
+      "grad_norm": 34.0,
+      "learning_rate": 1.8007313709487334e-05,
+      "loss": 13.0198,
+      "step": 710
+    },
+    {
+      "epoch": 0.288,
+      "grad_norm": 28.25,
+      "learning_rate": 1.792289643355191e-05,
+      "loss": 12.7704,
+      "step": 720
+    },
+    {
+      "epoch": 0.292,
+      "grad_norm": 23.125,
+      "learning_rate": 1.78369345732584e-05,
+      "loss": 12.857,
+      "step": 730
+    },
+    {
+      "epoch": 0.296,
+      "grad_norm": 23.5,
+      "learning_rate": 1.7749444887041797e-05,
+      "loss": 13.166,
+      "step": 740
+    },
+    {
+      "epoch": 0.3,
+      "grad_norm": 22.875,
+      "learning_rate": 1.766044443118978e-05,
+      "loss": 12.8836,
+      "step": 750
+    },
+    {
+      "epoch": 0.304,
+      "grad_norm": 28.75,
+      "learning_rate": 1.7569950556517566e-05,
+      "loss": 12.9149,
+      "step": 760
+    },
+    {
+      "epoch": 0.308,
+      "grad_norm": 28.0,
+      "learning_rate": 1.747798090498532e-05,
+      "loss": 12.6713,
+      "step": 770
+    },
+    {
+      "epoch": 0.312,
+      "grad_norm": 23.625,
+      "learning_rate": 1.7384553406258842e-05,
+      "loss": 12.755,
+      "step": 780
+    },
+    {
+      "epoch": 0.316,
+      "grad_norm": 23.375,
+      "learning_rate": 1.7289686274214116e-05,
+      "loss": 12.7228,
+      "step": 790
+    },
+    {
+      "epoch": 0.32,
+      "grad_norm": 23.625,
+      "learning_rate": 1.7193398003386514e-05,
+      "loss": 12.8029,
+      "step": 800
+    },
+    {
+      "epoch": 0.324,
+      "grad_norm": 29.75,
+      "learning_rate": 1.709570736536521e-05,
+      "loss": 13.0532,
+      "step": 810
+    },
+    {
+      "epoch": 0.328,
+      "grad_norm": 35.0,
+      "learning_rate": 1.6996633405133656e-05,
+      "loss": 11.925,
+      "step": 820
+    },
+    {
+      "epoch": 0.332,
+      "grad_norm": 29.25,
+      "learning_rate": 1.68961954373567e-05,
+      "loss": 13.2527,
+      "step": 830
+    },
+    {
+      "epoch": 0.336,
+      "grad_norm": 27.625,
+      "learning_rate": 1.6794413042615168e-05,
+      "loss": 12.9308,
+      "step": 840
+    },
+    {
+      "epoch": 0.34,
+      "grad_norm": 26.25,
+      "learning_rate": 1.6691306063588583e-05,
+      "loss": 12.9368,
+      "step": 850
+    },
+    {
+      "epoch": 0.344,
+      "grad_norm": 21.125,
+      "learning_rate": 1.6586894601186804e-05,
+      "loss": 12.3981,
+      "step": 860
+    },
+    {
+      "epoch": 0.348,
+      "grad_norm": 21.125,
+      "learning_rate": 1.6481199010631312e-05,
+      "loss": 12.7891,
+      "step": 870
+    },
+    {
+      "epoch": 0.352,
+      "grad_norm": 19.75,
+      "learning_rate": 1.63742398974869e-05,
+      "loss": 12.5167,
+      "step": 880
+    },
+    {
+      "epoch": 0.356,
+      "grad_norm": 26.875,
+      "learning_rate": 1.6266038113644605e-05,
+      "loss": 12.6286,
+      "step": 890
+    },
+    {
+      "epoch": 0.36,
+      "grad_norm": 19.875,
+      "learning_rate": 1.6156614753256583e-05,
+      "loss": 12.5022,
+      "step": 900
+    },
+    {
+      "epoch": 0.364,
+      "grad_norm": 22.625,
+      "learning_rate": 1.6045991148623752e-05,
+      "loss": 12.3728,
+      "step": 910
+    },
+    {
+      "epoch": 0.368,
+      "grad_norm": 29.0,
+      "learning_rate": 1.5934188866037017e-05,
+      "loss": 12.5975,
+      "step": 920
+    },
+    {
+      "epoch": 0.372,
+      "grad_norm": 24.0,
+      "learning_rate": 1.5821229701572897e-05,
+      "loss": 12.543,
+      "step": 930
+    },
+    {
+      "epoch": 0.376,
+      "grad_norm": 22.875,
+      "learning_rate": 1.570713567684432e-05,
+      "loss": 12.4243,
+      "step": 940
+    },
+    {
+      "epoch": 0.38,
+      "grad_norm": 24.0,
+      "learning_rate": 1.5591929034707468e-05,
+      "loss": 12.5414,
+      "step": 950
+    },
+    {
+      "epoch": 0.384,
+      "grad_norm": 22.125,
+      "learning_rate": 1.5475632234925505e-05,
+      "loss": 12.7494,
+      "step": 960
+    },
+    {
+      "epoch": 0.388,
+      "grad_norm": 22.375,
+      "learning_rate": 1.5358267949789968e-05,
+      "loss": 12.5341,
+      "step": 970
+    },
+    {
+      "epoch": 0.392,
+      "grad_norm": 19.875,
+      "learning_rate": 1.5239859059700794e-05,
+      "loss": 12.2114,
+      "step": 980
+    },
+    {
+      "epoch": 0.396,
+      "grad_norm": 24.125,
+      "learning_rate": 1.5120428648705716e-05,
+      "loss": 12.5967,
+      "step": 990
+    },
+    {
+      "epoch": 0.4,
+      "grad_norm": 27.375,
+      "learning_rate": 1.5000000000000002e-05,
+      "loss": 12.4666,
+      "step": 1000
+    },
+    {
+      "epoch": 0.404,
+      "grad_norm": 19.875,
+      "learning_rate": 1.4878596591387329e-05,
+      "loss": 12.467,
+      "step": 1010
+    },
+    {
+      "epoch": 0.408,
+      "grad_norm": 20.625,
+      "learning_rate": 1.4756242090702756e-05,
+      "loss": 12.5511,
+      "step": 1020
+    },
+    {
+      "epoch": 0.412,
+      "grad_norm": 20.5,
+      "learning_rate": 1.463296035119862e-05,
+      "loss": 12.6373,
+      "step": 1030
+    },
+    {
+      "epoch": 0.416,
+      "grad_norm": 17.25,
+      "learning_rate": 1.4508775406894308e-05,
+      "loss": 12.2578,
+      "step": 1040
+    },
+    {
+      "epoch": 0.42,
+      "grad_norm": 18.875,
+      "learning_rate": 1.4383711467890776e-05,
+      "loss": 12.1735,
+      "step": 1050
+    },
+    {
+      "epoch": 0.424,
+      "grad_norm": 44.5,
+      "learning_rate": 1.4257792915650728e-05,
+      "loss": 12.3356,
+      "step": 1060
+    },
+    {
+      "epoch": 0.428,
+      "grad_norm": 20.875,
+      "learning_rate": 1.413104429824542e-05,
+      "loss": 12.2474,
+      "step": 1070
+    },
+    {
+      "epoch": 0.432,
+      "grad_norm": 35.5,
+      "learning_rate": 1.4003490325568953e-05,
+      "loss": 12.1413,
+      "step": 1080
+    },
+    {
+      "epoch": 0.436,
+      "grad_norm": 17.125,
+      "learning_rate": 1.3875155864521031e-05,
+      "loss": 12.3009,
+      "step": 1090
+    },
+    {
+      "epoch": 0.44,
+      "grad_norm": 16.125,
+      "learning_rate": 1.3746065934159123e-05,
+      "loss": 12.5468,
+      "step": 1100
+    },
+    {
+      "epoch": 0.444,
+      "grad_norm": 17.125,
+      "learning_rate": 1.3616245700820922e-05,
+      "loss": 12.4364,
+      "step": 1110
+    },
+    {
+      "epoch": 0.448,
+      "grad_norm": 18.125,
+      "learning_rate": 1.3485720473218153e-05,
+      "loss": 12.5854,
+      "step": 1120
+    },
+    {
+      "epoch": 0.452,
+      "grad_norm": 18.75,
+      "learning_rate": 1.3354515697502552e-05,
+      "loss": 12.1944,
+      "step": 1130
+    },
+    {
+      "epoch": 0.456,
+      "grad_norm": 28.5,
+      "learning_rate": 1.3222656952305113e-05,
+      "loss": 11.8036,
+      "step": 1140
+    },
+    {
+      "epoch": 0.46,
+      "grad_norm": 19.25,
+      "learning_rate": 1.3090169943749475e-05,
+      "loss": 12.5166,
+      "step": 1150
+    },
+    {
+      "epoch": 0.464,
+      "grad_norm": 18.125,
+      "learning_rate": 1.2957080500440469e-05,
+      "loss": 12.3671,
+      "step": 1160
+    },
+    {
+      "epoch": 0.468,
+      "grad_norm": 20.25,
+      "learning_rate": 1.2823414568428767e-05,
+      "loss": 12.1979,
+      "step": 1170
+    },
+    {
+      "epoch": 0.472,
+      "grad_norm": 19.25,
+      "learning_rate": 1.2689198206152657e-05,
+      "loss": 11.9643,
+      "step": 1180
+    },
+    {
+      "epoch": 0.476,
+      "grad_norm": 20.125,
+      "learning_rate": 1.2554457579357906e-05,
+      "loss": 12.1654,
+      "step": 1190
+    },
+    {
+      "epoch": 0.48,
+      "grad_norm": 23.375,
+      "learning_rate": 1.2419218955996677e-05,
+      "loss": 12.1331,
+      "step": 1200
+    },
+    {
+      "epoch": 0.484,
+      "grad_norm": 22.0,
+      "learning_rate": 1.2283508701106559e-05,
+      "loss": 11.7672,
+      "step": 1210
+    },
+    {
+      "epoch": 0.488,
+      "grad_norm": 20.125,
+      "learning_rate": 1.2147353271670634e-05,
+      "loss": 12.0773,
+      "step": 1220
+    },
+    {
+      "epoch": 0.492,
+      "grad_norm": 34.75,
+      "learning_rate": 1.2010779211459649e-05,
+      "loss": 12.4496,
+      "step": 1230
+    },
+    {
+      "epoch": 0.496,
+      "grad_norm": 18.0,
+      "learning_rate": 1.187381314585725e-05,
+      "loss": 12.0919,
+      "step": 1240
+    },
+    {
+      "epoch": 0.5,
+      "grad_norm": 19.125,
+      "learning_rate": 1.1736481776669307e-05,
+      "loss": 12.2576,
+      "step": 1250
+    },
+    {
+      "epoch": 0.504,
+      "grad_norm": 17.25,
+      "learning_rate": 1.159881187691835e-05,
+      "loss": 12.5035,
+      "step": 1260
+    },
+    {
+      "epoch": 0.508,
+      "grad_norm": 17.875,
+      "learning_rate": 1.1460830285624119e-05,
+      "loss": 12.186,
+      "step": 1270
+    },
+    {
+      "epoch": 0.512,
+      "grad_norm": 20.0,
+      "learning_rate": 1.1322563902571227e-05,
+      "loss": 12.2628,
+      "step": 1280
+    },
+    {
+      "epoch": 0.516,
+      "grad_norm": 23.125,
+      "learning_rate": 1.1184039683065014e-05,
+      "loss": 12.1192,
+      "step": 1290
+    },
+    {
+      "epoch": 0.52,
+      "grad_norm": 19.5,
+      "learning_rate": 1.1045284632676535e-05,
+      "loss": 11.4056,
+      "step": 1300
+    },
+    {
+      "epoch": 0.524,
+      "grad_norm": 18.375,
+      "learning_rate": 1.0906325801977804e-05,
+      "loss": 11.909,
+      "step": 1310
+    },
+    {
+      "epoch": 0.528,
+      "grad_norm": 18.375,
+      "learning_rate": 1.0767190281268187e-05,
+      "loss": 11.9674,
+      "step": 1320
+    },
+    {
+      "epoch": 0.532,
+      "grad_norm": 22.0,
+      "learning_rate": 1.0627905195293135e-05,
+      "loss": 11.8993,
+      "step": 1330
+    },
+    {
+      "epoch": 0.536,
+      "grad_norm": 17.5,
+      "learning_rate": 1.0488497697956134e-05,
+      "loss": 12.0954,
+      "step": 1340
+    },
+    {
+      "epoch": 0.54,
+      "grad_norm": 17.125,
+      "learning_rate": 1.0348994967025012e-05,
+      "loss": 11.6242,
+      "step": 1350
+    },
+    {
+      "epoch": 0.544,
+      "grad_norm": 17.375,
+      "learning_rate": 1.0209424198833571e-05,
+      "loss": 12.0752,
+      "step": 1360
+    },
+    {
+      "epoch": 0.548,
+      "grad_norm": 20.375,
+      "learning_rate": 1.0069812602979617e-05,
+      "loss": 11.6246,
+      "step": 1370
+    },
+    {
+      "epoch": 0.552,
+      "grad_norm": 19.125,
+      "learning_rate": 9.930187397020385e-06,
+      "loss": 12.0839,
+      "step": 1380
+    },
+    {
+      "epoch": 0.556,
+      "grad_norm": 20.125,
+      "learning_rate": 9.790575801166432e-06,
+      "loss": 12.2101,
+      "step": 1390
+    },
+    {
+      "epoch": 0.56,
+      "grad_norm": 18.875,
+      "learning_rate": 9.651005032974994e-06,
+      "loss": 11.9135,
+      "step": 1400
+    },
+    {
+      "epoch": 0.564,
+      "grad_norm": 20.25,
+      "learning_rate": 9.511502302043867e-06,
+      "loss": 12.0097,
+      "step": 1410
+    },
+    {
+      "epoch": 0.568,
+      "grad_norm": 24.875,
+      "learning_rate": 9.372094804706867e-06,
+      "loss": 11.8345,
+      "step": 1420
+    },
+    {
+      "epoch": 0.572,
+      "grad_norm": 18.625,
+      "learning_rate": 9.232809718731815e-06,
+      "loss": 11.9064,
+      "step": 1430
+    },
+    {
+      "epoch": 0.576,
+      "grad_norm": 18.75,
+      "learning_rate": 9.093674198022201e-06,
+      "loss": 11.9602,
+      "step": 1440
+    },
+    {
+      "epoch": 0.58,
+      "grad_norm": 18.125,
+      "learning_rate": 8.954715367323468e-06,
+      "loss": 11.8622,
+      "step": 1450
+    },
+    {
+      "epoch": 0.584,
+      "grad_norm": 15.8125,
+      "learning_rate": 8.815960316934991e-06,
+      "loss": 11.7525,
+      "step": 1460
+    },
+    {
+      "epoch": 0.588,
+      "grad_norm": 19.625,
+      "learning_rate": 8.677436097428775e-06,
+      "loss": 11.6101,
+      "step": 1470
+    },
+    {
+      "epoch": 0.592,
+      "grad_norm": 16.125,
+      "learning_rate": 8.539169714375885e-06,
+      "loss": 11.9193,
+      "step": 1480
+    },
+    {
+      "epoch": 0.596,
+      "grad_norm": 17.375,
+      "learning_rate": 8.401188123081653e-06,
+      "loss": 11.8304,
+      "step": 1490
+    },
+    {
+      "epoch": 0.6,
+      "grad_norm": 24.375,
+      "learning_rate": 8.263518223330698e-06,
+      "loss": 11.6055,
+      "step": 1500
+    },
+    {
+      "epoch": 0.604,
+      "grad_norm": 18.375,
+      "learning_rate": 8.126186854142752e-06,
+      "loss": 11.714,
+      "step": 1510
+    },
+    {
+      "epoch": 0.608,
+      "grad_norm": 16.375,
+      "learning_rate": 7.989220788540356e-06,
+      "loss": 12.3469,
+      "step": 1520
+    },
+    {
+      "epoch": 0.612,
+      "grad_norm": 22.5,
+      "learning_rate": 7.852646728329368e-06,
+      "loss": 11.4557,
+      "step": 1530
+    },
+    {
+      "epoch": 0.616,
+      "grad_norm": 18.0,
+      "learning_rate": 7.716491298893443e-06,
+      "loss": 11.6784,
+      "step": 1540
+    },
+    {
+      "epoch": 0.62,
+      "grad_norm": 16.75,
+      "learning_rate": 7.580781044003324e-06,
+      "loss": 11.7863,
+      "step": 1550
+    },
+    {
+      "epoch": 0.624,
+      "grad_norm": 16.25,
+      "learning_rate": 7.445542420642097e-06,
+      "loss": 11.4727,
+      "step": 1560
+    },
+    {
+      "epoch": 0.628,
+      "grad_norm": 14.125,
+      "learning_rate": 7.310801793847344e-06,
+      "loss": 11.6379,
+      "step": 1570
+    },
+    {
+      "epoch": 0.632,
+      "grad_norm": 18.0,
+      "learning_rate": 7.176585431571235e-06,
+      "loss": 11.9072,
+      "step": 1580
+    },
+    {
+      "epoch": 0.636,
+      "grad_norm": 17.75,
+      "learning_rate": 7.042919499559538e-06,
+      "loss": 11.67,
+      "step": 1590
+    },
+    {
+      "epoch": 0.64,
+      "grad_norm": 15.5625,
+      "learning_rate": 6.909830056250527e-06,
+      "loss": 12.1847,
+      "step": 1600
+    },
+    {
+      "epoch": 0.644,
+      "grad_norm": 17.125,
+      "learning_rate": 6.777343047694891e-06,
+      "loss": 11.534,
+      "step": 1610
+    },
+    {
+      "epoch": 0.648,
+      "grad_norm": 13.75,
+      "learning_rate": 6.645484302497452e-06,
+      "loss": 11.8197,
+      "step": 1620
+    },
+    {
+      "epoch": 0.652,
+      "grad_norm": 14.75,
+      "learning_rate": 6.5142795267818505e-06,
+      "loss": 11.2645,
+      "step": 1630
+    },
+    {
+      "epoch": 0.656,
+      "grad_norm": 20.75,
+      "learning_rate": 6.383754299179079e-06,
+      "loss": 11.805,
+      "step": 1640
+    },
+    {
+      "epoch": 0.66,
+      "grad_norm": 14.5,
+      "learning_rate": 6.25393406584088e-06,
+      "loss": 11.5875,
+      "step": 1650
+    },
+    {
+      "epoch": 0.664,
+      "grad_norm": 15.3125,
+      "learning_rate": 6.124844135478971e-06,
+      "loss": 11.4167,
+      "step": 1660
+    },
+    {
+      "epoch": 0.668,
+      "grad_norm": 21.5,
+      "learning_rate": 5.996509674431053e-06,
+      "loss": 11.6036,
+      "step": 1670
+    },
+    {
+      "epoch": 0.672,
+      "grad_norm": 16.75,
+      "learning_rate": 5.868955701754584e-06,
+      "loss": 12.2395,
+      "step": 1680
+    },
+    {
+      "epoch": 0.676,
+      "grad_norm": 13.0,
+      "learning_rate": 5.742207084349274e-06,
+      "loss": 11.8518,
+      "step": 1690
+    },
+    {
+      "epoch": 0.68,
+      "grad_norm": 24.125,
+      "learning_rate": 5.616288532109225e-06,
+      "loss": 11.8566,
+      "step": 1700
+    },
+    {
+      "epoch": 0.684,
+      "grad_norm": 14.4375,
+      "learning_rate": 5.491224593105695e-06,
+      "loss": 11.4311,
+      "step": 1710
+    },
+    {
+      "epoch": 0.688,
+      "grad_norm": 13.875,
+      "learning_rate": 5.367039648801386e-06,
+      "loss": 11.3975,
+      "step": 1720
+    },
+    {
+      "epoch": 0.692,
+      "grad_norm": 18.5,
+      "learning_rate": 5.243757909297247e-06,
+      "loss": 11.5973,
+      "step": 1730
+    },
+    {
+      "epoch": 0.696,
+      "grad_norm": 18.25,
+      "learning_rate": 5.121403408612672e-06,
+      "loss": 11.7354,
+      "step": 1740
+    },
+    {
+      "epoch": 0.7,
+      "grad_norm": 15.375,
+      "learning_rate": 5.000000000000003e-06,
+      "loss": 11.8648,
+      "step": 1750
+    },
+    {
+      "epoch": 0.704,
+      "grad_norm": 17.75,
+      "learning_rate": 4.879571351294287e-06,
+      "loss": 11.788,
+      "step": 1760
+    },
+    {
+      "epoch": 0.708,
+      "grad_norm": 18.75,
+      "learning_rate": 4.76014094029921e-06,
+      "loss": 11.8052,
+      "step": 1770
+    },
+    {
+      "epoch": 0.712,
+      "grad_norm": 16.0,
+      "learning_rate": 4.641732050210032e-06,
+      "loss": 11.473,
+      "step": 1780
+    },
+    {
+      "epoch": 0.716,
+      "grad_norm": 14.1875,
+      "learning_rate": 4.524367765074499e-06,
+      "loss": 11.3811,
+      "step": 1790
+    },
+    {
+      "epoch": 0.72,
+      "grad_norm": 14.4375,
+      "learning_rate": 4.408070965292534e-06,
+      "loss": 11.2224,
+      "step": 1800
+    },
+    {
+      "epoch": 0.724,
+      "grad_norm": 19.125,
+      "learning_rate": 4.292864323155684e-06,
+      "loss": 11.3153,
+      "step": 1810
+    },
+    {
+      "epoch": 0.728,
+      "grad_norm": 17.0,
+      "learning_rate": 4.178770298427107e-06,
+      "loss": 11.5543,
+      "step": 1820
+    },
+    {
+      "epoch": 0.732,
+      "grad_norm": 16.125,
+      "learning_rate": 4.065811133962987e-06,
+      "loss": 11.6786,
+      "step": 1830
+    },
+    {
+      "epoch": 0.736,
+      "grad_norm": 17.75,
+      "learning_rate": 3.954008851376252e-06,
+      "loss": 11.5551,
+      "step": 1840
+    },
+    {
+      "epoch": 0.74,
+      "grad_norm": 15.75,
+      "learning_rate": 3.8433852467434175e-06,
+      "loss": 11.4994,
+      "step": 1850
+    },
+    {
+      "epoch": 0.744,
+      "grad_norm": 15.0,
+      "learning_rate": 3.7339618863553983e-06,
+      "loss": 11.7822,
+      "step": 1860
+    },
+    {
+      "epoch": 0.748,
+      "grad_norm": 16.25,
+      "learning_rate": 3.625760102513103e-06,
+      "loss": 11.7441,
+      "step": 1870
+    },
+    {
+      "epoch": 0.752,
+      "grad_norm": 16.125,
+      "learning_rate": 3.5188009893686916e-06,
+      "loss": 11.7047,
+      "step": 1880
+    },
+    {
+      "epoch": 0.756,
+      "grad_norm": 14.875,
+      "learning_rate": 3.4131053988131947e-06,
+      "loss": 11.6164,
+      "step": 1890
+    },
+    {
+      "epoch": 0.76,
+      "grad_norm": 13.875,
+      "learning_rate": 3.308693936411421e-06,
+      "loss": 11.8134,
+      "step": 1900
+    },
+    {
+      "epoch": 0.764,
+      "grad_norm": 16.125,
+      "learning_rate": 3.2055869573848374e-06,
+      "loss": 11.8393,
+      "step": 1910
+    },
+    {
+      "epoch": 0.768,
+      "grad_norm": 16.875,
+      "learning_rate": 3.103804562643302e-06,
+      "loss": 12.1629,
+      "step": 1920
+    },
+    {
+      "epoch": 0.772,
+      "grad_norm": 16.25,
+      "learning_rate": 3.003366594866345e-06,
+      "loss": 11.5428,
+      "step": 1930
+    },
+    {
+      "epoch": 0.776,
+      "grad_norm": 15.625,
+      "learning_rate": 2.9042926346347932e-06,
+      "loss": 11.3288,
+      "step": 1940
+    },
+    {
+      "epoch": 0.78,
+      "grad_norm": 14.625,
+      "learning_rate": 2.8066019966134907e-06,
+      "loss": 11.6534,
+      "step": 1950
+    },
+    {
+      "epoch": 0.784,
+      "grad_norm": 16.125,
+      "learning_rate": 2.7103137257858867e-06,
+      "loss": 11.4254,
+      "step": 1960
+    },
+    {
+      "epoch": 0.788,
+      "grad_norm": 17.25,
+      "learning_rate": 2.615446593741161e-06,
+      "loss": 11.8175,
+      "step": 1970
+    },
+    {
+      "epoch": 0.792,
+      "grad_norm": 13.8125,
+      "learning_rate": 2.522019095014683e-06,
+      "loss": 11.6752,
+      "step": 1980
+    },
+    {
+      "epoch": 0.796,
+      "grad_norm": 18.0,
+      "learning_rate": 2.4300494434824373e-06,
+      "loss": 11.692,
+      "step": 1990
+    },
+    {
+      "epoch": 0.8,
+      "grad_norm": 13.5,
+      "learning_rate": 2.339555568810221e-06,
+      "loss": 11.5219,
+      "step": 2000
+    },
+    {
+      "epoch": 0.804,
+      "grad_norm": 14.9375,
+      "learning_rate": 2.2505551129582047e-06,
+      "loss": 11.4298,
+      "step": 2010
+    },
+    {
+      "epoch": 0.808,
+      "grad_norm": 17.125,
+      "learning_rate": 2.163065426741603e-06,
+      "loss": 11.5041,
+      "step": 2020
+    },
+    {
+      "epoch": 0.812,
+      "grad_norm": 15.25,
+      "learning_rate": 2.0771035664480944e-06,
+      "loss": 11.5272,
+      "step": 2030
+    },
+    {
+      "epoch": 0.816,
+      "grad_norm": 16.75,
+      "learning_rate": 1.9926862905126663e-06,
+      "loss": 11.3027,
+      "step": 2040
+    },
+    {
+      "epoch": 0.82,
+      "grad_norm": 14.875,
+      "learning_rate": 1.9098300562505266e-06,
+      "loss": 11.6911,
+      "step": 2050
+    },
+    {
+      "epoch": 0.824,
+      "grad_norm": 13.375,
+      "learning_rate": 1.8285510166487154e-06,
+      "loss": 11.2328,
+      "step": 2060
+    },
+    {
+      "epoch": 0.828,
+      "grad_norm": 14.75,
+      "learning_rate": 1.7488650172170496e-06,
+      "loss": 11.7217,
+      "step": 2070
+    },
+    {
+      "epoch": 0.832,
+      "grad_norm": 14.0625,
+      "learning_rate": 1.6707875928990059e-06,
+      "loss": 11.7367,
+      "step": 2080
+    },
+    {
+      "epoch": 0.836,
+      "grad_norm": 13.5,
+      "learning_rate": 1.5943339650431578e-06,
+      "loss": 11.7986,
+      "step": 2090
+    },
+    {
+      "epoch": 0.84,
+      "grad_norm": 17.25,
+      "learning_rate": 1.5195190384357405e-06,
+      "loss": 11.6699,
+      "step": 2100
+    },
+    {
+      "epoch": 0.844,
+      "grad_norm": 16.5,
+      "learning_rate": 1.446357398394934e-06,
+      "loss": 11.8443,
+      "step": 2110
+    },
+    {
+      "epoch": 0.848,
+      "grad_norm": 14.0,
+      "learning_rate": 1.3748633079274254e-06,
+      "loss": 11.8157,
+      "step": 2120
+    },
+    {
+      "epoch": 0.852,
+      "grad_norm": 13.75,
+      "learning_rate": 1.30505070494781e-06,
+      "loss": 11.8902,
+      "step": 2130
+    },
+    {
+      "epoch": 0.856,
+      "grad_norm": 13.75,
+      "learning_rate": 1.2369331995613664e-06,
+      "loss": 12.1189,
+      "step": 2140
+    },
+    {
+      "epoch": 0.86,
+      "grad_norm": 14.1875,
+      "learning_rate": 1.1705240714107301e-06,
+      "loss": 11.692,
+      "step": 2150
+    },
+    {
+      "epoch": 0.864,
+      "grad_norm": 13.625,
+      "learning_rate": 1.1058362670870248e-06,
+      "loss": 11.5978,
+      "step": 2160
+    },
+    {
+      "epoch": 0.868,
+      "grad_norm": 13.875,
+      "learning_rate": 1.042882397605871e-06,
+      "loss": 11.6953,
+      "step": 2170
+    },
+    {
+      "epoch": 0.872,
+      "grad_norm": 13.4375,
+      "learning_rate": 9.816747359488632e-07,
+      "loss": 11.4016,
+      "step": 2180
+    },
+    {
+      "epoch": 0.876,
+      "grad_norm": 16.0,
+      "learning_rate": 9.222252146709143e-07,
+      "loss": 11.4873,
+      "step": 2190
+    },
+    {
+      "epoch": 0.88,
+      "grad_norm": 15.625,
+      "learning_rate": 8.645454235739903e-07,
+      "loss": 11.2464,
+      "step": 2200
+    },
+    {
+      "epoch": 0.884,
+      "grad_norm": 17.875,
+      "learning_rate": 8.086466074476562e-07,
+      "loss": 11.8022,
+      "step": 2210
+    },
+    {
+      "epoch": 0.888,
+      "grad_norm": 12.5,
+      "learning_rate": 7.545396638768698e-07,
+      "loss": 11.778,
+      "step": 2220
+    },
+    {
+      "epoch": 0.892,
+      "grad_norm": 16.0,
+      "learning_rate": 7.022351411174866e-07,
+      "loss": 11.5754,
+      "step": 2230
+    },
+    {
+      "epoch": 0.896,
+      "grad_norm": 15.0,
+      "learning_rate": 6.517432360398556e-07,
+      "loss": 11.9485,
+      "step": 2240
+    },
+    {
+      "epoch": 0.9,
+      "grad_norm": 16.125,
+      "learning_rate": 6.030737921409169e-07,
+      "loss": 11.8117,
+      "step": 2250
+    },
+    {
+      "epoch": 0.904,
+      "grad_norm": 17.0,
+      "learning_rate": 5.562362976251901e-07,
+      "loss": 11.3828,
+      "step": 2260
+    },
+    {
+      "epoch": 0.908,
+      "grad_norm": 17.25,
+      "learning_rate": 5.112398835550348e-07,
+      "loss": 11.5017,
+      "step": 2270
+    },
+    {
+      "epoch": 0.912,
+      "grad_norm": 13.625,
+      "learning_rate": 4.6809332207053083e-07,
+      "loss": 11.886,
+      "step": 2280
+    },
+    {
+      "epoch": 0.916,
+      "grad_norm": 16.625,
+      "learning_rate": 4.268050246793276e-07,
+      "loss": 11.5668,
+      "step": 2290
+    },
+    {
+      "epoch": 0.92,
+      "grad_norm": 16.875,
+      "learning_rate": 3.8738304061681107e-07,
+      "loss": 11.5818,
+      "step": 2300
+    },
+    {
+      "epoch": 0.924,
+      "grad_norm": 15.1875,
+      "learning_rate": 3.498350552768859e-07,
+      "loss": 11.2704,
+      "step": 2310
+    },
+    {
+      "epoch": 0.928,
+      "grad_norm": 15.875,
+      "learning_rate": 3.1416838871368925e-07,
+      "loss": 11.5949,
+      "step": 2320
+    },
+    {
+      "epoch": 0.932,
+      "grad_norm": 15.5625,
+      "learning_rate": 2.8038999421453827e-07,
+      "loss": 11.0745,
+      "step": 2330
+    },
+    {
+      "epoch": 0.936,
+      "grad_norm": 18.375,
+      "learning_rate": 2.4850645694436736e-07,
+      "loss": 11.762,
+      "step": 2340
+    },
+    {
+      "epoch": 0.94,
+      "grad_norm": 12.875,
+      "learning_rate": 2.1852399266194312e-07,
+      "loss": 11.7735,
+      "step": 2350
+    },
+    {
+      "epoch": 0.944,
+      "grad_norm": 17.75,
+      "learning_rate": 1.9044844650808468e-07,
+      "loss": 11.578,
+      "step": 2360
+    },
+    {
+      "epoch": 0.948,
+      "grad_norm": 17.375,
+      "learning_rate": 1.6428529186614195e-07,
+      "loss": 11.8724,
+      "step": 2370
+    },
+    {
+      "epoch": 0.952,
+      "grad_norm": 15.5,
+      "learning_rate": 1.400396292949513e-07,
+      "loss": 11.4199,
+      "step": 2380
+    },
+    {
+      "epoch": 0.956,
+      "grad_norm": 14.5,
+      "learning_rate": 1.1771618553447217e-07,
+      "loss": 11.3379,
+      "step": 2390
+    },
+    {
+      "epoch": 0.96,
+      "grad_norm": 14.5625,
+      "learning_rate": 9.731931258429638e-08,
+      "loss": 11.7067,
+      "step": 2400
+    },
+    {
+      "epoch": 0.964,
+      "grad_norm": 16.125,
+      "learning_rate": 7.885298685522235e-08,
+      "loss": 11.8713,
+      "step": 2410
+    },
+    {
+      "epoch": 0.968,
+      "grad_norm": 14.5,
+      "learning_rate": 6.232080839403631e-08,
+      "loss": 10.8336,
+      "step": 2420
+    },
+    {
+      "epoch": 0.972,
+      "grad_norm": 13.0,
+      "learning_rate": 4.772600018168816e-08,
+      "loss": 11.7092,
+      "step": 2430
+    },
+    {
+      "epoch": 0.976,
+      "grad_norm": 15.0625,
+      "learning_rate": 3.50714075049563e-08,
+      "loss": 11.4174,
+      "step": 2440
+    },
+    {
+      "epoch": 0.98,
+      "grad_norm": 15.0625,
+      "learning_rate": 2.4359497401758026e-08,
+      "loss": 11.9475,
+      "step": 2450
+    },
+    {
+      "epoch": 0.984,
+      "grad_norm": 14.6875,
+      "learning_rate": 1.5592358180189782e-08,
+      "loss": 11.7502,
+      "step": 2460
+    },
+    {
+      "epoch": 0.988,
+      "grad_norm": 13.75,
+      "learning_rate": 8.771699011416169e-09,
+      "loss": 11.3247,
+      "step": 2470
+    },
+    {
+      "epoch": 0.992,
+      "grad_norm": 14.0625,
+      "learning_rate": 3.898849596456477e-09,
+      "loss": 11.4935,
+      "step": 2480
+    },
+    {
+      "epoch": 0.996,
+      "grad_norm": 13.8125,
+      "learning_rate": 9.74759906957612e-10,
+      "loss": 11.274,
+      "step": 2490
+    },
+    {
+      "epoch": 1.0,
+      "grad_norm": 14.8125,
+      "learning_rate": 0.0,
+      "loss": 11.0717,
+      "step": 2500
+    }
+  ],
+  "logging_steps": 10,
+  "max_steps": 2500,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 9223372036854775807,
+  "save_steps": 2500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": true
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 6.4405861564416e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}

checkpoint-2500/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d3e6c1cd0a0dc623667f92357f5e99d5f1eee1a2ad00bef8d397716da4ab4b06
+size 5777