stulcrad's picture
Training in progress, step 500
6f92c53 verified
{
"_name_or_path": "FacebookAI/xlm-roberta-large",
"architectures": [
"XLMRobertaForTokenClassification"
],
"attention_probs_dropout_prob": 0.15,
"bos_token_id": 0,
"classifier_dropout": null,
"eos_token_id": 2,
"hidden_act": "gelu",
"hidden_dropout_prob": 0.2,
"hidden_size": 1024,
"id2label": {
"0": "O",
"1": "B-ah",
"2": "I-ah",
"3": "B-at",
"4": "I-at",
"5": "B-az",
"6": "I-az",
"7": "B-cb",
"8": "I-cb",
"9": "B-cn",
"10": "I-cn",
"11": "B-cp",
"12": "I-cp",
"13": "B-cr",
"14": "I-cr",
"15": "B-cs",
"16": "I-cs",
"17": "B-g_",
"18": "I-g_",
"19": "B-gc",
"20": "I-gc",
"21": "B-gh",
"22": "I-gh",
"23": "B-gl",
"24": "I-gl",
"25": "B-gp",
"26": "I-gp",
"27": "B-gq",
"28": "I-gq",
"29": "B-gr",
"30": "I-gr",
"31": "B-gs",
"32": "I-gs",
"33": "B-gt",
"34": "I-gt",
"35": "B-gu",
"36": "I-gu",
"37": "B-i_",
"38": "I-i_",
"39": "B-ia",
"40": "I-ia",
"41": "B-ic",
"42": "I-ic",
"43": "B-if",
"44": "I-if",
"45": "B-io",
"46": "I-io",
"47": "B-mi",
"48": "I-mi",
"49": "B-mn",
"50": "I-mn",
"51": "B-mr",
"52": "I-mr",
"53": "B-mt",
"54": "I-mt",
"55": "B-n_",
"56": "I-n_",
"57": "B-na",
"58": "I-na",
"59": "B-nc",
"60": "I-nc",
"61": "B-ni",
"62": "I-ni",
"63": "B-nm",
"64": "I-nm",
"65": "B-np",
"66": "I-np",
"67": "B-nq",
"68": "I-nq",
"69": "B-nr",
"70": "I-nr",
"71": "B-nw",
"72": "I-nw",
"73": "B-o_",
"74": "I-o_",
"75": "B-oa",
"76": "I-oa",
"77": "B-oc",
"78": "I-oc",
"79": "B-oe",
"80": "I-oe",
"81": "B-om",
"82": "I-om",
"83": "B-op",
"84": "I-op",
"85": "B-or",
"86": "I-or",
"87": "B-p_",
"88": "I-p_",
"89": "B-pb",
"90": "I-pb",
"91": "B-pc",
"92": "I-pc",
"93": "B-pd",
"94": "I-pd",
"95": "B-pf",
"96": "I-pf",
"97": "B-pm",
"98": "I-pm",
"99": "B-pp",
"100": "I-pp",
"101": "B-ps",
"102": "I-ps",
"103": "B-qc",
"104": "I-qc",
"105": "B-qo",
"106": "I-qo",
"107": "B-tc",
"108": "I-tc",
"109": "B-td",
"110": "I-td",
"111": "B-tf",
"112": "I-tf",
"113": "B-th",
"114": "I-th",
"115": "B-tm",
"116": "I-tm",
"117": "B-tn",
"118": "I-tn",
"119": "B-tp",
"120": "I-tp",
"121": "B-ts",
"122": "I-ts",
"123": "B-ty",
"124": "I-ty"
},
"initializer_range": 0.02,
"intermediate_size": 4096,
"label2id": {
"B-ah": 1,
"B-at": 3,
"B-az": 5,
"B-cb": 7,
"B-cn": 9,
"B-cp": 11,
"B-cr": 13,
"B-cs": 15,
"B-g_": 17,
"B-gc": 19,
"B-gh": 21,
"B-gl": 23,
"B-gp": 25,
"B-gq": 27,
"B-gr": 29,
"B-gs": 31,
"B-gt": 33,
"B-gu": 35,
"B-i_": 37,
"B-ia": 39,
"B-ic": 41,
"B-if": 43,
"B-io": 45,
"B-mi": 47,
"B-mn": 49,
"B-mr": 51,
"B-mt": 53,
"B-n_": 55,
"B-na": 57,
"B-nc": 59,
"B-ni": 61,
"B-nm": 63,
"B-np": 65,
"B-nq": 67,
"B-nr": 69,
"B-nw": 71,
"B-o_": 73,
"B-oa": 75,
"B-oc": 77,
"B-oe": 79,
"B-om": 81,
"B-op": 83,
"B-or": 85,
"B-p_": 87,
"B-pb": 89,
"B-pc": 91,
"B-pd": 93,
"B-pf": 95,
"B-pm": 97,
"B-pp": 99,
"B-ps": 101,
"B-qc": 103,
"B-qo": 105,
"B-tc": 107,
"B-td": 109,
"B-tf": 111,
"B-th": 113,
"B-tm": 115,
"B-tn": 117,
"B-tp": 119,
"B-ts": 121,
"B-ty": 123,
"I-ah": 2,
"I-at": 4,
"I-az": 6,
"I-cb": 8,
"I-cn": 10,
"I-cp": 12,
"I-cr": 14,
"I-cs": 16,
"I-g_": 18,
"I-gc": 20,
"I-gh": 22,
"I-gl": 24,
"I-gp": 26,
"I-gq": 28,
"I-gr": 30,
"I-gs": 32,
"I-gt": 34,
"I-gu": 36,
"I-i_": 38,
"I-ia": 40,
"I-ic": 42,
"I-if": 44,
"I-io": 46,
"I-mi": 48,
"I-mn": 50,
"I-mr": 52,
"I-mt": 54,
"I-n_": 56,
"I-na": 58,
"I-nc": 60,
"I-ni": 62,
"I-nm": 64,
"I-np": 66,
"I-nq": 68,
"I-nr": 70,
"I-nw": 72,
"I-o_": 74,
"I-oa": 76,
"I-oc": 78,
"I-oe": 80,
"I-om": 82,
"I-op": 84,
"I-or": 86,
"I-p_": 88,
"I-pb": 90,
"I-pc": 92,
"I-pd": 94,
"I-pf": 96,
"I-pm": 98,
"I-pp": 100,
"I-ps": 102,
"I-qc": 104,
"I-qo": 106,
"I-tc": 108,
"I-td": 110,
"I-tf": 112,
"I-th": 114,
"I-tm": 116,
"I-tn": 118,
"I-tp": 120,
"I-ts": 122,
"I-ty": 124,
"O": 0
},
"layer_norm_eps": 1e-05,
"max_position_embeddings": 514,
"model_type": "xlm-roberta",
"num_attention_heads": 16,
"num_hidden_layers": 24,
"output_past": true,
"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": 250002
}