luigisaetta commited on
Commit
094c6df
·
1 Parent(s): 94f7ef5

Training in progress, step 300

Browse files
fine-tune-atco2-non-streaming.ipynb CHANGED
@@ -2,7 +2,7 @@
2
  "cells": [
3
  {
4
  "cell_type": "markdown",
5
- "id": "ce1be10c",
6
  "metadata": {
7
  "id": "75b58048-7d14-4fc6-8085-1fc08c81b4a6"
8
  },
@@ -12,7 +12,7 @@
12
  },
13
  {
14
  "cell_type": "markdown",
15
- "id": "438ffd28",
16
  "metadata": {
17
  "id": "fbfa8ad5-4cdc-4512-9058-836cbbf65e1a"
18
  },
@@ -25,7 +25,7 @@
25
  },
26
  {
27
  "cell_type": "markdown",
28
- "id": "ca7b1c86",
29
  "metadata": {
30
  "id": "afe0d503-ae4e-4aa7-9af4-dbcba52db41e"
31
  },
@@ -35,7 +35,7 @@
35
  },
36
  {
37
  "cell_type": "markdown",
38
- "id": "9e151483",
39
  "metadata": {
40
  "id": "9ae91ed4-9c3e-4ade-938e-f4c2dcfbfdc0"
41
  },
@@ -67,7 +67,7 @@
67
  },
68
  {
69
  "cell_type": "markdown",
70
- "id": "bcd1e2ee",
71
  "metadata": {
72
  "id": "e59b91d6-be24-4b5e-bb38-4977ea143a72"
73
  },
@@ -86,7 +86,7 @@
86
  },
87
  {
88
  "cell_type": "markdown",
89
- "id": "65434750",
90
  "metadata": {
91
  "id": "21b6316e-8a55-4549-a154-66d3da2ab74a"
92
  },
@@ -115,7 +115,7 @@
115
  },
116
  {
117
  "cell_type": "markdown",
118
- "id": "4d476a78",
119
  "metadata": {
120
  "id": "3a680dfc-cbba-4f6c-8a1f-e1a5ff3f123a"
121
  },
@@ -127,7 +127,7 @@
127
  },
128
  {
129
  "cell_type": "markdown",
130
- "id": "6fb0f9c2",
131
  "metadata": {
132
  "id": "b219c9dd-39b6-4a95-b2a1-3f547a1e7bc0"
133
  },
@@ -137,7 +137,7 @@
137
  },
138
  {
139
  "cell_type": "markdown",
140
- "id": "b33f2737",
141
  "metadata": {
142
  "id": "674429c5-0ab4-4adf-975b-621bb69eca38"
143
  },
@@ -155,7 +155,7 @@
155
  {
156
  "cell_type": "code",
157
  "execution_count": 1,
158
- "id": "9f64f58b",
159
  "metadata": {
160
  "id": "a2787582-554f-44ce-9f38-4180a5ed6b44"
161
  },
@@ -171,7 +171,7 @@
171
  {
172
  "cell_type": "code",
173
  "execution_count": 2,
174
- "id": "7d3c2c02",
175
  "metadata": {},
176
  "outputs": [
177
  {
@@ -212,7 +212,7 @@
212
  },
213
  {
214
  "cell_type": "markdown",
215
- "id": "c21cb7d5",
216
  "metadata": {
217
  "id": "d5c7c3d6-7197-41e7-a088-49b753c1681f"
218
  },
@@ -227,7 +227,7 @@
227
  {
228
  "cell_type": "code",
229
  "execution_count": 3,
230
- "id": "3647bc36",
231
  "metadata": {
232
  "id": "20ba635d-518c-47ac-97ee-3cad25f1e0ce"
233
  },
@@ -257,7 +257,7 @@
257
  },
258
  {
259
  "cell_type": "markdown",
260
- "id": "c3f6227d",
261
  "metadata": {
262
  "id": "2d63b2d2-f68a-4d74-b7f1-5127f6d16605"
263
  },
@@ -267,7 +267,7 @@
267
  },
268
  {
269
  "cell_type": "markdown",
270
- "id": "66102d53",
271
  "metadata": {
272
  "id": "601c3099-1026-439e-93e2-5635b3ba5a73"
273
  },
@@ -287,7 +287,7 @@
287
  },
288
  {
289
  "cell_type": "markdown",
290
- "id": "2c57b7e0",
291
  "metadata": {
292
  "id": "560332eb-3558-41a1-b500-e83a9f695f84"
293
  },
@@ -297,7 +297,7 @@
297
  },
298
  {
299
  "cell_type": "markdown",
300
- "id": "d1d6a7b9",
301
  "metadata": {
302
  "id": "32ec8068-0bd7-412d-b662-0edb9d1e7365"
303
  },
@@ -309,7 +309,7 @@
309
  },
310
  {
311
  "cell_type": "markdown",
312
- "id": "a1aeaab5",
313
  "metadata": {
314
  "id": "589d9ec1-d12b-4b64-93f7-04c63997da19"
315
  },
@@ -324,7 +324,7 @@
324
  },
325
  {
326
  "cell_type": "markdown",
327
- "id": "15e15263",
328
  "metadata": {
329
  "id": "b2ef54d5-b946-4c1d-9fdc-adc5d01b46aa"
330
  },
@@ -335,7 +335,7 @@
335
  {
336
  "cell_type": "code",
337
  "execution_count": 4,
338
- "id": "2b45e282",
339
  "metadata": {
340
  "id": "bc77d7bb-f9e2-47f5-b663-30f7a4321ce5"
341
  },
@@ -348,7 +348,7 @@
348
  },
349
  {
350
  "cell_type": "markdown",
351
- "id": "40180db6",
352
  "metadata": {
353
  "id": "93748af7-b917-4ecf-a0c8-7d89077ff9cb"
354
  },
@@ -358,7 +358,7 @@
358
  },
359
  {
360
  "cell_type": "markdown",
361
- "id": "9a0ebaaa",
362
  "metadata": {
363
  "id": "2bc82609-a9fb-447a-a2af-99597c864029"
364
  },
@@ -372,7 +372,7 @@
372
  {
373
  "cell_type": "code",
374
  "execution_count": 5,
375
- "id": "399e8a0b",
376
  "metadata": {
377
  "id": "c7b07f9b-ae0e-4f89-98f0-0c50d432eab6",
378
  "outputId": "5c004b44-86e7-4e00-88be-39e0af5eed69"
@@ -386,7 +386,7 @@
386
  },
387
  {
388
  "cell_type": "markdown",
389
- "id": "0139da6c",
390
  "metadata": {
391
  "id": "d2ef23f3-f4a8-483a-a2dc-080a7496cb1b"
392
  },
@@ -396,7 +396,7 @@
396
  },
397
  {
398
  "cell_type": "markdown",
399
- "id": "74f25c66",
400
  "metadata": {
401
  "id": "5ff67654-5a29-4bb8-a69d-0228946c6f8d"
402
  },
@@ -412,7 +412,7 @@
412
  {
413
  "cell_type": "code",
414
  "execution_count": 6,
415
- "id": "9d176416",
416
  "metadata": {
417
  "id": "77d9f0c5-8607-4642-a8ac-c3ab2e223ea6"
418
  },
@@ -425,7 +425,7 @@
425
  },
426
  {
427
  "cell_type": "markdown",
428
- "id": "2d342376",
429
  "metadata": {
430
  "id": "381acd09-0b0f-4d04-9eb3-f028ac0e5f2c"
431
  },
@@ -435,7 +435,7 @@
435
  },
436
  {
437
  "cell_type": "markdown",
438
- "id": "4b043dd0",
439
  "metadata": {
440
  "id": "9649bf01-2e8a-45e5-8fca-441c13637b8f"
441
  },
@@ -447,7 +447,7 @@
447
  {
448
  "cell_type": "code",
449
  "execution_count": 7,
450
- "id": "75dd00a9",
451
  "metadata": {
452
  "id": "6e6b0ec5-0c94-4e2c-ae24-c791be1b2255"
453
  },
@@ -467,7 +467,7 @@
467
  },
468
  {
469
  "cell_type": "markdown",
470
- "id": "943f8968",
471
  "metadata": {
472
  "id": "5a679f05-063d-41b3-9b58-4fc9c6ccf4fd"
473
  },
@@ -486,7 +486,7 @@
486
  {
487
  "cell_type": "code",
488
  "execution_count": 8,
489
- "id": "2e8cd7e2",
490
  "metadata": {
491
  "id": "f12e2e57-156f-417b-8cfb-69221cc198e8"
492
  },
@@ -499,7 +499,7 @@
499
  },
500
  {
501
  "cell_type": "markdown",
502
- "id": "f3f7cf49",
503
  "metadata": {
504
  "id": "00382a3e-abec-4cdd-a54c-d1aaa3ea4707"
505
  },
@@ -511,7 +511,7 @@
511
  {
512
  "cell_type": "code",
513
  "execution_count": 9,
514
- "id": "82e4692d",
515
  "metadata": {
516
  "id": "87122d71-289a-466a-afcf-fa354b18946b"
517
  },
@@ -531,7 +531,7 @@
531
  },
532
  {
533
  "cell_type": "markdown",
534
- "id": "cbcf9e3c",
535
  "metadata": {},
536
  "source": [
537
  "We'll define our pre-processing strategy. We advise that you **do not** lower-case the transcriptions or remove punctuation unless mixing different datasets. This will enable you to fine-tune Whisper models that can predict punctuation and casing. Later, you will see how we can evaluate the predictions without punctuation or casing, so that the models benefit from the WER improvement obtained by normalising the transcriptions while still predicting fully formatted transcriptions."
@@ -540,7 +540,7 @@
540
  {
541
  "cell_type": "code",
542
  "execution_count": 10,
543
- "id": "be5abb8f",
544
  "metadata": {},
545
  "outputs": [],
546
  "source": [
@@ -554,7 +554,7 @@
554
  },
555
  {
556
  "cell_type": "markdown",
557
- "id": "9d0bd886",
558
  "metadata": {},
559
  "source": [
560
  "Now we can write a function to prepare our data ready for the model:\n",
@@ -567,7 +567,7 @@
567
  {
568
  "cell_type": "code",
569
  "execution_count": 11,
570
- "id": "b00578f6",
571
  "metadata": {},
572
  "outputs": [],
573
  "source": [
@@ -594,7 +594,7 @@
594
  },
595
  {
596
  "cell_type": "markdown",
597
- "id": "906d9efe",
598
  "metadata": {
599
  "id": "70b319fb-2439-4ef6-a70d-a47bf41c4a13"
600
  },
@@ -605,7 +605,7 @@
605
  {
606
  "cell_type": "code",
607
  "execution_count": 12,
608
- "id": "69066e5c",
609
  "metadata": {
610
  "id": "7b73ab39-ffaf-4b9e-86e5-782963c6134b"
611
  },
@@ -841,7 +841,7 @@
841
  },
842
  {
843
  "cell_type": "markdown",
844
- "id": "21198ebb",
845
  "metadata": {},
846
  "source": [
847
  "Finally, we filter any training data with audio samples longer than 30s. These samples would otherwise be truncated by the Whisper feature-extractor which could affect the stability of training. We define a function that returns `True` for samples that are less than 30s, and `False` for those that are longer:"
@@ -850,7 +850,7 @@
850
  {
851
  "cell_type": "code",
852
  "execution_count": 13,
853
- "id": "470c6b79",
854
  "metadata": {},
855
  "outputs": [],
856
  "source": [
@@ -862,7 +862,7 @@
862
  },
863
  {
864
  "cell_type": "markdown",
865
- "id": "b4e60e66",
866
  "metadata": {},
867
  "source": [
868
  "We apply our filter function to all samples of our training dataset through 🤗 Datasets' `.filter` method:"
@@ -871,7 +871,7 @@
871
  {
872
  "cell_type": "code",
873
  "execution_count": 14,
874
- "id": "6d4b19d2",
875
  "metadata": {},
876
  "outputs": [
877
  {
@@ -891,7 +891,7 @@
891
  },
892
  {
893
  "cell_type": "markdown",
894
- "id": "18f1ad42",
895
  "metadata": {
896
  "id": "263a5a58-0239-4a25-b0df-c625fc9c5810"
897
  },
@@ -901,7 +901,7 @@
901
  },
902
  {
903
  "cell_type": "markdown",
904
- "id": "9239da19",
905
  "metadata": {
906
  "id": "a693e768-c5a6-453f-89a1-b601dcf7daf7"
907
  },
@@ -924,7 +924,7 @@
924
  },
925
  {
926
  "cell_type": "markdown",
927
- "id": "df812100",
928
  "metadata": {
929
  "id": "8d230e6d-624c-400a-bbf5-fa660881df25"
930
  },
@@ -934,7 +934,7 @@
934
  },
935
  {
936
  "cell_type": "markdown",
937
- "id": "c268efce",
938
  "metadata": {
939
  "id": "04def221-0637-4a69-b242-d3f0c1d0ee78"
940
  },
@@ -960,7 +960,7 @@
960
  {
961
  "cell_type": "code",
962
  "execution_count": 15,
963
- "id": "24a62d76",
964
  "metadata": {
965
  "id": "8326221e-ec13-4731-bb4e-51e5fc1486c5"
966
  },
@@ -1001,7 +1001,7 @@
1001
  },
1002
  {
1003
  "cell_type": "markdown",
1004
- "id": "d3606cd8",
1005
  "metadata": {
1006
  "id": "3cae7dbf-8a50-456e-a3a8-7fd005390f86"
1007
  },
@@ -1012,7 +1012,7 @@
1012
  {
1013
  "cell_type": "code",
1014
  "execution_count": 16,
1015
- "id": "56477a6b",
1016
  "metadata": {
1017
  "id": "fc834702-c0d3-4a96-b101-7b87be32bf42"
1018
  },
@@ -1023,7 +1023,7 @@
1023
  },
1024
  {
1025
  "cell_type": "markdown",
1026
- "id": "2eafb33d",
1027
  "metadata": {
1028
  "id": "d62bb2ab-750a-45e7-82e9-61d6f4805698"
1029
  },
@@ -1033,7 +1033,7 @@
1033
  },
1034
  {
1035
  "cell_type": "markdown",
1036
- "id": "10a00fa5",
1037
  "metadata": {
1038
  "id": "66fee1a7-a44c-461e-b047-c3917221572e"
1039
  },
@@ -1045,7 +1045,7 @@
1045
  {
1046
  "cell_type": "code",
1047
  "execution_count": 17,
1048
- "id": "0eaf9eff",
1049
  "metadata": {
1050
  "id": "b22b4011-f31f-4b57-b684-c52332f92890"
1051
  },
@@ -1058,7 +1058,7 @@
1058
  },
1059
  {
1060
  "cell_type": "markdown",
1061
- "id": "b4d8b5f7",
1062
  "metadata": {
1063
  "id": "4f32cab6-31f0-4cb9-af4c-40ba0f5fc508"
1064
  },
@@ -1078,7 +1078,7 @@
1078
  {
1079
  "cell_type": "code",
1080
  "execution_count": 18,
1081
- "id": "e50e5f03",
1082
  "metadata": {
1083
  "id": "23959a70-22d0-4ffe-9fa1-72b61e75bb52"
1084
  },
@@ -1109,7 +1109,7 @@
1109
  },
1110
  {
1111
  "cell_type": "markdown",
1112
- "id": "8c95f83f",
1113
  "metadata": {
1114
  "id": "daf2a825-6d9f-4a23-b145-c37c0039075b"
1115
  },
@@ -1119,7 +1119,7 @@
1119
  },
1120
  {
1121
  "cell_type": "markdown",
1122
- "id": "6322c19d",
1123
  "metadata": {
1124
  "id": "437a97fa-4864-476b-8abc-f28b8166cfa5"
1125
  },
@@ -1131,7 +1131,7 @@
1131
  {
1132
  "cell_type": "code",
1133
  "execution_count": 19,
1134
- "id": "0d49936d",
1135
  "metadata": {
1136
  "id": "5a10cc4b-07ec-4ebd-ac1d-7c601023594f"
1137
  },
@@ -1144,7 +1144,7 @@
1144
  },
1145
  {
1146
  "cell_type": "markdown",
1147
- "id": "6848f5ce",
1148
  "metadata": {
1149
  "id": "a15ead5f-2277-4a39-937b-585c2497b2df"
1150
  },
@@ -1155,7 +1155,7 @@
1155
  {
1156
  "cell_type": "code",
1157
  "execution_count": 20,
1158
- "id": "1e3272a3",
1159
  "metadata": {
1160
  "id": "62038ba3-88ed-4fce-84db-338f50dcd04f"
1161
  },
@@ -1168,7 +1168,7 @@
1168
  },
1169
  {
1170
  "cell_type": "markdown",
1171
- "id": "96f83b5f",
1172
  "metadata": {
1173
  "id": "2178dea4-80ca-47b6-b6ea-ba1915c90c06"
1174
  },
@@ -1178,7 +1178,7 @@
1178
  },
1179
  {
1180
  "cell_type": "markdown",
1181
- "id": "f659e225",
1182
  "metadata": {
1183
  "id": "c21af1e9-0188-4134-ac82-defc7bdcc436"
1184
  },
@@ -1189,7 +1189,7 @@
1189
  {
1190
  "cell_type": "code",
1191
  "execution_count": 21,
1192
- "id": "cb5cdcfe",
1193
  "metadata": {
1194
  "id": "0ae3e9af-97b7-4aa0-ae85-20b23b5bcb3a"
1195
  },
@@ -1223,7 +1223,7 @@
1223
  },
1224
  {
1225
  "cell_type": "markdown",
1226
- "id": "d573456c",
1227
  "metadata": {
1228
  "id": "b3a944d8-3112-4552-82a0-be25988b3857"
1229
  },
@@ -1234,7 +1234,7 @@
1234
  },
1235
  {
1236
  "cell_type": "markdown",
1237
- "id": "757baf62",
1238
  "metadata": {
1239
  "id": "bac29114-d226-4f54-97cf-8718c9f94e1e"
1240
  },
@@ -1246,7 +1246,7 @@
1246
  {
1247
  "cell_type": "code",
1248
  "execution_count": 22,
1249
- "id": "3703befd",
1250
  "metadata": {
1251
  "id": "d546d7fe-0543-479a-b708-2ebabec19493"
1252
  },
@@ -1277,7 +1277,7 @@
1277
  },
1278
  {
1279
  "cell_type": "markdown",
1280
- "id": "44747a53",
1281
  "metadata": {
1282
  "id": "uOrRhDGtN5S4"
1283
  },
@@ -1288,7 +1288,7 @@
1288
  {
1289
  "cell_type": "code",
1290
  "execution_count": 23,
1291
- "id": "622349f3",
1292
  "metadata": {
1293
  "id": "-2zQwMfEOBJq"
1294
  },
@@ -1310,7 +1310,7 @@
1310
  },
1311
  {
1312
  "cell_type": "markdown",
1313
- "id": "bc26d1e9",
1314
  "metadata": {
1315
  "id": "7f404cf9-4345-468c-8196-4bd101d9bd51"
1316
  },
@@ -1320,7 +1320,7 @@
1320
  },
1321
  {
1322
  "cell_type": "markdown",
1323
- "id": "02415746",
1324
  "metadata": {
1325
  "id": "5e8b8d56-5a70-4f68-bd2e-f0752d0bd112"
1326
  },
@@ -1337,7 +1337,7 @@
1337
  {
1338
  "cell_type": "code",
1339
  "execution_count": null,
1340
- "id": "8b188180",
1341
  "metadata": {
1342
  "id": "ee8b7b8e-1c9a-4d77-9137-1778a629e6de"
1343
  },
@@ -1365,8 +1365,8 @@
1365
  "\n",
1366
  " <div>\n",
1367
  " \n",
1368
- " <progress value='201' max='500' style='width:300px; height:20px; vertical-align: middle;'></progress>\n",
1369
- " [201/500 1:09:30 < 1:44:26, 0.05 it/s, Epoch 4.25/11]\n",
1370
  " </div>\n",
1371
  " <table border=\"1\" class=\"dataframe\">\n",
1372
  " <thead>\n",
@@ -1402,6 +1402,18 @@
1402
  " <td>0.577473</td>\n",
1403
  " <td>19.811321</td>\n",
1404
  " </tr>\n",
 
 
 
 
 
 
 
 
 
 
 
 
1405
  " </tbody>\n",
1406
  "</table><p>"
1407
  ],
@@ -1441,6 +1453,19 @@
1441
  "Configuration saved in ./checkpoint-200/config.json\n",
1442
  "Model weights saved in ./checkpoint-200/pytorch_model.bin\n",
1443
  "Feature extractor saved in ./checkpoint-200/preprocessor_config.json\n",
 
 
 
 
 
 
 
 
 
 
 
 
 
1444
  "Feature extractor saved in ./preprocessor_config.json\n"
1445
  ]
1446
  }
@@ -1452,7 +1477,7 @@
1452
  {
1453
  "cell_type": "code",
1454
  "execution_count": null,
1455
- "id": "9d0664a8",
1456
  "metadata": {
1457
  "id": "c704f91e-241b-48c9-b8e0-f0da396a9663"
1458
  },
@@ -1470,7 +1495,7 @@
1470
  },
1471
  {
1472
  "cell_type": "markdown",
1473
- "id": "13e3ddd8",
1474
  "metadata": {
1475
  "id": "090d676a-f944-4297-a938-a40eda0b2b68"
1476
  },
@@ -1481,7 +1506,7 @@
1481
  {
1482
  "cell_type": "code",
1483
  "execution_count": null,
1484
- "id": "d8cf3a71",
1485
  "metadata": {
1486
  "id": "d7030622-caf7-4039-939b-6195cdaa2585"
1487
  },
@@ -1492,7 +1517,7 @@
1492
  },
1493
  {
1494
  "cell_type": "markdown",
1495
- "id": "f378904c",
1496
  "metadata": {
1497
  "id": "ca743fbd-602c-48d4-ba8d-a2fe60af64ba"
1498
  },
@@ -1502,7 +1527,7 @@
1502
  },
1503
  {
1504
  "cell_type": "markdown",
1505
- "id": "0a4c5fb9",
1506
  "metadata": {
1507
  "id": "7f737783-2870-4e35-aa11-86a42d7d997a"
1508
  },
 
2
  "cells": [
3
  {
4
  "cell_type": "markdown",
5
+ "id": "8308975c",
6
  "metadata": {
7
  "id": "75b58048-7d14-4fc6-8085-1fc08c81b4a6"
8
  },
 
12
  },
13
  {
14
  "cell_type": "markdown",
15
+ "id": "002e5c77",
16
  "metadata": {
17
  "id": "fbfa8ad5-4cdc-4512-9058-836cbbf65e1a"
18
  },
 
25
  },
26
  {
27
  "cell_type": "markdown",
28
+ "id": "293972e5",
29
  "metadata": {
30
  "id": "afe0d503-ae4e-4aa7-9af4-dbcba52db41e"
31
  },
 
35
  },
36
  {
37
  "cell_type": "markdown",
38
+ "id": "c197e7a9",
39
  "metadata": {
40
  "id": "9ae91ed4-9c3e-4ade-938e-f4c2dcfbfdc0"
41
  },
 
67
  },
68
  {
69
  "cell_type": "markdown",
70
+ "id": "3e9d693e",
71
  "metadata": {
72
  "id": "e59b91d6-be24-4b5e-bb38-4977ea143a72"
73
  },
 
86
  },
87
  {
88
  "cell_type": "markdown",
89
+ "id": "e5d2f792",
90
  "metadata": {
91
  "id": "21b6316e-8a55-4549-a154-66d3da2ab74a"
92
  },
 
115
  },
116
  {
117
  "cell_type": "markdown",
118
+ "id": "0bd16c3d",
119
  "metadata": {
120
  "id": "3a680dfc-cbba-4f6c-8a1f-e1a5ff3f123a"
121
  },
 
127
  },
128
  {
129
  "cell_type": "markdown",
130
+ "id": "38d1e3d5",
131
  "metadata": {
132
  "id": "b219c9dd-39b6-4a95-b2a1-3f547a1e7bc0"
133
  },
 
137
  },
138
  {
139
  "cell_type": "markdown",
140
+ "id": "fb6a5797",
141
  "metadata": {
142
  "id": "674429c5-0ab4-4adf-975b-621bb69eca38"
143
  },
 
155
  {
156
  "cell_type": "code",
157
  "execution_count": 1,
158
+ "id": "de423e45",
159
  "metadata": {
160
  "id": "a2787582-554f-44ce-9f38-4180a5ed6b44"
161
  },
 
171
  {
172
  "cell_type": "code",
173
  "execution_count": 2,
174
+ "id": "8f4df3c7",
175
  "metadata": {},
176
  "outputs": [
177
  {
 
212
  },
213
  {
214
  "cell_type": "markdown",
215
+ "id": "998a5233",
216
  "metadata": {
217
  "id": "d5c7c3d6-7197-41e7-a088-49b753c1681f"
218
  },
 
227
  {
228
  "cell_type": "code",
229
  "execution_count": 3,
230
+ "id": "d834bfb6",
231
  "metadata": {
232
  "id": "20ba635d-518c-47ac-97ee-3cad25f1e0ce"
233
  },
 
257
  },
258
  {
259
  "cell_type": "markdown",
260
+ "id": "589c1cf1",
261
  "metadata": {
262
  "id": "2d63b2d2-f68a-4d74-b7f1-5127f6d16605"
263
  },
 
267
  },
268
  {
269
  "cell_type": "markdown",
270
+ "id": "92e9bda0",
271
  "metadata": {
272
  "id": "601c3099-1026-439e-93e2-5635b3ba5a73"
273
  },
 
287
  },
288
  {
289
  "cell_type": "markdown",
290
+ "id": "e8f68ff5",
291
  "metadata": {
292
  "id": "560332eb-3558-41a1-b500-e83a9f695f84"
293
  },
 
297
  },
298
  {
299
  "cell_type": "markdown",
300
+ "id": "ba59d5f0",
301
  "metadata": {
302
  "id": "32ec8068-0bd7-412d-b662-0edb9d1e7365"
303
  },
 
309
  },
310
  {
311
  "cell_type": "markdown",
312
+ "id": "2b18e72b",
313
  "metadata": {
314
  "id": "589d9ec1-d12b-4b64-93f7-04c63997da19"
315
  },
 
324
  },
325
  {
326
  "cell_type": "markdown",
327
+ "id": "d6d354ea",
328
  "metadata": {
329
  "id": "b2ef54d5-b946-4c1d-9fdc-adc5d01b46aa"
330
  },
 
335
  {
336
  "cell_type": "code",
337
  "execution_count": 4,
338
+ "id": "c6452e65",
339
  "metadata": {
340
  "id": "bc77d7bb-f9e2-47f5-b663-30f7a4321ce5"
341
  },
 
348
  },
349
  {
350
  "cell_type": "markdown",
351
+ "id": "82d0ac42",
352
  "metadata": {
353
  "id": "93748af7-b917-4ecf-a0c8-7d89077ff9cb"
354
  },
 
358
  },
359
  {
360
  "cell_type": "markdown",
361
+ "id": "6db15460",
362
  "metadata": {
363
  "id": "2bc82609-a9fb-447a-a2af-99597c864029"
364
  },
 
372
  {
373
  "cell_type": "code",
374
  "execution_count": 5,
375
+ "id": "31c3f1fd",
376
  "metadata": {
377
  "id": "c7b07f9b-ae0e-4f89-98f0-0c50d432eab6",
378
  "outputId": "5c004b44-86e7-4e00-88be-39e0af5eed69"
 
386
  },
387
  {
388
  "cell_type": "markdown",
389
+ "id": "41e76f9a",
390
  "metadata": {
391
  "id": "d2ef23f3-f4a8-483a-a2dc-080a7496cb1b"
392
  },
 
396
  },
397
  {
398
  "cell_type": "markdown",
399
+ "id": "09603bd4",
400
  "metadata": {
401
  "id": "5ff67654-5a29-4bb8-a69d-0228946c6f8d"
402
  },
 
412
  {
413
  "cell_type": "code",
414
  "execution_count": 6,
415
+ "id": "1cb92985",
416
  "metadata": {
417
  "id": "77d9f0c5-8607-4642-a8ac-c3ab2e223ea6"
418
  },
 
425
  },
426
  {
427
  "cell_type": "markdown",
428
+ "id": "d9528e54",
429
  "metadata": {
430
  "id": "381acd09-0b0f-4d04-9eb3-f028ac0e5f2c"
431
  },
 
435
  },
436
  {
437
  "cell_type": "markdown",
438
+ "id": "2972da25",
439
  "metadata": {
440
  "id": "9649bf01-2e8a-45e5-8fca-441c13637b8f"
441
  },
 
447
  {
448
  "cell_type": "code",
449
  "execution_count": 7,
450
+ "id": "8ab54822",
451
  "metadata": {
452
  "id": "6e6b0ec5-0c94-4e2c-ae24-c791be1b2255"
453
  },
 
467
  },
468
  {
469
  "cell_type": "markdown",
470
+ "id": "1988343b",
471
  "metadata": {
472
  "id": "5a679f05-063d-41b3-9b58-4fc9c6ccf4fd"
473
  },
 
486
  {
487
  "cell_type": "code",
488
  "execution_count": 8,
489
+ "id": "f78f28f2",
490
  "metadata": {
491
  "id": "f12e2e57-156f-417b-8cfb-69221cc198e8"
492
  },
 
499
  },
500
  {
501
  "cell_type": "markdown",
502
+ "id": "4aa7d5d9",
503
  "metadata": {
504
  "id": "00382a3e-abec-4cdd-a54c-d1aaa3ea4707"
505
  },
 
511
  {
512
  "cell_type": "code",
513
  "execution_count": 9,
514
+ "id": "f50a2edb",
515
  "metadata": {
516
  "id": "87122d71-289a-466a-afcf-fa354b18946b"
517
  },
 
531
  },
532
  {
533
  "cell_type": "markdown",
534
+ "id": "e1d471fe",
535
  "metadata": {},
536
  "source": [
537
  "We'll define our pre-processing strategy. We advise that you **do not** lower-case the transcriptions or remove punctuation unless mixing different datasets. This will enable you to fine-tune Whisper models that can predict punctuation and casing. Later, you will see how we can evaluate the predictions without punctuation or casing, so that the models benefit from the WER improvement obtained by normalising the transcriptions while still predicting fully formatted transcriptions."
 
540
  {
541
  "cell_type": "code",
542
  "execution_count": 10,
543
+ "id": "756e987b",
544
  "metadata": {},
545
  "outputs": [],
546
  "source": [
 
554
  },
555
  {
556
  "cell_type": "markdown",
557
+ "id": "246c0ec3",
558
  "metadata": {},
559
  "source": [
560
  "Now we can write a function to prepare our data ready for the model:\n",
 
567
  {
568
  "cell_type": "code",
569
  "execution_count": 11,
570
+ "id": "b66341cd",
571
  "metadata": {},
572
  "outputs": [],
573
  "source": [
 
594
  },
595
  {
596
  "cell_type": "markdown",
597
+ "id": "6e81112a",
598
  "metadata": {
599
  "id": "70b319fb-2439-4ef6-a70d-a47bf41c4a13"
600
  },
 
605
  {
606
  "cell_type": "code",
607
  "execution_count": 12,
608
+ "id": "de797232",
609
  "metadata": {
610
  "id": "7b73ab39-ffaf-4b9e-86e5-782963c6134b"
611
  },
 
841
  },
842
  {
843
  "cell_type": "markdown",
844
+ "id": "ff5ca53b",
845
  "metadata": {},
846
  "source": [
847
  "Finally, we filter any training data with audio samples longer than 30s. These samples would otherwise be truncated by the Whisper feature-extractor which could affect the stability of training. We define a function that returns `True` for samples that are less than 30s, and `False` for those that are longer:"
 
850
  {
851
  "cell_type": "code",
852
  "execution_count": 13,
853
+ "id": "d99b0f59",
854
  "metadata": {},
855
  "outputs": [],
856
  "source": [
 
862
  },
863
  {
864
  "cell_type": "markdown",
865
+ "id": "99b7eb7e",
866
  "metadata": {},
867
  "source": [
868
  "We apply our filter function to all samples of our training dataset through 🤗 Datasets' `.filter` method:"
 
871
  {
872
  "cell_type": "code",
873
  "execution_count": 14,
874
+ "id": "2944ce59",
875
  "metadata": {},
876
  "outputs": [
877
  {
 
891
  },
892
  {
893
  "cell_type": "markdown",
894
+ "id": "0162b42e",
895
  "metadata": {
896
  "id": "263a5a58-0239-4a25-b0df-c625fc9c5810"
897
  },
 
901
  },
902
  {
903
  "cell_type": "markdown",
904
+ "id": "a02f599b",
905
  "metadata": {
906
  "id": "a693e768-c5a6-453f-89a1-b601dcf7daf7"
907
  },
 
924
  },
925
  {
926
  "cell_type": "markdown",
927
+ "id": "1e170a70",
928
  "metadata": {
929
  "id": "8d230e6d-624c-400a-bbf5-fa660881df25"
930
  },
 
934
  },
935
  {
936
  "cell_type": "markdown",
937
+ "id": "54e0500b",
938
  "metadata": {
939
  "id": "04def221-0637-4a69-b242-d3f0c1d0ee78"
940
  },
 
960
  {
961
  "cell_type": "code",
962
  "execution_count": 15,
963
+ "id": "88b48f61",
964
  "metadata": {
965
  "id": "8326221e-ec13-4731-bb4e-51e5fc1486c5"
966
  },
 
1001
  },
1002
  {
1003
  "cell_type": "markdown",
1004
+ "id": "e77fd74a",
1005
  "metadata": {
1006
  "id": "3cae7dbf-8a50-456e-a3a8-7fd005390f86"
1007
  },
 
1012
  {
1013
  "cell_type": "code",
1014
  "execution_count": 16,
1015
+ "id": "9d5293a5",
1016
  "metadata": {
1017
  "id": "fc834702-c0d3-4a96-b101-7b87be32bf42"
1018
  },
 
1023
  },
1024
  {
1025
  "cell_type": "markdown",
1026
+ "id": "ab5dfc21",
1027
  "metadata": {
1028
  "id": "d62bb2ab-750a-45e7-82e9-61d6f4805698"
1029
  },
 
1033
  },
1034
  {
1035
  "cell_type": "markdown",
1036
+ "id": "85ca98b1",
1037
  "metadata": {
1038
  "id": "66fee1a7-a44c-461e-b047-c3917221572e"
1039
  },
 
1045
  {
1046
  "cell_type": "code",
1047
  "execution_count": 17,
1048
+ "id": "c20575d7",
1049
  "metadata": {
1050
  "id": "b22b4011-f31f-4b57-b684-c52332f92890"
1051
  },
 
1058
  },
1059
  {
1060
  "cell_type": "markdown",
1061
+ "id": "b3cd5b3d",
1062
  "metadata": {
1063
  "id": "4f32cab6-31f0-4cb9-af4c-40ba0f5fc508"
1064
  },
 
1078
  {
1079
  "cell_type": "code",
1080
  "execution_count": 18,
1081
+ "id": "8b3b9e6e",
1082
  "metadata": {
1083
  "id": "23959a70-22d0-4ffe-9fa1-72b61e75bb52"
1084
  },
 
1109
  },
1110
  {
1111
  "cell_type": "markdown",
1112
+ "id": "aa47a845",
1113
  "metadata": {
1114
  "id": "daf2a825-6d9f-4a23-b145-c37c0039075b"
1115
  },
 
1119
  },
1120
  {
1121
  "cell_type": "markdown",
1122
+ "id": "d8d3980c",
1123
  "metadata": {
1124
  "id": "437a97fa-4864-476b-8abc-f28b8166cfa5"
1125
  },
 
1131
  {
1132
  "cell_type": "code",
1133
  "execution_count": 19,
1134
+ "id": "66fb8109",
1135
  "metadata": {
1136
  "id": "5a10cc4b-07ec-4ebd-ac1d-7c601023594f"
1137
  },
 
1144
  },
1145
  {
1146
  "cell_type": "markdown",
1147
+ "id": "6488d605",
1148
  "metadata": {
1149
  "id": "a15ead5f-2277-4a39-937b-585c2497b2df"
1150
  },
 
1155
  {
1156
  "cell_type": "code",
1157
  "execution_count": 20,
1158
+ "id": "bb9b6538",
1159
  "metadata": {
1160
  "id": "62038ba3-88ed-4fce-84db-338f50dcd04f"
1161
  },
 
1168
  },
1169
  {
1170
  "cell_type": "markdown",
1171
+ "id": "8c8ac68d",
1172
  "metadata": {
1173
  "id": "2178dea4-80ca-47b6-b6ea-ba1915c90c06"
1174
  },
 
1178
  },
1179
  {
1180
  "cell_type": "markdown",
1181
+ "id": "3ff54dd4",
1182
  "metadata": {
1183
  "id": "c21af1e9-0188-4134-ac82-defc7bdcc436"
1184
  },
 
1189
  {
1190
  "cell_type": "code",
1191
  "execution_count": 21,
1192
+ "id": "3c698301",
1193
  "metadata": {
1194
  "id": "0ae3e9af-97b7-4aa0-ae85-20b23b5bcb3a"
1195
  },
 
1223
  },
1224
  {
1225
  "cell_type": "markdown",
1226
+ "id": "f4e63b34",
1227
  "metadata": {
1228
  "id": "b3a944d8-3112-4552-82a0-be25988b3857"
1229
  },
 
1234
  },
1235
  {
1236
  "cell_type": "markdown",
1237
+ "id": "b109c728",
1238
  "metadata": {
1239
  "id": "bac29114-d226-4f54-97cf-8718c9f94e1e"
1240
  },
 
1246
  {
1247
  "cell_type": "code",
1248
  "execution_count": 22,
1249
+ "id": "119edab5",
1250
  "metadata": {
1251
  "id": "d546d7fe-0543-479a-b708-2ebabec19493"
1252
  },
 
1277
  },
1278
  {
1279
  "cell_type": "markdown",
1280
+ "id": "b7228116",
1281
  "metadata": {
1282
  "id": "uOrRhDGtN5S4"
1283
  },
 
1288
  {
1289
  "cell_type": "code",
1290
  "execution_count": 23,
1291
+ "id": "fabad1e6",
1292
  "metadata": {
1293
  "id": "-2zQwMfEOBJq"
1294
  },
 
1310
  },
1311
  {
1312
  "cell_type": "markdown",
1313
+ "id": "8feb0de8",
1314
  "metadata": {
1315
  "id": "7f404cf9-4345-468c-8196-4bd101d9bd51"
1316
  },
 
1320
  },
1321
  {
1322
  "cell_type": "markdown",
1323
+ "id": "1a63a349",
1324
  "metadata": {
1325
  "id": "5e8b8d56-5a70-4f68-bd2e-f0752d0bd112"
1326
  },
 
1337
  {
1338
  "cell_type": "code",
1339
  "execution_count": null,
1340
+ "id": "bc1ba2fe",
1341
  "metadata": {
1342
  "id": "ee8b7b8e-1c9a-4d77-9137-1778a629e6de"
1343
  },
 
1365
  "\n",
1366
  " <div>\n",
1367
  " \n",
1368
+ " <progress value='301' max='500' style='width:300px; height:20px; vertical-align: middle;'></progress>\n",
1369
+ " [301/500 1:46:59 < 1:11:12, 0.05 it/s, Epoch 6.38/11]\n",
1370
  " </div>\n",
1371
  " <table border=\"1\" class=\"dataframe\">\n",
1372
  " <thead>\n",
 
1402
  " <td>0.577473</td>\n",
1403
  " <td>19.811321</td>\n",
1404
  " </tr>\n",
1405
+ " <tr>\n",
1406
+ " <td>250</td>\n",
1407
+ " <td>0.010700</td>\n",
1408
+ " <td>0.555161</td>\n",
1409
+ " <td>19.706499</td>\n",
1410
+ " </tr>\n",
1411
+ " <tr>\n",
1412
+ " <td>300</td>\n",
1413
+ " <td>0.004400</td>\n",
1414
+ " <td>0.572261</td>\n",
1415
+ " <td>18.134172</td>\n",
1416
+ " </tr>\n",
1417
  " </tbody>\n",
1418
  "</table><p>"
1419
  ],
 
1453
  "Configuration saved in ./checkpoint-200/config.json\n",
1454
  "Model weights saved in ./checkpoint-200/pytorch_model.bin\n",
1455
  "Feature extractor saved in ./checkpoint-200/preprocessor_config.json\n",
1456
+ "Feature extractor saved in ./preprocessor_config.json\n",
1457
+ "The following columns in the evaluation set don't have a corresponding argument in `WhisperForConditionalGeneration.forward` and have been ignored: input_length. If input_length are not expected by `WhisperForConditionalGeneration.forward`, you can safely ignore this message.\n",
1458
+ "***** Running Evaluation *****\n",
1459
+ " Num examples = 56\n",
1460
+ " Batch size = 1\n",
1461
+ "The following columns in the evaluation set don't have a corresponding argument in `WhisperForConditionalGeneration.forward` and have been ignored: input_length. If input_length are not expected by `WhisperForConditionalGeneration.forward`, you can safely ignore this message.\n",
1462
+ "***** Running Evaluation *****\n",
1463
+ " Num examples = 56\n",
1464
+ " Batch size = 1\n",
1465
+ "Saving model checkpoint to ./checkpoint-300\n",
1466
+ "Configuration saved in ./checkpoint-300/config.json\n",
1467
+ "Model weights saved in ./checkpoint-300/pytorch_model.bin\n",
1468
+ "Feature extractor saved in ./checkpoint-300/preprocessor_config.json\n",
1469
  "Feature extractor saved in ./preprocessor_config.json\n"
1470
  ]
1471
  }
 
1477
  {
1478
  "cell_type": "code",
1479
  "execution_count": null,
1480
+ "id": "aa6aa51c",
1481
  "metadata": {
1482
  "id": "c704f91e-241b-48c9-b8e0-f0da396a9663"
1483
  },
 
1495
  },
1496
  {
1497
  "cell_type": "markdown",
1498
+ "id": "049805cf",
1499
  "metadata": {
1500
  "id": "090d676a-f944-4297-a938-a40eda0b2b68"
1501
  },
 
1506
  {
1507
  "cell_type": "code",
1508
  "execution_count": null,
1509
+ "id": "b014da90",
1510
  "metadata": {
1511
  "id": "d7030622-caf7-4039-939b-6195cdaa2585"
1512
  },
 
1517
  },
1518
  {
1519
  "cell_type": "markdown",
1520
+ "id": "61f78720",
1521
  "metadata": {
1522
  "id": "ca743fbd-602c-48d4-ba8d-a2fe60af64ba"
1523
  },
 
1527
  },
1528
  {
1529
  "cell_type": "markdown",
1530
+ "id": "c5d37e16",
1531
  "metadata": {
1532
  "id": "7f737783-2870-4e35-aa11-86a42d7d997a"
1533
  },
pytorch_model.bin CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:d8c43348411ba678b826cf38385cd78495c359c731740c58034b085acd1f2c80
3
  size 3055754841
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:de2e9a057a78b661960fe361039e46b708d14f8b441499a4cbee1ddb5ae92f5e
3
  size 3055754841
runs/Dec10_19-43-08_28bc7e304ae9/events.out.tfevents.1670701396.28bc7e304ae9.29234.0 CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:7b2d739035757f3b68e2b86c57c832ebb92c5e4a81b2bef2430bf41d9e5b6200
3
- size 6784
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ff9b841276431c9b975b3d2df6f580f374598cc181a22b0568a9627ce8638feb
3
+ size 8048