Javtor's picture
Training in progress, epoch 1
803c944
raw history blame
No virus
5.04 kB
{
"_name_or_path": "Javtor/biomedical-topic-categorization-2022only",
"architectures": [
"BertForSequenceClassification"
],
"attention_probs_dropout_prob": 0.1,
"classifier_dropout": null,
"gradient_checkpointing": false,
"hidden_act": "gelu",
"hidden_dropout_prob": 0.1,
"hidden_size": 1024,
"id2label": {
"0": "T001",
"1": "T002",
"2": "T004",
"3": "T005",
"4": "T007",
"5": "T008",
"6": "T010",
"7": "T011",
"8": "T012",
"9": "T013",
"10": "T014",
"11": "T015",
"12": "T016",
"13": "T017",
"14": "T018",
"15": "T019",
"16": "T020",
"17": "T022",
"18": "T023",
"19": "T024",
"20": "T025",
"21": "T026",
"22": "T028",
"23": "T029",
"24": "T030",
"25": "T031",
"26": "T032",
"27": "T033",
"28": "T034",
"29": "T037",
"30": "T038",
"31": "T039",
"32": "T040",
"33": "T041",
"34": "T042",
"35": "T043",
"36": "T044",
"37": "T045",
"38": "T046",
"39": "T047",
"40": "T048",
"41": "T049",
"42": "T050",
"43": "T051",
"44": "T052",
"45": "T053",
"46": "T054",
"47": "T055",
"48": "T056",
"49": "T057",
"50": "T058",
"51": "T059",
"52": "T060",
"53": "T061",
"54": "T062",
"55": "T063",
"56": "T064",
"57": "T065",
"58": "T066",
"59": "T067",
"60": "T068",
"61": "T069",
"62": "T070",
"63": "T071",
"64": "T072",
"65": "T073",
"66": "T074",
"67": "T075",
"68": "T077",
"69": "T078",
"70": "T079",
"71": "T080",
"72": "T081",
"73": "T082",
"74": "T083",
"75": "T085",
"76": "T086",
"77": "T087",
"78": "T089",
"79": "T090",
"80": "T091",
"81": "T092",
"82": "T093",
"83": "T094",
"84": "T095",
"85": "T096",
"86": "T097",
"87": "T098",
"88": "T099",
"89": "T100",
"90": "T101",
"91": "T102",
"92": "T103",
"93": "T104",
"94": "T109",
"95": "T114",
"96": "T116",
"97": "T120",
"98": "T121",
"99": "T122",
"100": "T123",
"101": "T125",
"102": "T126",
"103": "T127",
"104": "T129",
"105": "T130",
"106": "T131",
"107": "T167",
"108": "T168",
"109": "T169",
"110": "T170",
"111": "T171",
"112": "T184",
"113": "T185",
"114": "T190",
"115": "T191",
"116": "T192",
"117": "T194",
"118": "T195",
"119": "T196",
"120": "T197",
"121": "T200",
"122": "T201",
"123": "T204"
},
"initializer_range": 0.02,
"intermediate_size": 4096,
"label2id": {
"T001": 0,
"T002": 1,
"T004": 2,
"T005": 3,
"T007": 4,
"T008": 5,
"T010": 6,
"T011": 7,
"T012": 8,
"T013": 9,
"T014": 10,
"T015": 11,
"T016": 12,
"T017": 13,
"T018": 14,
"T019": 15,
"T020": 16,
"T022": 17,
"T023": 18,
"T024": 19,
"T025": 20,
"T026": 21,
"T028": 22,
"T029": 23,
"T030": 24,
"T031": 25,
"T032": 26,
"T033": 27,
"T034": 28,
"T037": 29,
"T038": 30,
"T039": 31,
"T040": 32,
"T041": 33,
"T042": 34,
"T043": 35,
"T044": 36,
"T045": 37,
"T046": 38,
"T047": 39,
"T048": 40,
"T049": 41,
"T050": 42,
"T051": 43,
"T052": 44,
"T053": 45,
"T054": 46,
"T055": 47,
"T056": 48,
"T057": 49,
"T058": 50,
"T059": 51,
"T060": 52,
"T061": 53,
"T062": 54,
"T063": 55,
"T064": 56,
"T065": 57,
"T066": 58,
"T067": 59,
"T068": 60,
"T069": 61,
"T070": 62,
"T071": 63,
"T072": 64,
"T073": 65,
"T074": 66,
"T075": 67,
"T077": 68,
"T078": 69,
"T079": 70,
"T080": 71,
"T081": 72,
"T082": 73,
"T083": 74,
"T085": 75,
"T086": 76,
"T087": 77,
"T089": 78,
"T090": 79,
"T091": 80,
"T092": 81,
"T093": 82,
"T094": 83,
"T095": 84,
"T096": 85,
"T097": 86,
"T098": 87,
"T099": 88,
"T100": 89,
"T101": 90,
"T102": 91,
"T103": 92,
"T104": 93,
"T109": 94,
"T114": 95,
"T116": 96,
"T120": 97,
"T121": 98,
"T122": 99,
"T123": 100,
"T125": 101,
"T126": 102,
"T127": 103,
"T129": 104,
"T130": 105,
"T131": 106,
"T167": 107,
"T168": 108,
"T169": 109,
"T170": 110,
"T171": 111,
"T184": 112,
"T185": 113,
"T190": 114,
"T191": 115,
"T192": 116,
"T194": 117,
"T195": 118,
"T196": 119,
"T197": 120,
"T200": 121,
"T201": 122,
"T204": 123
},
"layer_norm_eps": 1e-12,
"max_position_embeddings": 512,
"model_type": "bert",
"num_attention_heads": 16,
"num_hidden_layers": 24,
"pad_token_id": 0,
"position_embedding_type": "absolute",
"problem_type": "multi_label_classification",
"torch_dtype": "float32",
"transformers_version": "4.23.1",
"type_vocab_size": 2,
"use_cache": true,
"vocab_size": 28895
}