Xunzi-Qwen1.5-7B-upos / config.json
KoichiYasuoka's picture
model improved for transformers 4.42
000444e
{
"architectures": [
"Qwen2ForTokenClassification"
],
"attention_dropout": 0.0,
"bos_token_id": 151643,
"custom_pipelines": {
"upos": {
"impl": "upos.BellmanFordTokenClassificationPipeline",
"pt": "AutoModelForTokenClassification"
}
},
"eos_token_id": 151643,
"hidden_act": "silu",
"hidden_size": 4096,
"id2label": {
"0": "ADP",
"1": "ADP|Degree=Equ",
"2": "ADV",
"3": "ADV|AdvType=Cau",
"4": "ADV|AdvType=Deg|Degree=Cmp",
"5": "ADV|AdvType=Deg|Degree=Pos",
"6": "ADV|AdvType=Deg|Degree=Sup",
"7": "ADV|AdvType=Tim",
"8": "ADV|AdvType=Tim|Aspect=Perf",
"9": "ADV|AdvType=Tim|Tense=Fut",
"10": "ADV|AdvType=Tim|Tense=Past",
"11": "ADV|AdvType=Tim|Tense=Pres",
"12": "ADV|Degree=Equ|VerbForm=Conv",
"13": "ADV|Degree=Pos|VerbForm=Conv",
"14": "ADV|Polarity=Neg",
"15": "ADV|Polarity=Neg|VerbForm=Conv",
"16": "ADV|VerbForm=Conv",
"17": "AUX|Mood=Des",
"18": "AUX|Mood=Nec",
"19": "AUX|Mood=Pot",
"20": "AUX|VerbType=Cop",
"21": "AUX|Voice=Pass",
"22": "B-ADP",
"23": "B-ADP|Degree=Equ",
"24": "B-ADV",
"25": "B-ADV|AdvType=Cau",
"26": "B-ADV|AdvType=Deg|Degree=Cmp",
"27": "B-ADV|AdvType=Deg|Degree=Pos",
"28": "B-ADV|AdvType=Deg|Degree=Sup",
"29": "B-ADV|AdvType=Tim",
"30": "B-ADV|AdvType=Tim|Aspect=Perf",
"31": "B-ADV|AdvType=Tim|Tense=Fut",
"32": "B-ADV|AdvType=Tim|Tense=Past",
"33": "B-ADV|AdvType=Tim|Tense=Pres",
"34": "B-ADV|Degree=Equ|VerbForm=Conv",
"35": "B-ADV|Degree=Pos|VerbForm=Conv",
"36": "B-ADV|Polarity=Neg",
"37": "B-ADV|Polarity=Neg|VerbForm=Conv",
"38": "B-ADV|VerbForm=Conv",
"39": "B-AUX|Mood=Des",
"40": "B-AUX|Mood=Nec",
"41": "B-AUX|Mood=Pot",
"42": "B-AUX|VerbType=Cop",
"43": "B-AUX|Voice=Pass",
"44": "B-CCONJ",
"45": "B-INTJ",
"46": "B-NOUN",
"47": "B-NOUN|Case=Loc",
"48": "B-NOUN|Case=Tem",
"49": "B-NOUN|Degree=Pos",
"50": "B-NOUN|NounType=Clf",
"51": "B-NUM",
"52": "B-NUM|NumType=Ord",
"53": "B-PART",
"54": "B-PRON|Person=1|PronType=Prs",
"55": "B-PRON|Person=2|PronType=Prs",
"56": "B-PRON|Person=3|PronType=Prs",
"57": "B-PRON|PronType=Dem",
"58": "B-PRON|PronType=Int",
"59": "B-PRON|PronType=Prs",
"60": "B-PRON|PronType=Prs|Reflex=Yes",
"61": "B-PROPN",
"62": "B-PROPN|Case=Loc|NameType=Geo",
"63": "B-PROPN|Case=Loc|NameType=Nat",
"64": "B-PROPN|NameType=Giv",
"65": "B-PROPN|NameType=Prs",
"66": "B-PROPN|NameType=Sur",
"67": "B-PUNCT",
"68": "B-SCONJ",
"69": "B-SYM",
"70": "B-VERB",
"71": "B-VERB|Degree=Equ",
"72": "B-VERB|Degree=Equ|VerbForm=Part",
"73": "B-VERB|Degree=Pos",
"74": "B-VERB|Degree=Pos|VerbForm=Part",
"75": "B-VERB|Polarity=Neg",
"76": "B-VERB|Polarity=Neg|VerbForm=Part",
"77": "B-VERB|VerbForm=Part",
"78": "CCONJ",
"79": "I-ADP",
"80": "I-ADP|Degree=Equ",
"81": "I-ADV",
"82": "I-ADV|AdvType=Cau",
"83": "I-ADV|AdvType=Deg|Degree=Cmp",
"84": "I-ADV|AdvType=Deg|Degree=Pos",
"85": "I-ADV|AdvType=Deg|Degree=Sup",
"86": "I-ADV|AdvType=Tim",
"87": "I-ADV|AdvType=Tim|Aspect=Perf",
"88": "I-ADV|AdvType=Tim|Tense=Fut",
"89": "I-ADV|AdvType=Tim|Tense=Past",
"90": "I-ADV|AdvType=Tim|Tense=Pres",
"91": "I-ADV|Degree=Equ|VerbForm=Conv",
"92": "I-ADV|Degree=Pos|VerbForm=Conv",
"93": "I-ADV|Polarity=Neg",
"94": "I-ADV|Polarity=Neg|VerbForm=Conv",
"95": "I-ADV|VerbForm=Conv",
"96": "I-AUX|Mood=Des",
"97": "I-AUX|Mood=Nec",
"98": "I-AUX|Mood=Pot",
"99": "I-AUX|VerbType=Cop",
"100": "I-AUX|Voice=Pass",
"101": "I-CCONJ",
"102": "I-INTJ",
"103": "I-NOUN",
"104": "I-NOUN|Case=Loc",
"105": "I-NOUN|Case=Tem",
"106": "I-NOUN|Degree=Pos",
"107": "I-NOUN|NounType=Clf",
"108": "I-NUM",
"109": "I-NUM|NumType=Ord",
"110": "I-PART",
"111": "I-PRON|Person=1|PronType=Prs",
"112": "I-PRON|Person=2|PronType=Prs",
"113": "I-PRON|Person=3|PronType=Prs",
"114": "I-PRON|PronType=Dem",
"115": "I-PRON|PronType=Int",
"116": "I-PRON|PronType=Prs",
"117": "I-PRON|PronType=Prs|Reflex=Yes",
"118": "I-PROPN",
"119": "I-PROPN|Case=Loc|NameType=Geo",
"120": "I-PROPN|Case=Loc|NameType=Nat",
"121": "I-PROPN|NameType=Giv",
"122": "I-PROPN|NameType=Prs",
"123": "I-PROPN|NameType=Sur",
"124": "I-PUNCT",
"125": "I-SCONJ",
"126": "I-SYM",
"127": "I-VERB",
"128": "I-VERB|Degree=Equ",
"129": "I-VERB|Degree=Equ|VerbForm=Part",
"130": "I-VERB|Degree=Pos",
"131": "I-VERB|Degree=Pos|VerbForm=Part",
"132": "I-VERB|Polarity=Neg",
"133": "I-VERB|Polarity=Neg|VerbForm=Part",
"134": "I-VERB|VerbForm=Part",
"135": "INTJ",
"136": "NOUN",
"137": "NOUN|Case=Loc",
"138": "NOUN|Case=Tem",
"139": "NOUN|Degree=Pos",
"140": "NOUN|NounType=Clf",
"141": "NUM",
"142": "NUM|NumType=Ord",
"143": "PART",
"144": "PRON|Person=1|PronType=Prs",
"145": "PRON|Person=2|PronType=Prs",
"146": "PRON|Person=3|PronType=Prs",
"147": "PRON|PronType=Dem",
"148": "PRON|PronType=Int",
"149": "PRON|PronType=Prs",
"150": "PRON|PronType=Prs|Reflex=Yes",
"151": "PROPN",
"152": "PROPN|Case=Loc|NameType=Geo",
"153": "PROPN|Case=Loc|NameType=Nat",
"154": "PROPN|NameType=Giv",
"155": "PROPN|NameType=Prs",
"156": "PROPN|NameType=Sur",
"157": "PUNCT",
"158": "SCONJ",
"159": "SYM",
"160": "VERB",
"161": "VERB|Degree=Equ",
"162": "VERB|Degree=Equ|VerbForm=Part",
"163": "VERB|Degree=Pos",
"164": "VERB|Degree=Pos|VerbForm=Part",
"165": "VERB|Polarity=Neg",
"166": "VERB|Polarity=Neg|VerbForm=Part",
"167": "VERB|VerbForm=Part"
},
"initializer_range": 0.02,
"intermediate_size": 11008,
"label2id": {
"ADP": 0,
"ADP|Degree=Equ": 1,
"ADV": 2,
"ADV|AdvType=Cau": 3,
"ADV|AdvType=Deg|Degree=Cmp": 4,
"ADV|AdvType=Deg|Degree=Pos": 5,
"ADV|AdvType=Deg|Degree=Sup": 6,
"ADV|AdvType=Tim": 7,
"ADV|AdvType=Tim|Aspect=Perf": 8,
"ADV|AdvType=Tim|Tense=Fut": 9,
"ADV|AdvType=Tim|Tense=Past": 10,
"ADV|AdvType=Tim|Tense=Pres": 11,
"ADV|Degree=Equ|VerbForm=Conv": 12,
"ADV|Degree=Pos|VerbForm=Conv": 13,
"ADV|Polarity=Neg": 14,
"ADV|Polarity=Neg|VerbForm=Conv": 15,
"ADV|VerbForm=Conv": 16,
"AUX|Mood=Des": 17,
"AUX|Mood=Nec": 18,
"AUX|Mood=Pot": 19,
"AUX|VerbType=Cop": 20,
"AUX|Voice=Pass": 21,
"B-ADP": 22,
"B-ADP|Degree=Equ": 23,
"B-ADV": 24,
"B-ADV|AdvType=Cau": 25,
"B-ADV|AdvType=Deg|Degree=Cmp": 26,
"B-ADV|AdvType=Deg|Degree=Pos": 27,
"B-ADV|AdvType=Deg|Degree=Sup": 28,
"B-ADV|AdvType=Tim": 29,
"B-ADV|AdvType=Tim|Aspect=Perf": 30,
"B-ADV|AdvType=Tim|Tense=Fut": 31,
"B-ADV|AdvType=Tim|Tense=Past": 32,
"B-ADV|AdvType=Tim|Tense=Pres": 33,
"B-ADV|Degree=Equ|VerbForm=Conv": 34,
"B-ADV|Degree=Pos|VerbForm=Conv": 35,
"B-ADV|Polarity=Neg": 36,
"B-ADV|Polarity=Neg|VerbForm=Conv": 37,
"B-ADV|VerbForm=Conv": 38,
"B-AUX|Mood=Des": 39,
"B-AUX|Mood=Nec": 40,
"B-AUX|Mood=Pot": 41,
"B-AUX|VerbType=Cop": 42,
"B-AUX|Voice=Pass": 43,
"B-CCONJ": 44,
"B-INTJ": 45,
"B-NOUN": 46,
"B-NOUN|Case=Loc": 47,
"B-NOUN|Case=Tem": 48,
"B-NOUN|Degree=Pos": 49,
"B-NOUN|NounType=Clf": 50,
"B-NUM": 51,
"B-NUM|NumType=Ord": 52,
"B-PART": 53,
"B-PRON|Person=1|PronType=Prs": 54,
"B-PRON|Person=2|PronType=Prs": 55,
"B-PRON|Person=3|PronType=Prs": 56,
"B-PRON|PronType=Dem": 57,
"B-PRON|PronType=Int": 58,
"B-PRON|PronType=Prs": 59,
"B-PRON|PronType=Prs|Reflex=Yes": 60,
"B-PROPN": 61,
"B-PROPN|Case=Loc|NameType=Geo": 62,
"B-PROPN|Case=Loc|NameType=Nat": 63,
"B-PROPN|NameType=Giv": 64,
"B-PROPN|NameType=Prs": 65,
"B-PROPN|NameType=Sur": 66,
"B-PUNCT": 67,
"B-SCONJ": 68,
"B-SYM": 69,
"B-VERB": 70,
"B-VERB|Degree=Equ": 71,
"B-VERB|Degree=Equ|VerbForm=Part": 72,
"B-VERB|Degree=Pos": 73,
"B-VERB|Degree=Pos|VerbForm=Part": 74,
"B-VERB|Polarity=Neg": 75,
"B-VERB|Polarity=Neg|VerbForm=Part": 76,
"B-VERB|VerbForm=Part": 77,
"CCONJ": 78,
"I-ADP": 79,
"I-ADP|Degree=Equ": 80,
"I-ADV": 81,
"I-ADV|AdvType=Cau": 82,
"I-ADV|AdvType=Deg|Degree=Cmp": 83,
"I-ADV|AdvType=Deg|Degree=Pos": 84,
"I-ADV|AdvType=Deg|Degree=Sup": 85,
"I-ADV|AdvType=Tim": 86,
"I-ADV|AdvType=Tim|Aspect=Perf": 87,
"I-ADV|AdvType=Tim|Tense=Fut": 88,
"I-ADV|AdvType=Tim|Tense=Past": 89,
"I-ADV|AdvType=Tim|Tense=Pres": 90,
"I-ADV|Degree=Equ|VerbForm=Conv": 91,
"I-ADV|Degree=Pos|VerbForm=Conv": 92,
"I-ADV|Polarity=Neg": 93,
"I-ADV|Polarity=Neg|VerbForm=Conv": 94,
"I-ADV|VerbForm=Conv": 95,
"I-AUX|Mood=Des": 96,
"I-AUX|Mood=Nec": 97,
"I-AUX|Mood=Pot": 98,
"I-AUX|VerbType=Cop": 99,
"I-AUX|Voice=Pass": 100,
"I-CCONJ": 101,
"I-INTJ": 102,
"I-NOUN": 103,
"I-NOUN|Case=Loc": 104,
"I-NOUN|Case=Tem": 105,
"I-NOUN|Degree=Pos": 106,
"I-NOUN|NounType=Clf": 107,
"I-NUM": 108,
"I-NUM|NumType=Ord": 109,
"I-PART": 110,
"I-PRON|Person=1|PronType=Prs": 111,
"I-PRON|Person=2|PronType=Prs": 112,
"I-PRON|Person=3|PronType=Prs": 113,
"I-PRON|PronType=Dem": 114,
"I-PRON|PronType=Int": 115,
"I-PRON|PronType=Prs": 116,
"I-PRON|PronType=Prs|Reflex=Yes": 117,
"I-PROPN": 118,
"I-PROPN|Case=Loc|NameType=Geo": 119,
"I-PROPN|Case=Loc|NameType=Nat": 120,
"I-PROPN|NameType=Giv": 121,
"I-PROPN|NameType=Prs": 122,
"I-PROPN|NameType=Sur": 123,
"I-PUNCT": 124,
"I-SCONJ": 125,
"I-SYM": 126,
"I-VERB": 127,
"I-VERB|Degree=Equ": 128,
"I-VERB|Degree=Equ|VerbForm=Part": 129,
"I-VERB|Degree=Pos": 130,
"I-VERB|Degree=Pos|VerbForm=Part": 131,
"I-VERB|Polarity=Neg": 132,
"I-VERB|Polarity=Neg|VerbForm=Part": 133,
"I-VERB|VerbForm=Part": 134,
"INTJ": 135,
"NOUN": 136,
"NOUN|Case=Loc": 137,
"NOUN|Case=Tem": 138,
"NOUN|Degree=Pos": 139,
"NOUN|NounType=Clf": 140,
"NUM": 141,
"NUM|NumType=Ord": 142,
"PART": 143,
"PRON|Person=1|PronType=Prs": 144,
"PRON|Person=2|PronType=Prs": 145,
"PRON|Person=3|PronType=Prs": 146,
"PRON|PronType=Dem": 147,
"PRON|PronType=Int": 148,
"PRON|PronType=Prs": 149,
"PRON|PronType=Prs|Reflex=Yes": 150,
"PROPN": 151,
"PROPN|Case=Loc|NameType=Geo": 152,
"PROPN|Case=Loc|NameType=Nat": 153,
"PROPN|NameType=Giv": 154,
"PROPN|NameType=Prs": 155,
"PROPN|NameType=Sur": 156,
"PUNCT": 157,
"SCONJ": 158,
"SYM": 159,
"VERB": 160,
"VERB|Degree=Equ": 161,
"VERB|Degree=Equ|VerbForm=Part": 162,
"VERB|Degree=Pos": 163,
"VERB|Degree=Pos|VerbForm=Part": 164,
"VERB|Polarity=Neg": 165,
"VERB|Polarity=Neg|VerbForm=Part": 166,
"VERB|VerbForm=Part": 167
},
"max_position_embeddings": 32768,
"max_window_layers": 28,
"model_type": "qwen2",
"num_attention_heads": 32,
"num_hidden_layers": 32,
"num_key_value_heads": 32,
"rms_norm_eps": 1e-06,
"rope_theta": 1000000.0,
"sliding_window": 32768,
"tie_word_embeddings": false,
"tokenizer_class": "Qwen2Tokenizer",
"torch_dtype": "float32",
"transformers_version": "4.42.4",
"use_cache": false,
"use_sliding_window": false,
"vocab_size": 151936
}