samabos's picture
Training in progress, epoch 1
5b81c09 verified
{
"_name_or_path": "microsoft/deberta-v3-small",
"architectures": [
"DebertaV2ForSequenceClassification"
],
"attention_probs_dropout_prob": 0.1,
"hidden_act": "gelu",
"hidden_dropout_prob": 0.1,
"hidden_size": 768,
"id2label": {
"0": 105.13,
"1": 105.94,
"2": 106.0,
"3": 106.12,
"4": 106.2,
"5": 106.31,
"6": 205.0,
"7": 301.93,
"8": 302.84,
"9": 305.54,
"10": 305.72,
"11": 601.1,
"12": 603.19,
"13": 603.9,
"14": 710.21,
"15": 710.22,
"16": 710.8,
"17": 713.1,
"18": 713.3,
"19": 713.39,
"20": 801.11,
"21": 802.9,
"22": 811.2,
"23": 811.9,
"24": 907.2,
"25": 908.12,
"26": 910.12,
"27": 1209.24,
"28": 1522.0,
"29": 1601.0,
"30": 1702.6,
"31": 1901.1,
"32": 2006.0,
"33": 2101.12,
"34": 2101.2,
"35": 2101.3,
"36": 2207.2,
"37": 2208.4,
"38": 2305.0,
"39": 2309.1,
"40": 2515.12,
"41": 2516.12,
"42": 2521.0,
"43": 2803.0,
"44": 2918.91,
"45": 2931.35,
"46": 3002.13,
"47": 3002.14,
"48": 3002.15,
"49": 3207.2,
"50": 3405.2,
"51": 3914.0,
"52": 3917.31,
"53": 3922.2,
"54": 4003.0,
"55": 4013.1,
"56": 4017.0,
"57": 4101.9,
"58": 4115.1,
"59": 4301.9,
"60": 4302.2,
"61": 4413.0,
"62": 4415.1,
"63": 4504.1,
"64": 4707.2,
"65": 4817.2,
"66": 4819.1,
"67": 4819.2,
"68": 4819.3,
"69": 4820.4,
"70": 4903.0,
"71": 4909.0,
"72": 5104.0,
"73": 5702.5,
"74": 5801.21,
"75": 5801.23,
"76": 5801.27,
"77": 5801.31,
"78": 5801.33,
"79": 5801.37,
"80": 5806.1,
"81": 6001.1,
"82": 6502.0,
"83": 6507.0,
"84": 6602.0,
"85": 6603.2,
"86": 7001.0,
"87": 7003.12,
"88": 7004.2,
"89": 7005.21,
"90": 7009.1,
"91": 7016.1,
"92": 7107.0,
"93": 7109.0,
"94": 7111.0,
"95": 7204.41,
"96": 7208.1,
"97": 7208.4,
"98": 7211.13,
"99": 7302.3,
"100": 7323.1,
"101": 7403.21,
"102": 7407.21,
"103": 7408.21,
"104": 7408.22,
"105": 7409.4,
"106": 7411.21,
"107": 7411.22,
"108": 7609.0,
"109": 8102.95,
"110": 8205.3,
"111": 8209.0,
"112": 8306.3,
"113": 8430.1,
"114": 8441.2,
"115": 8443.91,
"116": 8452.9,
"117": 8483.1,
"118": 8483.3,
"119": 8483.6,
"120": 8483.9,
"121": 8507.1,
"122": 8517.11,
"123": 8519.2,
"124": 8525.8,
"125": 8528.52,
"126": 8540.2,
"127": 8543.7,
"128": 8545.9,
"129": 8607.3,
"130": 8703.21,
"131": 8703.22,
"132": 8703.23,
"133": 8703.24,
"134": 8703.31,
"135": 8703.32,
"136": 8703.33,
"137": 8703.4,
"138": 8708.21,
"139": 8708.5,
"140": 8714.91,
"141": 8714.96,
"142": 8907.0,
"143": 9002.11,
"144": 9018.41,
"145": 9021.4,
"146": 9033.0,
"147": 9111.2,
"148": 9301.2,
"149": 9401.3,
"150": 9603.5,
"151": 9606.1,
"152": 9607.11,
"153": 9616.1,
"154": 9616.2,
"155": 9617.0,
"156": 9618.0
},
"initializer_range": 0.02,
"intermediate_size": 3072,
"label2id": {
"105.13": 0,
"105.94": 1,
"106.0": 2,
"106.12": 3,
"106.2": 4,
"106.31": 5,
"205.0": 6,
"301.93": 7,
"302.84": 8,
"305.54": 9,
"305.72": 10,
"601.1": 11,
"603.19": 12,
"603.9": 13,
"710.21": 14,
"710.22": 15,
"710.8": 16,
"713.1": 17,
"713.3": 18,
"713.39": 19,
"801.11": 20,
"802.9": 21,
"811.2": 22,
"811.9": 23,
"907.2": 24,
"908.12": 25,
"910.12": 26,
"1209.24": 27,
"1522.0": 28,
"1601.0": 29,
"1702.6": 30,
"1901.1": 31,
"2006.0": 32,
"2101.12": 33,
"2101.2": 34,
"2101.3": 35,
"2207.2": 36,
"2208.4": 37,
"2305.0": 38,
"2309.1": 39,
"2515.12": 40,
"2516.12": 41,
"2521.0": 42,
"2803.0": 43,
"2918.91": 44,
"2931.35": 45,
"3002.13": 46,
"3002.14": 47,
"3002.15": 48,
"3207.2": 49,
"3405.2": 50,
"3914.0": 51,
"3917.31": 52,
"3922.2": 53,
"4003.0": 54,
"4013.1": 55,
"4017.0": 56,
"4101.9": 57,
"4115.1": 58,
"4301.9": 59,
"4302.2": 60,
"4413.0": 61,
"4415.1": 62,
"4504.1": 63,
"4707.2": 64,
"4817.2": 65,
"4819.1": 66,
"4819.2": 67,
"4819.3": 68,
"4820.4": 69,
"4903.0": 70,
"4909.0": 71,
"5104.0": 72,
"5702.5": 73,
"5801.21": 74,
"5801.23": 75,
"5801.27": 76,
"5801.31": 77,
"5801.33": 78,
"5801.37": 79,
"5806.1": 80,
"6001.1": 81,
"6502.0": 82,
"6507.0": 83,
"6602.0": 84,
"6603.2": 85,
"7001.0": 86,
"7003.12": 87,
"7004.2": 88,
"7005.21": 89,
"7009.1": 90,
"7016.1": 91,
"7107.0": 92,
"7109.0": 93,
"7111.0": 94,
"7204.41": 95,
"7208.1": 96,
"7208.4": 97,
"7211.13": 98,
"7302.3": 99,
"7323.1": 100,
"7403.21": 101,
"7407.21": 102,
"7408.21": 103,
"7408.22": 104,
"7409.4": 105,
"7411.21": 106,
"7411.22": 107,
"7609.0": 108,
"8102.95": 109,
"8205.3": 110,
"8209.0": 111,
"8306.3": 112,
"8430.1": 113,
"8441.2": 114,
"8443.91": 115,
"8452.9": 116,
"8483.1": 117,
"8483.3": 118,
"8483.6": 119,
"8483.9": 120,
"8507.1": 121,
"8517.11": 122,
"8519.2": 123,
"8525.8": 124,
"8528.52": 125,
"8540.2": 126,
"8543.7": 127,
"8545.9": 128,
"8607.3": 129,
"8703.21": 130,
"8703.22": 131,
"8703.23": 132,
"8703.24": 133,
"8703.31": 134,
"8703.32": 135,
"8703.33": 136,
"8703.4": 137,
"8708.21": 138,
"8708.5": 139,
"8714.91": 140,
"8714.96": 141,
"8907.0": 142,
"9002.11": 143,
"9018.41": 144,
"9021.4": 145,
"9033.0": 146,
"9111.2": 147,
"9301.2": 148,
"9401.3": 149,
"9603.5": 150,
"9606.1": 151,
"9607.11": 152,
"9616.1": 153,
"9616.2": 154,
"9617.0": 155,
"9618.0": 156
},
"layer_norm_eps": 1e-07,
"max_position_embeddings": 512,
"max_relative_positions": -1,
"model_type": "deberta-v2",
"norm_rel_ebd": "layer_norm",
"num_attention_heads": 12,
"num_hidden_layers": 6,
"pad_token_id": 0,
"pooler_dropout": 0,
"pooler_hidden_act": "gelu",
"pooler_hidden_size": 768,
"pos_att_type": [
"p2c",
"c2p"
],
"position_biased_input": false,
"position_buckets": 256,
"problem_type": "multi_label_classification",
"relative_attention": true,
"share_att_key": true,
"torch_dtype": "float32",
"transformers_version": "4.44.2",
"type_vocab_size": 0,
"vocab_size": 128100
}