Update config.json

config.json

@@ -1,90 +1,51 @@
 {
-    "allow_embedding_resizing": true,
-    "_name_or_path": "ModernBERT-dec-32m",
-    "architectures": [
-      "ModernBertForCausalLM"
-    ],
-    "attention_layer": "rope",
-    "attention_probs_dropout_prob": 0.0,
-    "attn_out_bias": false,
-    "attn_out_dropout_prob": 0.1,
-    "attn_qkv_bias": false,
-    "bert_layer": "prenorm",
-    "classifier_dropout": null,
-    "compile_model": true,
-    "reference_compile": false,
-    "decoder_bias": true,
-    "deterministic_fa2": true,
-    "embed_dropout_prob": 0.0,
-    "embed_norm": true,
-    "embedding_layer": "sans_pos",
-    "encoder_layer": "base",
-    "final_norm": true,
-    "global_attn_every_n_layers": 3,
-    "gradient_checkpointing": false,
-    "head_class_act": "silu",
-    "head_class_bias": false,
-    "head_class_dropout": 0.0,
-    "head_class_norm": false,
-    "head_pred_act": "gelu",
-    "head_pred_bias": false,
-    "head_pred_dropout": 0.0,
-    "head_pred_norm": true,
-    "hidden_act": "gelu",
-    "hidden_dropout_prob": 0.1,
-    "hidden_size": 384,
-    "init_cutoff_factor": 2.0,
-    "init_method": "full_megatron",
-    "init_small_embedding": false,
-    "init_std": 0.02,
-    "initial_attention_layer": null,
-    "initial_bert_layer": null,
-    "initial_mlp_layer": null,
-    "initializer_range": 0.02,
-    "intermediate_size": 576,
-    "layer_norm_eps": 1e-12,
-    "local_attn_rotary_emb_base": -1,
-    "local_attn_rotary_emb_dim": null,
-    "loss_function": "fa_cross_entropy",
-    "deterministic_flash_attn": false,
-    "loss_kwargs": {
-        "reduction": "mean"
-    },
-    "masked_prediction": false,
-    "causal_mask": true,
-    "max_position_embeddings": 1024,
-    "mlp_dropout_prob": 0.0,
-    "mlp_in_bias": false,
-    "mlp_layer": "glu",
-    "mlp_out_bias": false,
-    "model_type": "modernbert",
-    "norm_kwargs": {
-        "bias": false,
-        "eps": 1e-05
-    },
-    "normalization": "layernorm",
-    "num_attention_heads": 6,
-    "num_hidden_layers": 10,
-    "num_initial_layers": 1,
-    "pad_logits": true,
-    "pad_token_id": null,
-    "padding": "unpadded",
-    "pooling_type": "cls",
-    "position_embedding_type": "absolute",
-    "global_rope_theta": 160000.0,
-    "local_rope_theta": 160000.0,
-    "rotary_emb_base": 160000.0,
-    "rotary_emb_dim": 64,
-    "rotary_emb_interleaved": false,
-    "rotary_emb_scale_base": null,
-    "skip_first_prenorm": true,
-    "sliding_window": 128,
-    "transformers_version": "4.44.1",
-    "type_vocab_size": 2,
-    "unpad_embeddings": true,
-    "use_cache": true,
-    "use_fa2": true,
-    "use_sdpa_attn_mask": false,
-    "vocab_size": 50368,
-    "is_causal": true
+  "_name_or_path": "ettin-decoder-32m",
+  "architectures": [
+    "ModernBERTDecoder"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "bos_token_id": 50281,
+  "classifier_activation": "gelu",
+  "classifier_bias": false,
+  "classifier_dropout": 0.0,
+  "classifier_pooling": "mean",
+  "cls_token_id": 50281,
+  "decoder_bias": true,
+  "deterministic_flash_attn": false,
+  "embedding_dropout": 0.0,
+  "eos_token_id": 50282,
+  "global_attn_every_n_layers": 3,
+  "global_rope_theta": 160000.0,
+  "gradient_checkpointing": false,
+  "hidden_activation": "gelu",
+  "hidden_size": 384,
+  "initializer_cutoff_factor": 2.0,
+  "initializer_range": 0.02,
+  "intermediate_size": 576,
+  "layer_norm_eps": 1e-05,
+  "local_attention": 128,
+  "local_rope_theta": 160000.0,
+  "max_position_embeddings": 7999,
+  "mlp_bias": false,
+  "mlp_dropout": 0.0,
+  "model_type": "modernbert-decoder",
+  "norm_bias": false,
+  "norm_eps": 1e-05,
+  "num_attention_heads": 6,
+  "num_hidden_layers": 10,
+  "pad_token_id": 50283,
+  "position_embedding_type": "sans_pos",
+  "sep_token_id": 50282,
+  "tie_word_embeddings": true,
+  "torch_dtype": "float32",
+  "transformers_version": "4.47.0.dev0",
+  "vocab_size": 50368,
+  "is_causal": true,
+  "masked_prediction": false,
+  "causal_mask": true,
+  "eos_token": "[SEP]",
+  "bos_token": "[CLS]",
+  "tokenizer_class": "PreTrainedTokenizerFast",
+  "unk_token": "[UNK]"
 }