config.json

{
  "activation_function": "gelu",
  "allow_embedding_resizing": true,
  "architectures": [
    "ModernBertForTokenClassification"
  ],
  "attention_bias": false,
  "attention_dropout": 0.0,
  "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",
  "bos_token_id": 50281,
  "classifier_activation": "gelu",
  "classifier_bias": false,
  "classifier_dropout": 0.0,
  "classifier_pooling": "cls",
  "cls_token_id": 50281,
  "compile_model": true,
  "decoder_bias": true,
  "defect_label_names": [
    "O",
    "BAD-script",
    "TAIL-script",
    "BAD-grammar",
    "TAIL-grammar"
  ],
  "defect_threshold": [
    0.5,
    0.9
  ],
  "deterministic_flash_attn": false,
  "dtype": "float32",
  "embed_dropout_prob": 0.0,
  "embed_norm": true,
  "embedding_dropout": 0.0,
  "embedding_layer": "sans_pos",
  "eos_token_id": 50282,
  "final_norm": true,
  "global_attn_every_n_layers": 3,
  "head_pred_act": "gelu",
  "hidden_act": "gelu",
  "hidden_activation": "gelu",
  "hidden_size": 768,
  "id2label": {
    "0": "O",
    "1": "BAD-script",
    "2": "TAIL-script",
    "3": "BAD-grammar",
    "4": "TAIL-grammar"
  },
  "init_method": "full_megatron",
  "initializer_cutoff_factor": 2.0,
  "initializer_range": 0.02,
  "intermediate_size": 1152,
  "label2id": {
    "BAD-grammar": 3,
    "BAD-script": 1,
    "O": 0,
    "TAIL-grammar": 4,
    "TAIL-script": 2
  },
  "layer_types": [
    "full_attention",
    "sliding_attention",
    "sliding_attention",
    "full_attention",
    "sliding_attention",
    "sliding_attention",
    "full_attention",
    "sliding_attention",
    "sliding_attention",
    "full_attention",
    "sliding_attention",
    "sliding_attention",
    "full_attention",
    "sliding_attention",
    "sliding_attention",
    "full_attention",
    "sliding_attention",
    "sliding_attention",
    "full_attention",
    "sliding_attention",
    "sliding_attention",
    "full_attention"
  ],
  "local_attention": 256,
  "local_attn_rotary_emb_base": 10000.0,
  "loss_function": "fa_cross_entropy",
  "loss_kwargs": {
    "reduction": "mean"
  },
  "masked_prediction": true,
  "max_position_embeddings": 8192,
  "mlp_bias": false,
  "mlp_dropout": 0.0,
  "mlp_dropout_prob": 0.0,
  "mlp_in_bias": false,
  "mlp_layer": "glu",
  "mlp_out_bias": false,
  "model_type": "modernbert",
  "native_text_input": true,
  "norm_bias": false,
  "norm_eps": 1e-05,
  "norm_kwargs": {
    "bias": false,
    "eps": 1e-05
  },
  "normalization": "layernorm",
  "num_attention_heads": 12,
  "num_hidden_layers": 22,
  "pad_token_id": 50283,
  "padding": "unpadded",
  "prompt_format": "[CLS] query [SEP] decoded answer text [SEP]",
  "repad_logits_with_grad": false,
  "rope_parameters": {
    "full_attention": {
      "rope_theta": 160000.0,
      "rope_type": "default"
    },
    "sliding_attention": {
      "rope_theta": 10000.0,
      "rope_type": "default"
    }
  },
  "rotary_emb_base": 160000.0,
  "rotary_emb_dim": null,
  "rotary_emb_interleaved": false,
  "rotary_emb_scale_base": null,
  "sep_token_id": 50282,
  "skip_first_prenorm": true,
  "sparse_pred_ignore_index": -100,
  "sparse_prediction": false,
  "tie_word_embeddings": true,
  "transformers_version": "5.16.1",
  "unpad_embeddings": true,
  "use_cache": false,
  "vocab_size": 50368,
  "defect_threshold_val": [
    0.99,
    0.98
  ],
  "defect_threshold_note": "[script, grammar]; tuned on out-of-fold CV predictions (validation-tuned value in defect_threshold_val)",
  "tagger_max_len": 2048,
  "tagger_stride": 256
}