Commit
•
01d9dfe
1
Parent(s):
595bc93
Training in progress, epoch 1
Browse files- .gitignore +1 -0
- config.json +353 -0
- preprocessor_config.json +14 -0
- pytorch_model.bin +3 -0
- runs/Feb23_13-37-14_sanchit-t4-fleurs/1677159503.584576/events.out.tfevents.1677159503.sanchit-t4-fleurs.26232.1 +3 -0
- runs/Feb23_13-37-14_sanchit-t4-fleurs/events.out.tfevents.1677159503.sanchit-t4-fleurs.26232.0 +3 -0
- training_args.bin +3 -0
.gitignore
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
checkpoint-*/
|
config.json
ADDED
@@ -0,0 +1,353 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"_name_or_path": "openai/whisper-medium",
|
3 |
+
"activation_dropout": 0.0,
|
4 |
+
"activation_function": "gelu",
|
5 |
+
"architectures": [
|
6 |
+
"WhisperForAudioClassification"
|
7 |
+
],
|
8 |
+
"attention_dropout": 0.0,
|
9 |
+
"begin_suppress_tokens": [
|
10 |
+
220,
|
11 |
+
50257
|
12 |
+
],
|
13 |
+
"bos_token_id": 50257,
|
14 |
+
"classifier_proj_size": 256,
|
15 |
+
"d_model": 1024,
|
16 |
+
"decoder_attention_heads": 16,
|
17 |
+
"decoder_ffn_dim": 4096,
|
18 |
+
"decoder_layerdrop": 0.0,
|
19 |
+
"decoder_layers": 24,
|
20 |
+
"decoder_start_token_id": 50258,
|
21 |
+
"dropout": 0.0,
|
22 |
+
"encoder_attention_heads": 16,
|
23 |
+
"encoder_ffn_dim": 4096,
|
24 |
+
"encoder_layerdrop": 0.0,
|
25 |
+
"encoder_layers": 24,
|
26 |
+
"eos_token_id": 50257,
|
27 |
+
"finetuning_task": "audio-classification",
|
28 |
+
"forced_decoder_ids": [
|
29 |
+
[
|
30 |
+
1,
|
31 |
+
50259
|
32 |
+
],
|
33 |
+
[
|
34 |
+
2,
|
35 |
+
50359
|
36 |
+
],
|
37 |
+
[
|
38 |
+
3,
|
39 |
+
50363
|
40 |
+
]
|
41 |
+
],
|
42 |
+
"id2label": {
|
43 |
+
"0": "af_za",
|
44 |
+
"1": "am_et",
|
45 |
+
"10": "ca_es",
|
46 |
+
"100": "yue_hant_hk",
|
47 |
+
"101": "zu_za",
|
48 |
+
"11": "ceb_ph",
|
49 |
+
"12": "ckb_iq",
|
50 |
+
"13": "cmn_hans_cn",
|
51 |
+
"14": "cs_cz",
|
52 |
+
"15": "cy_gb",
|
53 |
+
"16": "da_dk",
|
54 |
+
"17": "de_de",
|
55 |
+
"18": "el_gr",
|
56 |
+
"19": "en_us",
|
57 |
+
"2": "ar_eg",
|
58 |
+
"20": "es_419",
|
59 |
+
"21": "et_ee",
|
60 |
+
"22": "fa_ir",
|
61 |
+
"23": "ff_sn",
|
62 |
+
"24": "fi_fi",
|
63 |
+
"25": "fil_ph",
|
64 |
+
"26": "fr_fr",
|
65 |
+
"27": "ga_ie",
|
66 |
+
"28": "gl_es",
|
67 |
+
"29": "gu_in",
|
68 |
+
"3": "as_in",
|
69 |
+
"30": "ha_ng",
|
70 |
+
"31": "he_il",
|
71 |
+
"32": "hi_in",
|
72 |
+
"33": "hr_hr",
|
73 |
+
"34": "hu_hu",
|
74 |
+
"35": "hy_am",
|
75 |
+
"36": "id_id",
|
76 |
+
"37": "ig_ng",
|
77 |
+
"38": "is_is",
|
78 |
+
"39": "it_it",
|
79 |
+
"4": "ast_es",
|
80 |
+
"40": "ja_jp",
|
81 |
+
"41": "jv_id",
|
82 |
+
"42": "ka_ge",
|
83 |
+
"43": "kam_ke",
|
84 |
+
"44": "kea_cv",
|
85 |
+
"45": "kk_kz",
|
86 |
+
"46": "km_kh",
|
87 |
+
"47": "kn_in",
|
88 |
+
"48": "ko_kr",
|
89 |
+
"49": "ky_kg",
|
90 |
+
"5": "az_az",
|
91 |
+
"50": "lb_lu",
|
92 |
+
"51": "lg_ug",
|
93 |
+
"52": "ln_cd",
|
94 |
+
"53": "lo_la",
|
95 |
+
"54": "lt_lt",
|
96 |
+
"55": "luo_ke",
|
97 |
+
"56": "lv_lv",
|
98 |
+
"57": "mi_nz",
|
99 |
+
"58": "mk_mk",
|
100 |
+
"59": "ml_in",
|
101 |
+
"6": "be_by",
|
102 |
+
"60": "mn_mn",
|
103 |
+
"61": "mr_in",
|
104 |
+
"62": "ms_my",
|
105 |
+
"63": "mt_mt",
|
106 |
+
"64": "my_mm",
|
107 |
+
"65": "nb_no",
|
108 |
+
"66": "ne_np",
|
109 |
+
"67": "nl_nl",
|
110 |
+
"68": "nso_za",
|
111 |
+
"69": "ny_mw",
|
112 |
+
"7": "bg_bg",
|
113 |
+
"70": "oc_fr",
|
114 |
+
"71": "om_et",
|
115 |
+
"72": "or_in",
|
116 |
+
"73": "pa_in",
|
117 |
+
"74": "pl_pl",
|
118 |
+
"75": "ps_af",
|
119 |
+
"76": "pt_br",
|
120 |
+
"77": "ro_ro",
|
121 |
+
"78": "ru_ru",
|
122 |
+
"79": "sd_in",
|
123 |
+
"8": "bn_in",
|
124 |
+
"80": "sk_sk",
|
125 |
+
"81": "sl_si",
|
126 |
+
"82": "sn_zw",
|
127 |
+
"83": "so_so",
|
128 |
+
"84": "sr_rs",
|
129 |
+
"85": "sv_se",
|
130 |
+
"86": "sw_ke",
|
131 |
+
"87": "ta_in",
|
132 |
+
"88": "te_in",
|
133 |
+
"89": "tg_tj",
|
134 |
+
"9": "bs_ba",
|
135 |
+
"90": "th_th",
|
136 |
+
"91": "tr_tr",
|
137 |
+
"92": "uk_ua",
|
138 |
+
"93": "umb_ao",
|
139 |
+
"94": "ur_pk",
|
140 |
+
"95": "uz_uz",
|
141 |
+
"96": "vi_vn",
|
142 |
+
"97": "wo_sn",
|
143 |
+
"98": "xh_za",
|
144 |
+
"99": "yo_ng"
|
145 |
+
},
|
146 |
+
"init_std": 0.02,
|
147 |
+
"is_encoder_decoder": true,
|
148 |
+
"label2id": {
|
149 |
+
"af_za": "0",
|
150 |
+
"am_et": "1",
|
151 |
+
"ar_eg": "2",
|
152 |
+
"as_in": "3",
|
153 |
+
"ast_es": "4",
|
154 |
+
"az_az": "5",
|
155 |
+
"be_by": "6",
|
156 |
+
"bg_bg": "7",
|
157 |
+
"bn_in": "8",
|
158 |
+
"bs_ba": "9",
|
159 |
+
"ca_es": "10",
|
160 |
+
"ceb_ph": "11",
|
161 |
+
"ckb_iq": "12",
|
162 |
+
"cmn_hans_cn": "13",
|
163 |
+
"cs_cz": "14",
|
164 |
+
"cy_gb": "15",
|
165 |
+
"da_dk": "16",
|
166 |
+
"de_de": "17",
|
167 |
+
"el_gr": "18",
|
168 |
+
"en_us": "19",
|
169 |
+
"es_419": "20",
|
170 |
+
"et_ee": "21",
|
171 |
+
"fa_ir": "22",
|
172 |
+
"ff_sn": "23",
|
173 |
+
"fi_fi": "24",
|
174 |
+
"fil_ph": "25",
|
175 |
+
"fr_fr": "26",
|
176 |
+
"ga_ie": "27",
|
177 |
+
"gl_es": "28",
|
178 |
+
"gu_in": "29",
|
179 |
+
"ha_ng": "30",
|
180 |
+
"he_il": "31",
|
181 |
+
"hi_in": "32",
|
182 |
+
"hr_hr": "33",
|
183 |
+
"hu_hu": "34",
|
184 |
+
"hy_am": "35",
|
185 |
+
"id_id": "36",
|
186 |
+
"ig_ng": "37",
|
187 |
+
"is_is": "38",
|
188 |
+
"it_it": "39",
|
189 |
+
"ja_jp": "40",
|
190 |
+
"jv_id": "41",
|
191 |
+
"ka_ge": "42",
|
192 |
+
"kam_ke": "43",
|
193 |
+
"kea_cv": "44",
|
194 |
+
"kk_kz": "45",
|
195 |
+
"km_kh": "46",
|
196 |
+
"kn_in": "47",
|
197 |
+
"ko_kr": "48",
|
198 |
+
"ky_kg": "49",
|
199 |
+
"lb_lu": "50",
|
200 |
+
"lg_ug": "51",
|
201 |
+
"ln_cd": "52",
|
202 |
+
"lo_la": "53",
|
203 |
+
"lt_lt": "54",
|
204 |
+
"luo_ke": "55",
|
205 |
+
"lv_lv": "56",
|
206 |
+
"mi_nz": "57",
|
207 |
+
"mk_mk": "58",
|
208 |
+
"ml_in": "59",
|
209 |
+
"mn_mn": "60",
|
210 |
+
"mr_in": "61",
|
211 |
+
"ms_my": "62",
|
212 |
+
"mt_mt": "63",
|
213 |
+
"my_mm": "64",
|
214 |
+
"nb_no": "65",
|
215 |
+
"ne_np": "66",
|
216 |
+
"nl_nl": "67",
|
217 |
+
"nso_za": "68",
|
218 |
+
"ny_mw": "69",
|
219 |
+
"oc_fr": "70",
|
220 |
+
"om_et": "71",
|
221 |
+
"or_in": "72",
|
222 |
+
"pa_in": "73",
|
223 |
+
"pl_pl": "74",
|
224 |
+
"ps_af": "75",
|
225 |
+
"pt_br": "76",
|
226 |
+
"ro_ro": "77",
|
227 |
+
"ru_ru": "78",
|
228 |
+
"sd_in": "79",
|
229 |
+
"sk_sk": "80",
|
230 |
+
"sl_si": "81",
|
231 |
+
"sn_zw": "82",
|
232 |
+
"so_so": "83",
|
233 |
+
"sr_rs": "84",
|
234 |
+
"sv_se": "85",
|
235 |
+
"sw_ke": "86",
|
236 |
+
"ta_in": "87",
|
237 |
+
"te_in": "88",
|
238 |
+
"tg_tj": "89",
|
239 |
+
"th_th": "90",
|
240 |
+
"tr_tr": "91",
|
241 |
+
"uk_ua": "92",
|
242 |
+
"umb_ao": "93",
|
243 |
+
"ur_pk": "94",
|
244 |
+
"uz_uz": "95",
|
245 |
+
"vi_vn": "96",
|
246 |
+
"wo_sn": "97",
|
247 |
+
"xh_za": "98",
|
248 |
+
"yo_ng": "99",
|
249 |
+
"yue_hant_hk": "100",
|
250 |
+
"zu_za": "101"
|
251 |
+
},
|
252 |
+
"max_length": 448,
|
253 |
+
"max_source_positions": 1500,
|
254 |
+
"max_target_positions": 448,
|
255 |
+
"model_type": "whisper",
|
256 |
+
"num_hidden_layers": 24,
|
257 |
+
"num_mel_bins": 80,
|
258 |
+
"pad_token_id": 50257,
|
259 |
+
"scale_embedding": false,
|
260 |
+
"suppress_tokens": [
|
261 |
+
1,
|
262 |
+
2,
|
263 |
+
7,
|
264 |
+
8,
|
265 |
+
9,
|
266 |
+
10,
|
267 |
+
14,
|
268 |
+
25,
|
269 |
+
26,
|
270 |
+
27,
|
271 |
+
28,
|
272 |
+
29,
|
273 |
+
31,
|
274 |
+
58,
|
275 |
+
59,
|
276 |
+
60,
|
277 |
+
61,
|
278 |
+
62,
|
279 |
+
63,
|
280 |
+
90,
|
281 |
+
91,
|
282 |
+
92,
|
283 |
+
93,
|
284 |
+
359,
|
285 |
+
503,
|
286 |
+
522,
|
287 |
+
542,
|
288 |
+
873,
|
289 |
+
893,
|
290 |
+
902,
|
291 |
+
918,
|
292 |
+
922,
|
293 |
+
931,
|
294 |
+
1350,
|
295 |
+
1853,
|
296 |
+
1982,
|
297 |
+
2460,
|
298 |
+
2627,
|
299 |
+
3246,
|
300 |
+
3253,
|
301 |
+
3268,
|
302 |
+
3536,
|
303 |
+
3846,
|
304 |
+
3961,
|
305 |
+
4183,
|
306 |
+
4667,
|
307 |
+
6585,
|
308 |
+
6647,
|
309 |
+
7273,
|
310 |
+
9061,
|
311 |
+
9383,
|
312 |
+
10428,
|
313 |
+
10929,
|
314 |
+
11938,
|
315 |
+
12033,
|
316 |
+
12331,
|
317 |
+
12562,
|
318 |
+
13793,
|
319 |
+
14157,
|
320 |
+
14635,
|
321 |
+
15265,
|
322 |
+
15618,
|
323 |
+
16553,
|
324 |
+
16604,
|
325 |
+
18362,
|
326 |
+
18956,
|
327 |
+
20075,
|
328 |
+
21675,
|
329 |
+
22520,
|
330 |
+
26130,
|
331 |
+
26161,
|
332 |
+
26435,
|
333 |
+
28279,
|
334 |
+
29464,
|
335 |
+
31650,
|
336 |
+
32302,
|
337 |
+
32470,
|
338 |
+
36865,
|
339 |
+
42863,
|
340 |
+
47425,
|
341 |
+
49870,
|
342 |
+
50254,
|
343 |
+
50258,
|
344 |
+
50360,
|
345 |
+
50361,
|
346 |
+
50362
|
347 |
+
],
|
348 |
+
"torch_dtype": "float16",
|
349 |
+
"transformers_version": "4.27.0.dev0",
|
350 |
+
"use_cache": true,
|
351 |
+
"use_weighted_layer_sum": false,
|
352 |
+
"vocab_size": 51865
|
353 |
+
}
|
preprocessor_config.json
ADDED
@@ -0,0 +1,14 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"chunk_length": 30,
|
3 |
+
"feature_extractor_type": "WhisperFeatureExtractor",
|
4 |
+
"feature_size": 80,
|
5 |
+
"hop_length": 160,
|
6 |
+
"n_fft": 400,
|
7 |
+
"n_samples": 480000,
|
8 |
+
"nb_max_frames": 3000,
|
9 |
+
"padding_side": "right",
|
10 |
+
"padding_value": 0.0,
|
11 |
+
"processor_class": "WhisperProcessor",
|
12 |
+
"return_attention_mask": false,
|
13 |
+
"sampling_rate": 16000
|
14 |
+
}
|
pytorch_model.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:3630bd8791baa8f7c8d725ebb8a257ba32420c7741a9fe6448826cb2effea411
|
3 |
+
size 615058493
|
runs/Feb23_13-37-14_sanchit-t4-fleurs/1677159503.584576/events.out.tfevents.1677159503.sanchit-t4-fleurs.26232.1
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:da751da5bf36be59098df00add531f0e8e31baf9fd0f8c91d17c5a67691b091f
|
3 |
+
size 5840
|
runs/Feb23_13-37-14_sanchit-t4-fleurs/events.out.tfevents.1677159503.sanchit-t4-fleurs.26232.0
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:9ac8cb4dd3c6de5e0fece304cc822f5c19e5897e8bad8a23610fbb244d01f929
|
3 |
+
size 62898
|
training_args.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:056cee6adb1b2f1fcd2f38aa61d20cb381ad85614636d6b75ea4483a58612531
|
3 |
+
size 4731
|