HeRo-finetuned-ner / config.json
msperka's picture
Training in progress, epoch 1
2046c10
raw
history blame
2.15 kB
{
"_name_or_path": "HeNLP/HeRo",
"architectures": [
"RobertaForTokenClassification"
],
"attention_probs_dropout_prob": 0.1,
"bos_token_id": 0,
"classifier_dropout": null,
"device_map": "dik",
"eos_token_id": 2,
"hidden_act": "gelu",
"hidden_dropout_prob": 0.1,
"hidden_size": 768,
"id2label": {
"0": "S-ANG",
"1": "B-ANG",
"2": "I-ANG",
"3": "E-ANG",
"4": "S-DUC",
"5": "B-DUC",
"6": "I-DUC",
"7": "E-DUC",
"8": "B-EVE",
"9": "E-EVE",
"10": "S-EVE",
"11": "I-EVE",
"12": "S-FAC",
"13": "B-FAC",
"14": "E-FAC",
"15": "I-FAC",
"16": "S-GPE",
"17": "B-GPE",
"18": "E-GPE",
"19": "I-GPE",
"20": "S-LOC",
"21": "B-LOC",
"22": "E-LOC",
"23": "I-LOC",
"24": "O",
"25": "S-ORG",
"26": "B-ORG",
"27": "E-ORG",
"28": "I-ORG",
"29": "B-PER",
"30": "I-PER",
"31": "E-PER",
"32": "S-PER",
"33": "B-WOA",
"34": "E-WOA",
"35": "I-WOA",
"36": "S-WOA"
},
"initializer_range": 0.02,
"intermediate_size": 3072,
"label2id": {
"B-ANG": 1,
"B-DUC": 5,
"B-EVE": 8,
"B-FAC": 13,
"B-GPE": 17,
"B-LOC": 21,
"B-ORG": 26,
"B-PER": 29,
"B-WOA": 33,
"E-ANG": 3,
"E-DUC": 7,
"E-EVE": 9,
"E-FAC": 14,
"E-GPE": 18,
"E-LOC": 22,
"E-ORG": 27,
"E-PER": 31,
"E-WOA": 34,
"I-ANG": 2,
"I-DUC": 6,
"I-EVE": 11,
"I-FAC": 15,
"I-GPE": 19,
"I-LOC": 23,
"I-ORG": 28,
"I-PER": 30,
"I-WOA": 35,
"O": 24,
"S-ANG": 0,
"S-DUC": 4,
"S-EVE": 10,
"S-FAC": 12,
"S-GPE": 16,
"S-LOC": 20,
"S-ORG": 25,
"S-PER": 32,
"S-WOA": 36
},
"layer_norm_eps": 1e-05,
"max_position_embeddings": 514,
"model_type": "roberta",
"num_attention_heads": 12,
"num_hidden_layers": 12,
"offload_folder": "os.path.join(path.parent.parent.absolute(), 'results/RoBERTa/model_and_tokenizer')",
"pad_token_id": 1,
"position_embedding_type": "absolute",
"torch_dtype": "float32",
"transformers_version": "4.35.2",
"type_vocab_size": 1,
"use_cache": true,
"vocab_size": 50265
}