| { |
| "architectures": [ |
| "ModernBertForSequenceClassification" |
| ], |
| "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": 768, |
| "id2label": { |
| "0": "T1001", |
| "1": "T1005", |
| "2": "T1020", |
| "3": "T1021", |
| "4": "T1027", |
| "5": "T1033", |
| "6": "T1036", |
| "7": "T1040", |
| "8": "T1041", |
| "9": "T1046", |
| "10": "T1049", |
| "11": "T1053", |
| "12": "T1056", |
| "13": "T1059", |
| "14": "T1068", |
| "15": "T1069", |
| "16": "T1074", |
| "17": "T1078", |
| "18": "T1080", |
| "19": "T1082", |
| "20": "T1087", |
| "21": "T1098", |
| "22": "T1105", |
| "23": "T1106", |
| "24": "T1119", |
| "25": "T1132", |
| "26": "T1133", |
| "27": "T1134", |
| "28": "T1136", |
| "29": "T1187", |
| "30": "T1189", |
| "31": "T1190", |
| "32": "T1195", |
| "33": "T1200", |
| "34": "T1203", |
| "35": "T1204", |
| "36": "T1213", |
| "37": "T1485", |
| "38": "T1486", |
| "39": "T1489", |
| "40": "T1491", |
| "41": "T1495", |
| "42": "T1498", |
| "43": "T1499", |
| "44": "T1518", |
| "45": "T1529", |
| "46": "T1531", |
| "47": "T1548", |
| "48": "T1552", |
| "49": "T1557", |
| "50": "T1559", |
| "51": "T1560", |
| "52": "T1562", |
| "53": "T1565", |
| "54": "T1566", |
| "55": "T1595", |
| "56": "T1614", |
| "57": "T1652", |
| "58": "T1654", |
| "59": "T1657", |
| "60": "T1659" |
| }, |
| "initializer_cutoff_factor": 2.0, |
| "initializer_range": 0.02, |
| "intermediate_size": 1152, |
| "label2id": { |
| "T1001": 0, |
| "T1005": 1, |
| "T1020": 2, |
| "T1021": 3, |
| "T1027": 4, |
| "T1033": 5, |
| "T1036": 6, |
| "T1040": 7, |
| "T1041": 8, |
| "T1046": 9, |
| "T1049": 10, |
| "T1053": 11, |
| "T1056": 12, |
| "T1059": 13, |
| "T1068": 14, |
| "T1069": 15, |
| "T1074": 16, |
| "T1078": 17, |
| "T1080": 18, |
| "T1082": 19, |
| "T1087": 20, |
| "T1098": 21, |
| "T1105": 22, |
| "T1106": 23, |
| "T1119": 24, |
| "T1132": 25, |
| "T1133": 26, |
| "T1134": 27, |
| "T1136": 28, |
| "T1187": 29, |
| "T1189": 30, |
| "T1190": 31, |
| "T1195": 32, |
| "T1200": 33, |
| "T1203": 34, |
| "T1204": 35, |
| "T1213": 36, |
| "T1485": 37, |
| "T1486": 38, |
| "T1489": 39, |
| "T1491": 40, |
| "T1495": 41, |
| "T1498": 42, |
| "T1499": 43, |
| "T1518": 44, |
| "T1529": 45, |
| "T1531": 46, |
| "T1548": 47, |
| "T1552": 48, |
| "T1557": 49, |
| "T1559": 50, |
| "T1560": 51, |
| "T1562": 52, |
| "T1565": 53, |
| "T1566": 54, |
| "T1595": 55, |
| "T1614": 56, |
| "T1652": 57, |
| "T1654": 58, |
| "T1657": 59, |
| "T1659": 60 |
| }, |
| "layer_norm_eps": 1e-05, |
| "local_attention": 128, |
| "local_rope_theta": 10000.0, |
| "max_position_embeddings": 8192, |
| "mlp_bias": false, |
| "mlp_dropout": 0.0, |
| "model_type": "modernbert", |
| "norm_bias": false, |
| "norm_eps": 1e-05, |
| "num_attention_heads": 12, |
| "num_hidden_layers": 22, |
| "pad_token_id": 50283, |
| "position_embedding_type": "absolute", |
| "problem_type": "multi_label_classification", |
| "repad_logits_with_grad": false, |
| "sep_token_id": 50282, |
| "sparse_pred_ignore_index": -100, |
| "sparse_prediction": false, |
| "torch_dtype": "float32", |
| "transformers_version": "4.55.0", |
| "vocab_size": 50368 |
| } |
|
|