Upload config

config.json

@@ -1,7 +1,8 @@
 {
-  "_name_or_path": "openai/whisper-medium",
+  "_name_or_path": "sanchit-gandhi/whisper-medium-fleurs-lang-id",
   "activation_dropout": 0.0,
   "activation_function": "gelu",
+  "apply_spec_augment": false,
   "architectures": [
     "WhisperForAudioClassification"
   ],
@@ -40,215 +41,221 @@
     ]
   ],
   "id2label": {
-    "0": "af_za",
-    "1": "am_et",
-    "10": "ca_es",
-    "100": "yue_hant_hk",
-    "101": "zu_za",
-    "11": "ceb_ph",
-    "12": "ckb_iq",
-    "13": "cmn_hans_cn",
-    "14": "cs_cz",
-    "15": "cy_gb",
-    "16": "da_dk",
-    "17": "de_de",
-    "18": "el_gr",
-    "19": "en_us",
-    "2": "ar_eg",
-    "20": "es_419",
-    "21": "et_ee",
-    "22": "fa_ir",
-    "23": "ff_sn",
-    "24": "fi_fi",
-    "25": "fil_ph",
-    "26": "fr_fr",
-    "27": "ga_ie",
-    "28": "gl_es",
-    "29": "gu_in",
-    "3": "as_in",
-    "30": "ha_ng",
-    "31": "he_il",
-    "32": "hi_in",
-    "33": "hr_hr",
-    "34": "hu_hu",
-    "35": "hy_am",
-    "36": "id_id",
-    "37": "ig_ng",
-    "38": "is_is",
-    "39": "it_it",
-    "4": "ast_es",
-    "40": "ja_jp",
-    "41": "jv_id",
-    "42": "ka_ge",
-    "43": "kam_ke",
-    "44": "kea_cv",
-    "45": "kk_kz",
-    "46": "km_kh",
-    "47": "kn_in",
-    "48": "ko_kr",
-    "49": "ky_kg",
-    "5": "az_az",
-    "50": "lb_lu",
-    "51": "lg_ug",
-    "52": "ln_cd",
-    "53": "lo_la",
-    "54": "lt_lt",
-    "55": "luo_ke",
-    "56": "lv_lv",
-    "57": "mi_nz",
-    "58": "mk_mk",
-    "59": "ml_in",
-    "6": "be_by",
-    "60": "mn_mn",
-    "61": "mr_in",
-    "62": "ms_my",
-    "63": "mt_mt",
-    "64": "my_mm",
-    "65": "nb_no",
-    "66": "ne_np",
-    "67": "nl_nl",
-    "68": "nso_za",
-    "69": "ny_mw",
-    "7": "bg_bg",
-    "70": "oc_fr",
-    "71": "om_et",
-    "72": "or_in",
-    "73": "pa_in",
-    "74": "pl_pl",
-    "75": "ps_af",
-    "76": "pt_br",
-    "77": "ro_ro",
-    "78": "ru_ru",
-    "79": "sd_in",
-    "8": "bn_in",
-    "80": "sk_sk",
-    "81": "sl_si",
-    "82": "sn_zw",
-    "83": "so_so",
-    "84": "sr_rs",
-    "85": "sv_se",
-    "86": "sw_ke",
-    "87": "ta_in",
-    "88": "te_in",
-    "89": "tg_tj",
-    "9": "bs_ba",
-    "90": "th_th",
-    "91": "tr_tr",
-    "92": "uk_ua",
-    "93": "umb_ao",
-    "94": "ur_pk",
-    "95": "uz_uz",
-    "96": "vi_vn",
-    "97": "wo_sn",
-    "98": "xh_za",
-    "99": "yo_ng"
+    "0": "Afrikaans",
+    "1": "Amharic",
+    "2": "Arabic",
+    "3": "Assamese",
+    "4": "Asturian",
+    "5": "Azerbaijani",
+    "6": "Belarusian",
+    "7": "Bulgarian",
+    "8": "Bengali",
+    "9": "Bosnian",
+    "10": "Catalan",
+    "11": "Cebuano",
+    "12": "Sorani-Kurdish",
+    "13": "Mandarin Chinese",
+    "14": "Czech",
+    "15": "Welsh",
+    "16": "Danish",
+    "17": "German",
+    "18": "Greek",
+    "19": "English",
+    "20": "Spanish",
+    "21": "Estonian",
+    "22": "Persian",
+    "23": "Fula",
+    "24": "Finnish",
+    "25": "Filipino",
+    "26": "French",
+    "27": "Irish",
+    "28": "Galician",
+    "29": "Gujarati",
+    "30": "Hausa",
+    "31": "Hebrew",
+    "32": "Hindi",
+    "33": "Croatian",
+    "34": "Hungarian",
+    "35": "Armenian",
+    "36": "Indonesian",
+    "37": "Igbo",
+    "38": "Icelandic",
+    "39": "Italian",
+    "40": "Japanese",
+    "41": "Javanese",
+    "42": "Georgian",
+    "43": "Kamba",
+    "44": "Kabuverdianu",
+    "45": "Kazakh",
+    "46": "Khmer",
+    "47": "Kannada",
+    "48": "Korean",
+    "49": "Kyrgyz",
+    "50": "Luxembourgish",
+    "51": "Ganda",
+    "52": "Lingala",
+    "53": "Lao",
+    "54": "Lithuanian",
+    "55": "Luo",
+    "56": "Latvian",
+    "57": "Maori",
+    "58": "Macedonian",
+    "59": "Malayalam",
+    "60": "Mongolian",
+    "61": "Marathi",
+    "62": "Malay",
+    "63": "Maltese",
+    "64": "Burmese",
+    "65": "Norwegian",
+    "66": "Nepali",
+    "67": "Dutch",
+    "68": "Northern-Sotho",
+    "69": "Nyanja",
+    "70": "Occitan",
+    "71": "Oromo",
+    "72": "Oriya",
+    "73": "Punjabi",
+    "74": "Polish",
+    "75": "Pashto",
+    "76": "Portuguese",
+    "77": "Romanian",
+    "78": "Russian",
+    "79": "Sindhi",
+    "80": "Slovak",
+    "81": "Slovenian",
+    "82": "Shona",
+    "83": "Somali",
+    "84": "Serbian",
+    "85": "Swedish",
+    "86": "Swahili",
+    "87": "Tamil",
+    "88": "Telugu",
+    "89": "Tajik",
+    "90": "Thai",
+    "91": "Turkish",
+    "92": "Ukrainian",
+    "93": "Umbundu",
+    "94": "Urdu",
+    "95": "Uzbek",
+    "96": "Vietnamese",
+    "97": "Wolof",
+    "98": "Xhosa",
+    "99": "Yoruba",
+    "100": "Cantonese Chinese",
+    "101": "Zulu"
   },
   "init_std": 0.02,
   "is_encoder_decoder": true,
   "label2id": {
-    "af_za": "0",
-    "am_et": "1",
-    "ar_eg": "2",
-    "as_in": "3",
-    "ast_es": "4",
-    "az_az": "5",
-    "be_by": "6",
-    "bg_bg": "7",
-    "bn_in": "8",
-    "bs_ba": "9",
-    "ca_es": "10",
-    "ceb_ph": "11",
-    "ckb_iq": "12",
-    "cmn_hans_cn": "13",
-    "cs_cz": "14",
-    "cy_gb": "15",
-    "da_dk": "16",
-    "de_de": "17",
-    "el_gr": "18",
-    "en_us": "19",
-    "es_419": "20",
-    "et_ee": "21",
-    "fa_ir": "22",
-    "ff_sn": "23",
-    "fi_fi": "24",
-    "fil_ph": "25",
-    "fr_fr": "26",
-    "ga_ie": "27",
-    "gl_es": "28",
-    "gu_in": "29",
-    "ha_ng": "30",
-    "he_il": "31",
-    "hi_in": "32",
-    "hr_hr": "33",
-    "hu_hu": "34",
-    "hy_am": "35",
-    "id_id": "36",
-    "ig_ng": "37",
-    "is_is": "38",
-    "it_it": "39",
-    "ja_jp": "40",
-    "jv_id": "41",
-    "ka_ge": "42",
-    "kam_ke": "43",
-    "kea_cv": "44",
-    "kk_kz": "45",
-    "km_kh": "46",
-    "kn_in": "47",
-    "ko_kr": "48",
-    "ky_kg": "49",
-    "lb_lu": "50",
-    "lg_ug": "51",
-    "ln_cd": "52",
-    "lo_la": "53",
-    "lt_lt": "54",
-    "luo_ke": "55",
-    "lv_lv": "56",
-    "mi_nz": "57",
-    "mk_mk": "58",
-    "ml_in": "59",
-    "mn_mn": "60",
-    "mr_in": "61",
-    "ms_my": "62",
-    "mt_mt": "63",
-    "my_mm": "64",
-    "nb_no": "65",
-    "ne_np": "66",
-    "nl_nl": "67",
-    "nso_za": "68",
-    "ny_mw": "69",
-    "oc_fr": "70",
-    "om_et": "71",
-    "or_in": "72",
-    "pa_in": "73",
-    "pl_pl": "74",
-    "ps_af": "75",
-    "pt_br": "76",
-    "ro_ro": "77",
-    "ru_ru": "78",
-    "sd_in": "79",
-    "sk_sk": "80",
-    "sl_si": "81",
-    "sn_zw": "82",
-    "so_so": "83",
-    "sr_rs": "84",
-    "sv_se": "85",
-    "sw_ke": "86",
-    "ta_in": "87",
-    "te_in": "88",
-    "tg_tj": "89",
-    "th_th": "90",
-    "tr_tr": "91",
-    "uk_ua": "92",
-    "umb_ao": "93",
-    "ur_pk": "94",
-    "uz_uz": "95",
-    "vi_vn": "96",
-    "wo_sn": "97",
-    "xh_za": "98",
-    "yo_ng": "99",
-    "yue_hant_hk": "100",
-    "zu_za": "101"
+    "Afrikaans": "0",
+    "Amharic": "1",
+    "Arabic": "2",
+    "Armenian": "35",
+    "Assamese": "3",
+    "Asturian": "4",
+    "Azerbaijani": "5",
+    "Belarusian": "6",
+    "Bengali": "8",
+    "Bosnian": "9",
+    "Bulgarian": "7",
+    "Burmese": "64",
+    "Cantonese Chinese": "100",
+    "Catalan": "10",
+    "Cebuano": "11",
+    "Croatian": "33",
+    "Czech": "14",
+    "Danish": "16",
+    "Dutch": "67",
+    "English": "19",
+    "Estonian": "21",
+    "Filipino": "25",
+    "Finnish": "24",
+    "French": "26",
+    "Fula": "23",
+    "Galician": "28",
+    "Ganda": "51",
+    "Georgian": "42",
+    "German": "17",
+    "Greek": "18",
+    "Gujarati": "29",
+    "Hausa": "30",
+    "Hebrew": "31",
+    "Hindi": "32",
+    "Hungarian": "34",
+    "Icelandic": "38",
+    "Igbo": "37",
+    "Indonesian": "36",
+    "Irish": "27",
+    "Italian": "39",
+    "Japanese": "40",
+    "Javanese": "41",
+    "Kabuverdianu": "44",
+    "Kamba": "43",
+    "Kannada": "47",
+    "Kazakh": "45",
+    "Khmer": "46",
+    "Korean": "48",
+    "Kyrgyz": "49",
+    "Lao": "53",
+    "Latvian": "56",
+    "Lingala": "52",
+    "Lithuanian": "54",
+    "Luo": "55",
+    "Luxembourgish": "50",
+    "Macedonian": "58",
+    "Malay": "62",
+    "Malayalam": "59",
+    "Maltese": "63",
+    "Mandarin Chinese": "13",
+    "Maori": "57",
+    "Marathi": "61",
+    "Mongolian": "60",
+    "Nepali": "66",
+    "Northern-Sotho": "68",
+    "Norwegian": "65",
+    "Nyanja": "69",
+    "Occitan": "70",
+    "Oriya": "72",
+    "Oromo": "71",
+    "Pashto": "75",
+    "Persian": "22",
+    "Polish": "74",
+    "Portuguese": "76",
+    "Punjabi": "73",
+    "Romanian": "77",
+    "Russian": "78",
+    "Serbian": "84",
+    "Shona": "82",
+    "Sindhi": "79",
+    "Slovak": "80",
+    "Slovenian": "81",
+    "Somali": "83",
+    "Sorani-Kurdish": "12",
+    "Spanish": "20",
+    "Swahili": "86",
+    "Swedish": "85",
+    "Tajik": "89",
+    "Tamil": "87",
+    "Telugu": "88",
+    "Thai": "90",
+    "Turkish": "91",
+    "Ukrainian": "92",
+    "Umbundu": "93",
+    "Urdu": "94",
+    "Uzbek": "95",
+    "Vietnamese": "96",
+    "Welsh": "15",
+    "Wolof": "97",
+    "Xhosa": "98",
+    "Yoruba": "99",
+    "Zulu": "101"
   },
+  "mask_feature_length": 10,
+  "mask_feature_min_masks": 0,
+  "mask_feature_prob": 0.0,
+  "mask_time_length": 10,
+  "mask_time_min_masks": 2,
+  "mask_time_prob": 0.05,
   "max_length": 448,
   "max_source_positions": 1500,
   "max_target_positions": 448,
@@ -346,7 +353,7 @@
     50362
   ],
   "torch_dtype": "float16",
-  "transformers_version": "4.27.0.dev0",
+  "transformers_version": "4.30.0.dev0",
   "use_cache": true,
   "use_weighted_layer_sum": false,
   "vocab_size": 51865