stulcrad's picture
Training in progress, step 1000
9a85902 verified
raw
history blame
3.63 kB
{
"_name_or_path": "ufal/robeczech-base",
"architectures": [
"RobertaForTokenClassification"
],
"attention_probs_dropout_prob": 0.2,
"bos_token_id": 0,
"classifier_dropout": null,
"eos_token_id": 2,
"gradient_checkpointing": false,
"hidden_act": "gelu",
"hidden_dropout_prob": 0.3,
"hidden_size": 768,
"id2label": {
"0": "O",
"1": "B-ah",
"2": "I-ah",
"3": "B-at",
"4": "I-at",
"5": "B-az",
"6": "I-az",
"7": "B-g_",
"8": "I-g_",
"9": "B-gc",
"10": "I-gc",
"11": "B-gh",
"12": "I-gh",
"13": "B-gl",
"14": "I-gl",
"15": "B-gp",
"16": "I-gp",
"17": "B-gq",
"18": "I-gq",
"19": "B-gr",
"20": "I-gr",
"21": "B-gs",
"22": "I-gs",
"23": "B-gt",
"24": "I-gt",
"25": "B-gu",
"26": "I-gu",
"27": "B-i_",
"28": "I-i_",
"29": "B-ia",
"30": "I-ia",
"31": "B-ic",
"32": "I-ic",
"33": "B-if",
"34": "I-if",
"35": "B-io",
"36": "I-io",
"37": "B-mn",
"38": "I-mn",
"39": "B-mt",
"40": "I-mt",
"41": "B-mr",
"42": "I-mr",
"43": "B-o_",
"44": "I-o_",
"45": "B-oa",
"46": "I-oa",
"47": "B-oc",
"48": "I-oc",
"49": "B-oe",
"50": "I-oe",
"51": "B-om",
"52": "I-om",
"53": "B-op",
"54": "I-op",
"55": "B-or",
"56": "I-or",
"57": "B-p_",
"58": "I-p_",
"59": "B-pb",
"60": "I-pb",
"61": "B-pc",
"62": "I-pc",
"63": "B-pd",
"64": "I-pd",
"65": "B-pf",
"66": "I-pf",
"67": "B-pm",
"68": "I-pm",
"69": "B-pp",
"70": "I-pp",
"71": "B-ps",
"72": "I-ps",
"73": "B-td",
"74": "I-td",
"75": "B-tf",
"76": "I-tf",
"77": "B-th",
"78": "I-th",
"79": "B-ti",
"80": "I-ti",
"81": "B-tm",
"82": "I-tm",
"83": "B-ty",
"84": "I-ty"
},
"initializer_range": 0.02,
"intermediate_size": 3072,
"label2id": {
"B-ah": 1,
"B-at": 3,
"B-az": 5,
"B-g_": 7,
"B-gc": 9,
"B-gh": 11,
"B-gl": 13,
"B-gp": 15,
"B-gq": 17,
"B-gr": 19,
"B-gs": 21,
"B-gt": 23,
"B-gu": 25,
"B-i_": 27,
"B-ia": 29,
"B-ic": 31,
"B-if": 33,
"B-io": 35,
"B-mn": 37,
"B-mr": 41,
"B-mt": 39,
"B-o_": 43,
"B-oa": 45,
"B-oc": 47,
"B-oe": 49,
"B-om": 51,
"B-op": 53,
"B-or": 55,
"B-p_": 57,
"B-pb": 59,
"B-pc": 61,
"B-pd": 63,
"B-pf": 65,
"B-pm": 67,
"B-pp": 69,
"B-ps": 71,
"B-td": 73,
"B-tf": 75,
"B-th": 77,
"B-ti": 79,
"B-tm": 81,
"B-ty": 83,
"I-ah": 2,
"I-at": 4,
"I-az": 6,
"I-g_": 8,
"I-gc": 10,
"I-gh": 12,
"I-gl": 14,
"I-gp": 16,
"I-gq": 18,
"I-gr": 20,
"I-gs": 22,
"I-gt": 24,
"I-gu": 26,
"I-i_": 28,
"I-ia": 30,
"I-ic": 32,
"I-if": 34,
"I-io": 36,
"I-mn": 38,
"I-mr": 42,
"I-mt": 40,
"I-o_": 44,
"I-oa": 46,
"I-oc": 48,
"I-oe": 50,
"I-om": 52,
"I-op": 54,
"I-or": 56,
"I-p_": 58,
"I-pb": 60,
"I-pc": 62,
"I-pd": 64,
"I-pf": 66,
"I-pm": 68,
"I-pp": 70,
"I-ps": 72,
"I-td": 74,
"I-tf": 76,
"I-th": 78,
"I-ti": 80,
"I-tm": 82,
"I-ty": 84,
"O": 0
},
"layer_norm_eps": 1e-05,
"max_position_embeddings": 514,
"model_type": "roberta",
"num_attention_heads": 12,
"num_hidden_layers": 12,
"pad_token_id": 1,
"position_embedding_type": "absolute",
"torch_dtype": "float32",
"transformers_version": "4.36.2",
"type_vocab_size": 1,
"use_cache": true,
"vocab_size": 51997
}