{ "activation_dropout": 0.0, "apply_spec_augment": true, "architectures": [ "Wav2Vec2ForCTC" ], "attention_dropout": 0.1, "bos_token_id": 1, "classifier_proj_size": 256, "codevector_dim": 768, "contrastive_logits_temperature": 0.1, "conv_bias": true, "conv_dim": [ 512, 512, 512, 512, 512, 512, 512 ], "conv_kernel": [ 10, 3, 3, 3, 3, 2, 2 ], "conv_stride": [ 5, 2, 2, 2, 2, 2, 2 ], "ctc_loss_reduction": "sum", "ctc_zero_infinity": false, "diversity_loss_weight": 0.1, "do_stable_layer_norm": true, "eos_token_id": 2, "feat_extract_activation": "gelu", "feat_extract_dropout": 0.0, "feat_extract_norm": "layer", "feat_proj_dropout": 0.1, "feat_quantizer_dropout": 0.0, "final_dropout": 0.0, "gradient_checkpointing": false, "hidden_act": "gelu", "hidden_dropout": 0.1, "hidden_size": 1024, "initializer_range": 0.02, "intermediate_size": 4096, "layer_norm_eps": 1e-05, "layerdrop": 0.1, "mask_channel_length": 10, "mask_channel_min_space": 1, "mask_channel_other": 0.0, "mask_channel_prob": 0.0, "mask_channel_selection": "static", "mask_feature_length": 10, "mask_feature_prob": 0.0, "mask_time_length": 10, "mask_time_min_space": 1, "mask_time_other": 0.0, "mask_time_prob": 0.075, "mask_time_selection": "static", "model_type": "wav2vec2", "num_attention_heads": 16, "num_codevector_groups": 2, "num_codevectors_per_group": 320, "num_conv_pos_embedding_groups": 16, "num_conv_pos_embeddings": 128, "num_feat_extract_layers": 7, "num_hidden_layers": 24, "num_negatives": 100, "pad_token_id": 0, "proj_codevector_dim": 768, "torch_dtype": "float32", "transformers_version": "4.13.0.dev0", "use_weighted_layer_sum": false, "vocab_size": 392, "id2label": { "1": "", "0": "", "2": "", "3": "", "4": "n", "5": "s", "6": "t", "7": "ə", "8": "l", "9": "a", "10": "i", "11": "k", "12": "d", "13": "m", "14": "ɛ", "15": "ɾ", "16": "e", "17": "ɪ", "18": "p", "19": "o", "20": "ɐ", "21": "z", "22": "ð", "23": "f", "24": "j", "25": "v", "26": "b", "27": "ɹ", "28": "ʁ", "29": "ʊ", "30": "iː", "31": "r", "32": "w", "33": "ʌ", "34": "u", "35": "ɡ", "36": "æ", "37": "aɪ", "38": "ʃ", "39": "h", "40": "ɔ", "41": "ɑː", "42": "ŋ", "43": "ɚ", "44": "eɪ", "45": "β", "46": "uː", "47": "y", "48": "ɑ̃", "49": "oʊ", "50": "ᵻ", "51": "eː", "52": "θ", "53": "aʊ", "54": "ts", "55": "oː", "56": "ɔ̃", "57": "ɣ", "58": "ɜ", "59": "ɑ", "60": "dʒ", "61": "əl", "62": "x", "63": "ɜː", "64": "ç", "65": "ʒ", "66": "tʃ", "67": "ɔː", "68": "ɑːɹ", "69": "ɛ̃", "70": "ʎ", "71": "ɔːɹ", "72": "ʋ", "73": "aː", "74": "ɕ", "75": "œ", "76": "ø", "77": "oːɹ", "78": "ɲ", "79": "yː", "80": "ʔ", "81": "iə", "82": "i5", "83": "s.", "84": "tɕ", "85": "??", "86": "nʲ", "87": "ɛː", "88": "œ̃", "89": "ɭ", "90": "ɔø", "91": "ʑ", "92": "tʲ", "93": "ɨ", "94": "ɛɹ", "95": "ts.", "96": "rʲ", "97": "ɪɹ", "98": "ɭʲ", "99": "i.5", "100": "ɔɪ", "101": "q", "102": "sʲ", "103": "u5", "104": "ʊɹ", "105": "iɜ", "106": "a5", "107": "iɛ5", "108": "øː", "109": "ʕ", "110": "ja", "111": "əɜ", "112": "th", "113": "ɑ5", "114": "oɪ", "115": "dʲ", "116": "ə5", "117": "tɕh", "118": "ts.h", "119": "mʲ", "120": "ɯ", "121": "dʑ", "122": "vʲ", "123": "e̞", "124": "tʃʲ", "125": "ei5", "126": "o5", "127": "onɡ5", "128": "ɑu5", "129": "iɑ5", "130": "ai5", "131": "aɪɚ", "132": "kh", "133": "ə1", "134": "ʐ", "135": "i2", "136": "ʉ", "137": "ħ", "138": "t[", "139": "aɪə", "140": "ʲ", "141": "ju", "142": "ə2", "143": "u2", "144": "oɜ", "145": "pː", "146": "iɛɜ", "147": "ou5", "148": "y5", "149": "uɜ", "150": "tː", "151": "uo5", "152": "d[", "153": "uoɜ", "154": "tsh", "155": "ɑɜ", "156": "ɵ", "157": "i̪5", "158": "uei5", "159": "ɟ", "160": "aɜ", "161": "ɑɨ", "162": "i.ɜ", "163": "eʊ", "164": "o2", "165": "ɐ̃", "166": "ä", "167": "pʲ", "168": "kʲ", "169": "n̩", "170": "ɒ", "171": "ph", "172": "ɑu2", "173": "uɨ", "174": "əɪ", "175": "ɫ", "176": "ɬ", "177": "yɜ", "178": "bʲ", "179": "ɑ2", "180": "s̪", "181": "aiɜ", "182": "χ", "183": "ɐ̃ʊ̃", "184": "1", "185": "ə4", "186": "yæɜ", "187": "a2", "188": "ɨː", "189": "t̪", "190": "iouɜ", "191": "ũ", "192": "onɡɜ", "193": "aɨ", "194": "iɛ2", "195": "ɔɨ", "196": "ɑuɜ", "197": "o̞", "198": "ei2", "199": "iou2", "200": "c", "201": "kː", "202": "y2", "203": "ɖ", "204": "oe", "205": "dˤ", "206": "yɛɜ", "207": "əʊ", "208": "S", "209": "ɡʲ", "210": "onɡ2", "211": "u\"", "212": "eiɜ", "213": "ʈ", "214": "ɯᵝ", "215": "iou5", "216": "dZ", "217": "r̝̊", "218": "i.2", "219": "tS", "220": "s^", "221": "ʝ", "222": "yə5", "223": "iɑɜ", "224": "uə5", "225": "pf", "226": "ɨu", "227": "iɑ2", "228": "ou2", "229": "ər2", "230": "fʲ", "231": "ai2", "232": "r̝", "233": "uəɜ", "234": "ɳ", "235": "əɨ", "236": "ua5", "237": "uɪ", "238": "ɽ", "239": "bː", "240": "yu5", "241": "uo2", "242": "yɛ5", "243": "l̩", "244": "ɻ", "245": "ərɜ", "246": "ʂ", "247": "i̪2", "248": "ouɜ", "249": "uaɜ", "250": "a.", "251": "a.ː", "252": "yæ5", "253": "dː", "254": "r̩", "255": "ee", "256": "ɪu", "257": "ər5", "258": "i̪ɜ", "259": "æi", "260": "u:", "261": "i.ː", "262": "t^", "263": "o1", "264": "ɪ^", "265": "ai", "266": "ueiɜ", "267": "æː", "268": "ɛɪ", "269": "eə", "270": "i.", "271": "ɴ", "272": "ie", "273": "ua2", "274": "ɑ1", "275": "o4", "276": "tʃː", "277": "o:", "278": "ɑ:", "279": "u1", "280": "N", "281": "i̪1", "282": "au", "283": "yæ2", "284": "u.", "285": "qː", "286": "yəɜ", "287": "y:", "288": "kʰ", "289": "tʃʰ", "290": "iʊ", "291": "sx", "292": "õ", "293": "uo", "294": "tʰ", "295": "uai5", "296": "bʰ", "297": "u.ː", "298": "uə2", "299": "ʊə", "300": "d^", "301": "s̪ː", "302": "yiɜ", "303": "dʰ", "304": "r.", "305": "oe:", "306": "i1", "307": "ɟː", "308": "yu2", "309": "nʲʲ", "310": "i̪4", "311": "uei2", "312": "tsʲ", "313": "ɸ", "314": "ĩ", "315": "ɑ4", "316": "t̪ː", "317": "eɑ", "318": "u4", "319": "e:", "320": "tsː", "321": "ʈʰ", "322": "ɡʰ", "323": "ɯɯ", "324": "dʒʲ", "325": "ʂʲ", "326": "X", "327": "ɵː", "328": "uaiɜ", "329": "tɕʲ", "330": "ã", "331": "t^ː", "332": "ẽː", "333": "yɛ2", "334": "cː", "335": "i.1", "336": "ɛʊ", "337": "dˤdˤ", "338": "dʒː", "339": "i4", "340": "ɡː", "341": "yi", "342": "ɕʲ", "343": "ɟʰ", "344": "pʰ", "345": "dʑʲ", "346": "yuɜ", "347": "ua1", "348": "ua4", "349": "æiː", "350": "ɐɐ", "351": "ui", "352": "iou1", "353": "ʊː", "354": "a1", "355": "iou4", "356": "cʰ", "357": "iɛ1", "358": "yə2", "359": "ɖʰ", "360": "ẽ", "361": "ʒʲ", "362": "ää", "363": "ər4", "364": "iːː", "365": "ɪː", "366": "iɑ1", "367": "ər1", "368": "œː", "369": "øi", "370": "ɪuː", "371": "cʰcʰ", "372": "əː1", "373": "iː1", "374": "ũ", "375": "kʰː", "376": "o̞o̞", "377": "xʲ", "378": "ou1", "379": "iɛ4", "380": "e̞e̞", "381": "y1", "382": "dzː", "383": "dʲʲ", "384": "dʰː", "385": "ɯᵝɯᵝ", "386": "lː", "387": "uo1", "388": "i.4", "389": "i:", "390": "yɛ5ʲ", "391": "a4" }, "label2id": { "": 1, "": 0, "": 2, "": 3, "n": 4, "s": 5, "t": 6, "ə": 7, "l": 8, "a": 9, "i": 10, "k": 11, "d": 12, "m": 13, "ɛ": 14, "ɾ": 15, "e": 16, "ɪ": 17, "p": 18, "o": 19, "ɐ": 20, "z": 21, "ð": 22, "f": 23, "j": 24, "v": 25, "b": 26, "ɹ": 27, "ʁ": 28, "ʊ": 29, "iː": 30, "r": 31, "w": 32, "ʌ": 33, "u": 34, "ɡ": 35, "æ": 36, "aɪ": 37, "ʃ": 38, "h": 39, "ɔ": 40, "ɑː": 41, "ŋ": 42, "ɚ": 43, "eɪ": 44, "β": 45, "uː": 46, "y": 47, "ɑ̃": 48, "oʊ": 49, "ᵻ": 50, "eː": 51, "θ": 52, "aʊ": 53, "ts": 54, "oː": 55, "ɔ̃": 56, "ɣ": 57, "ɜ": 58, "ɑ": 59, "dʒ": 60, "əl": 61, "x": 62, "ɜː": 63, "ç": 64, "ʒ": 65, "tʃ": 66, "ɔː": 67, "ɑːɹ": 68, "ɛ̃": 69, "ʎ": 70, "ɔːɹ": 71, "ʋ": 72, "aː": 73, "ɕ": 74, "œ": 75, "ø": 76, "oːɹ": 77, "ɲ": 78, "yː": 79, "ʔ": 80, "iə": 81, "i5": 82, "s.": 83, "tɕ": 84, "??": 85, "nʲ": 86, "ɛː": 87, "œ̃": 88, "ɭ": 89, "ɔø": 90, "ʑ": 91, "tʲ": 92, "ɨ": 93, "ɛɹ": 94, "ts.": 95, "rʲ": 96, "ɪɹ": 97, "ɭʲ": 98, "i.5": 99, "ɔɪ": 100, "q": 101, "sʲ": 102, "u5": 103, "ʊɹ": 104, "iɜ": 105, "a5": 106, "iɛ5": 107, "øː": 108, "ʕ": 109, "ja": 110, "əɜ": 111, "th": 112, "ɑ5": 113, "oɪ": 114, "dʲ": 115, "ə5": 116, "tɕh": 117, "ts.h": 118, "mʲ": 119, "ɯ": 120, "dʑ": 121, "vʲ": 122, "e̞": 123, "tʃʲ": 124, "ei5": 125, "o5": 126, "onɡ5": 127, "ɑu5": 128, "iɑ5": 129, "ai5": 130, "aɪɚ": 131, "kh": 132, "ə1": 133, "ʐ": 134, "i2": 135, "ʉ": 136, "ħ": 137, "t[": 138, "aɪə": 139, "ʲ": 140, "ju": 141, "ə2": 142, "u2": 143, "oɜ": 144, "pː": 145, "iɛɜ": 146, "ou5": 147, "y5": 148, "uɜ": 149, "tː": 150, "uo5": 151, "d[": 152, "uoɜ": 153, "tsh": 154, "ɑɜ": 155, "ɵ": 156, "i̪5": 157, "uei5": 158, "ɟ": 159, "aɜ": 160, "ɑɨ": 161, "i.ɜ": 162, "eʊ": 163, "o2": 164, "ɐ̃": 165, "ä": 166, "pʲ": 167, "kʲ": 168, "n̩": 169, "ɒ": 170, "ph": 171, "ɑu2": 172, "uɨ": 173, "əɪ": 174, "ɫ": 175, "ɬ": 176, "yɜ": 177, "bʲ": 178, "ɑ2": 179, "s̪": 180, "aiɜ": 181, "χ": 182, "ɐ̃ʊ̃": 183, "1": 184, "ə4": 185, "yæɜ": 186, "a2": 187, "ɨː": 188, "t̪": 189, "iouɜ": 190, "ũ": 191, "onɡɜ": 192, "aɨ": 193, "iɛ2": 194, "ɔɨ": 195, "ɑuɜ": 196, "o̞": 197, "ei2": 198, "iou2": 199, "c": 200, "kː": 201, "y2": 202, "ɖ": 203, "oe": 204, "dˤ": 205, "yɛɜ": 206, "əʊ": 207, "S": 208, "ɡʲ": 209, "onɡ2": 210, "u\"": 211, "eiɜ": 212, "ʈ": 213, "ɯᵝ": 214, "iou5": 215, "dZ": 216, "r̝̊": 217, "i.2": 218, "tS": 219, "s^": 220, "ʝ": 221, "yə5": 222, "iɑɜ": 223, "uə5": 224, "pf": 225, "ɨu": 226, "iɑ2": 227, "ou2": 228, "ər2": 229, "fʲ": 230, "ai2": 231, "r̝": 232, "uəɜ": 233, "ɳ": 234, "əɨ": 235, "ua5": 236, "uɪ": 237, "ɽ": 238, "bː": 239, "yu5": 240, "uo2": 241, "yɛ5": 242, "l̩": 243, "ɻ": 244, "ərɜ": 245, "ʂ": 246, "i̪2": 247, "ouɜ": 248, "uaɜ": 249, "a.": 250, "a.ː": 251, "yæ5": 252, "dː": 253, "r̩": 254, "ee": 255, "ɪu": 256, "ər5": 257, "i̪ɜ": 258, "æi": 259, "u:": 260, "i.ː": 261, "t^": 262, "o1": 263, "ɪ^": 264, "ai": 265, "ueiɜ": 266, "æː": 267, "ɛɪ": 268, "eə": 269, "i.": 270, "ɴ": 271, "ie": 272, "ua2": 273, "ɑ1": 274, "o4": 275, "tʃː": 276, "o:": 277, "ɑ:": 278, "u1": 279, "N": 280, "i̪1": 281, "au": 282, "yæ2": 283, "u.": 284, "qː": 285, "yəɜ": 286, "y:": 287, "kʰ": 288, "tʃʰ": 289, "iʊ": 290, "sx": 291, "õ": 292, "uo": 293, "tʰ": 294, "uai5": 295, "bʰ": 296, "u.ː": 297, "uə2": 298, "ʊə": 299, "d^": 300, "s̪ː": 301, "yiɜ": 302, "dʰ": 303, "r.": 304, "oe:": 305, "i1": 306, "ɟː": 307, "yu2": 308, "nʲʲ": 309, "i̪4": 310, "uei2": 311, "tsʲ": 312, "ɸ": 313, "ĩ": 314, "ɑ4": 315, "t̪ː": 316, "eɑ": 317, "u4": 318, "e:": 319, "tsː": 320, "ʈʰ": 321, "ɡʰ": 322, "ɯɯ": 323, "dʒʲ": 324, "ʂʲ": 325, "X": 326, "ɵː": 327, "uaiɜ": 328, "tɕʲ": 329, "ã": 330, "t^ː": 331, "ẽː": 332, "yɛ2": 333, "cː": 334, "i.1": 335, "ɛʊ": 336, "dˤdˤ": 337, "dʒː": 338, "i4": 339, "ɡː": 340, "yi": 341, "ɕʲ": 342, "ɟʰ": 343, "pʰ": 344, "dʑʲ": 345, "yuɜ": 346, "ua1": 347, "ua4": 348, "æiː": 349, "ɐɐ": 350, "ui": 351, "iou1": 352, "ʊː": 353, "a1": 354, "iou4": 355, "cʰ": 356, "iɛ1": 357, "yə2": 358, "ɖʰ": 359, "ẽ": 360, "ʒʲ": 361, "ää": 362, "ər4": 363, "iːː": 364, "ɪː": 365, "iɑ1": 366, "ər1": 367, "œː": 368, "øi": 369, "ɪuː": 370, "cʰcʰ": 371, "əː1": 372, "iː1": 373, "ũ": 374, "kʰː": 375, "o̞o̞": 376, "xʲ": 377, "ou1": 378, "iɛ4": 379, "e̞e̞": 380, "y1": 381, "dzː": 382, "dʲʲ": 383, "dʰː": 384, "ɯᵝɯᵝ": 385, "lː": 386, "uo1": 387, "i.4": 388, "i:": 389, "yɛ5ʲ": 390, "a4": 391 } }