Hugging Face's logo

makhataei
/
Whisper-Small-Common-Voice

Automatic Speech Recognition
Transformers
TensorBoard
Safetensors
Persian
whisper
fa-asr
Generated from Trainer
Inference Endpoints
Model card Files Metrics Community
makhataei commited on
Commit
5aea8a5
•
1 Parent(s): 47054e3

Training in progress, step 1200

Browse files
This view is limited to 50 files because it contains too many changes.   See raw diff
Files changed (50) hide show
  1. 1/checkpoint-100/config.json +52 -0
  2. 1/checkpoint-100/generation_config.json +263 -0
  3. 1/checkpoint-100/model.safetensors +3 -0
  4. 1/checkpoint-100/optimizer.pt +3 -0
  5. 1/checkpoint-100/preprocessor_config.json +14 -0
  6. 1/checkpoint-100/rng_state.pth +3 -0
  7. 1/checkpoint-100/scheduler.pt +3 -0
  8. 1/checkpoint-100/trainer_state.json +52 -0
  9. 1/checkpoint-100/training_args.bin +3 -0
  10. 1/checkpoint-1000/config.json +52 -0
  11. 1/checkpoint-1000/generation_config.json +263 -0
  12. 1/checkpoint-1000/model.safetensors +3 -0
  13. 1/checkpoint-1000/optimizer.pt +3 -0
  14. 1/checkpoint-1000/preprocessor_config.json +14 -0
  15. 1/checkpoint-1000/rng_state.pth +3 -0
  16. 1/checkpoint-1000/scheduler.pt +3 -0
  17. 1/checkpoint-1000/trainer_state.json +349 -0
  18. 1/checkpoint-1000/training_args.bin +3 -0
  19. 1/checkpoint-10000/config.json +52 -0
  20. 1/checkpoint-10000/generation_config.json +263 -0
  21. 1/checkpoint-10000/model.safetensors +3 -0
  22. 1/checkpoint-10000/optimizer.pt +3 -0
  23. 1/checkpoint-10000/preprocessor_config.json +14 -0
  24. 1/checkpoint-10000/rng_state.pth +3 -0
  25. 1/checkpoint-10000/scheduler.pt +3 -0
  26. 1/checkpoint-10000/trainer_state.json +3319 -0
  27. 1/checkpoint-10000/training_args.bin +3 -0
  28. 1/checkpoint-1100/config.json +52 -0
  29. 1/checkpoint-1100/generation_config.json +263 -0
  30. 1/checkpoint-1100/model.safetensors +3 -0
  31. 1/checkpoint-1100/optimizer.pt +3 -0
  32. 1/checkpoint-1100/preprocessor_config.json +14 -0
  33. 1/checkpoint-1100/rng_state.pth +3 -0
  34. 1/checkpoint-1100/scheduler.pt +3 -0
  35. 1/checkpoint-1100/trainer_state.json +382 -0
  36. 1/checkpoint-1100/training_args.bin +3 -0
  37. 1/checkpoint-1200/config.json +52 -0
  38. 1/checkpoint-1200/generation_config.json +263 -0
  39. 1/checkpoint-1200/model.safetensors +3 -0
  40. 1/checkpoint-1200/optimizer.pt +3 -0
  41. 1/checkpoint-1200/preprocessor_config.json +14 -0
  42. 1/checkpoint-1200/rng_state.pth +3 -0
  43. 1/checkpoint-1200/scheduler.pt +3 -0
  44. 1/checkpoint-1200/trainer_state.json +415 -0
  45. 1/checkpoint-1200/training_args.bin +3 -0
  46. 1/checkpoint-1300/config.json +52 -0
  47. 1/checkpoint-1300/generation_config.json +263 -0
  48. 1/checkpoint-1300/model.safetensors +3 -0
  49. 1/checkpoint-1300/optimizer.pt +3 -0
  50. 1/checkpoint-1300/preprocessor_config.json +14 -0
1/checkpoint-100/config.json ADDED
@@ -0,0 +1,52 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "openai/whisper-small",
3
+ "activation_dropout": 0.0,
4
+ "activation_function": "gelu",
5
+ "apply_spec_augment": false,
6
+ "architectures": [
7
+ "WhisperForConditionalGeneration"
8
+ ],
9
+ "attention_dropout": 0.0,
10
+ "begin_suppress_tokens": [
11
+ 220,
12
+ 50257
13
+ ],
14
+ "bos_token_id": 50257,
15
+ "classifier_proj_size": 256,
16
+ "d_model": 768,
17
+ "decoder_attention_heads": 12,
18
+ "decoder_ffn_dim": 3072,
19
+ "decoder_layerdrop": 0.0,
20
+ "decoder_layers": 12,
21
+ "decoder_start_token_id": 50258,
22
+ "dropout": 0.0,
23
+ "encoder_attention_heads": 12,
24
+ "encoder_ffn_dim": 3072,
25
+ "encoder_layerdrop": 0.0,
26
+ "encoder_layers": 12,
27
+ "eos_token_id": 50257,
28
+ "forced_decoder_ids": null,
29
+ "init_std": 0.02,
30
+ "is_encoder_decoder": true,
31
+ "mask_feature_length": 10,
32
+ "mask_feature_min_masks": 0,
33
+ "mask_feature_prob": 0.0,
34
+ "mask_time_length": 10,
35
+ "mask_time_min_masks": 2,
36
+ "mask_time_prob": 0.05,
37
+ "max_length": 448,
38
+ "max_source_positions": 1500,
39
+ "max_target_positions": 448,
40
+ "median_filter_width": 7,
41
+ "model_type": "whisper",
42
+ "num_hidden_layers": 12,
43
+ "num_mel_bins": 80,
44
+ "pad_token_id": 50257,
45
+ "scale_embedding": false,
46
+ "suppress_tokens": [],
47
+ "torch_dtype": "float32",
48
+ "transformers_version": "4.35.0",
49
+ "use_cache": true,
50
+ "use_weighted_layer_sum": false,
51
+ "vocab_size": 51865
52
+ }
1/checkpoint-100/generation_config.json ADDED
@@ -0,0 +1,263 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "alignment_heads": [
3
+ [
4
+ 5,
5
+ 3
6
+ ],
7
+ [
8
+ 5,
9
+ 9
10
+ ],
11
+ [
12
+ 8,
13
+ 0
14
+ ],
15
+ [
16
+ 8,
17
+ 4
18
+ ],
19
+ [
20
+ 8,
21
+ 7
22
+ ],
23
+ [
24
+ 8,
25
+ 8
26
+ ],
27
+ [
28
+ 9,
29
+ 0
30
+ ],
31
+ [
32
+ 9,
33
+ 7
34
+ ],
35
+ [
36
+ 9,
37
+ 9
38
+ ],
39
+ [
40
+ 10,
41
+ 5
42
+ ]
43
+ ],
44
+ "begin_suppress_tokens": [
45
+ 220,
46
+ 50257
47
+ ],
48
+ "bos_token_id": 50257,
49
+ "decoder_start_token_id": 50258,
50
+ "eos_token_id": 50257,
51
+ "forced_decoder_ids": [
52
+ [
53
+ 1,
54
+ null
55
+ ],
56
+ [
57
+ 2,
58
+ 50359
59
+ ]
60
+ ],
61
+ "is_multilingual": true,
62
+ "lang_to_id": {
63
+ "<|af|>": 50327,
64
+ "<|am|>": 50334,
65
+ "<|ar|>": 50272,
66
+ "<|as|>": 50350,
67
+ "<|az|>": 50304,
68
+ "<|ba|>": 50355,
69
+ "<|be|>": 50330,
70
+ "<|bg|>": 50292,
71
+ "<|bn|>": 50302,
72
+ "<|bo|>": 50347,
73
+ "<|br|>": 50309,
74
+ "<|bs|>": 50315,
75
+ "<|ca|>": 50270,
76
+ "<|cs|>": 50283,
77
+ "<|cy|>": 50297,
78
+ "<|da|>": 50285,
79
+ "<|de|>": 50261,
80
+ "<|el|>": 50281,
81
+ "<|en|>": 50259,
82
+ "<|es|>": 50262,
83
+ "<|et|>": 50307,
84
+ "<|eu|>": 50310,
85
+ "<|fa|>": 50300,
86
+ "<|fi|>": 50277,
87
+ "<|fo|>": 50338,
88
+ "<|fr|>": 50265,
89
+ "<|gl|>": 50319,
90
+ "<|gu|>": 50333,
91
+ "<|haw|>": 50352,
92
+ "<|ha|>": 50354,
93
+ "<|he|>": 50279,
94
+ "<|hi|>": 50276,
95
+ "<|hr|>": 50291,
96
+ "<|ht|>": 50339,
97
+ "<|hu|>": 50286,
98
+ "<|hy|>": 50312,
99
+ "<|id|>": 50275,
100
+ "<|is|>": 50311,
101
+ "<|it|>": 50274,
102
+ "<|ja|>": 50266,
103
+ "<|jw|>": 50356,
104
+ "<|ka|>": 50329,
105
+ "<|kk|>": 50316,
106
+ "<|km|>": 50323,
107
+ "<|kn|>": 50306,
108
+ "<|ko|>": 50264,
109
+ "<|la|>": 50294,
110
+ "<|lb|>": 50345,
111
+ "<|ln|>": 50353,
112
+ "<|lo|>": 50336,
113
+ "<|lt|>": 50293,
114
+ "<|lv|>": 50301,
115
+ "<|mg|>": 50349,
116
+ "<|mi|>": 50295,
117
+ "<|mk|>": 50308,
118
+ "<|ml|>": 50296,
119
+ "<|mn|>": 50314,
120
+ "<|mr|>": 50320,
121
+ "<|ms|>": 50282,
122
+ "<|mt|>": 50343,
123
+ "<|my|>": 50346,
124
+ "<|ne|>": 50313,
125
+ "<|nl|>": 50271,
126
+ "<|nn|>": 50342,
127
+ "<|no|>": 50288,
128
+ "<|oc|>": 50328,
129
+ "<|pa|>": 50321,
130
+ "<|pl|>": 50269,
131
+ "<|ps|>": 50340,
132
+ "<|pt|>": 50267,
133
+ "<|ro|>": 50284,
134
+ "<|ru|>": 50263,
135
+ "<|sa|>": 50344,
136
+ "<|sd|>": 50332,
137
+ "<|si|>": 50322,
138
+ "<|sk|>": 50298,
139
+ "<|sl|>": 50305,
140
+ "<|sn|>": 50324,
141
+ "<|so|>": 50326,
142
+ "<|sq|>": 50317,
143
+ "<|sr|>": 50303,
144
+ "<|su|>": 50357,
145
+ "<|sv|>": 50273,
146
+ "<|sw|>": 50318,
147
+ "<|ta|>": 50287,
148
+ "<|te|>": 50299,
149
+ "<|tg|>": 50331,
150
+ "<|th|>": 50289,
151
+ "<|tk|>": 50341,
152
+ "<|tl|>": 50348,
153
+ "<|tr|>": 50268,
154
+ "<|tt|>": 50351,
155
+ "<|uk|>": 50280,
156
+ "<|ur|>": 50290,
157
+ "<|uz|>": 50337,
158
+ "<|vi|>": 50278,
159
+ "<|yi|>": 50335,
160
+ "<|yo|>": 50325,
161
+ "<|zh|>": 50260
162
+ },
163
+ "max_initial_timestamp_index": 1,
164
+ "max_length": 448,
165
+ "no_timestamps_token_id": 50363,
166
+ "pad_token_id": 50257,
167
+ "return_timestamps": false,
168
+ "suppress_tokens": [
169
+ 1,
170
+ 2,
171
+ 7,
172
+ 8,
173
+ 9,
174
+ 10,
175
+ 14,
176
+ 25,
177
+ 26,
178
+ 27,
179
+ 28,
180
+ 29,
181
+ 31,
182
+ 58,
183
+ 59,
184
+ 60,
185
+ 61,
186
+ 62,
187
+ 63,
188
+ 90,
189
+ 91,
190
+ 92,
191
+ 93,
192
+ 359,
193
+ 503,
194
+ 522,
195
+ 542,
196
+ 873,
197
+ 893,
198
+ 902,
199
+ 918,
200
+ 922,
201
+ 931,
202
+ 1350,
203
+ 1853,
204
+ 1982,
205
+ 2460,
206
+ 2627,
207
+ 3246,
208
+ 3253,
209
+ 3268,
210
+ 3536,
211
+ 3846,
212
+ 3961,
213
+ 4183,
214
+ 4667,
215
+ 6585,
216
+ 6647,
217
+ 7273,
218
+ 9061,
219
+ 9383,
220
+ 10428,
221
+ 10929,
222
+ 11938,
223
+ 12033,
224
+ 12331,
225
+ 12562,
226
+ 13793,
227
+ 14157,
228
+ 14635,
229
+ 15265,
230
+ 15618,
231
+ 16553,
232
+ 16604,
233
+ 18362,
234
+ 18956,
235
+ 20075,
236
+ 21675,
237
+ 22520,
238
+ 26130,
239
+ 26161,
240
+ 26435,
241
+ 28279,
242
+ 29464,
243
+ 31650,
244
+ 32302,
245
+ 32470,
246
+ 36865,
247
+ 42863,
248
+ 47425,
249
+ 49870,
250
+ 50254,
251
+ 50258,
252
+ 50358,
253
+ 50359,
254
+ 50360,
255
+ 50361,
256
+ 50362
257
+ ],
258
+ "task_to_id": {
259
+ "transcribe": 50359,
260
+ "translate": 50358
261
+ },
262
+ "transformers_version": "4.35.0"
263
+ }
1/checkpoint-100/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3e18335ec5da776fe2b0b7acd7ab199ff873c1e33ce374a03636f96c75112dd6
3
+ size 966995080
1/checkpoint-100/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c418ac9599634bb255d0564c060a059d52536e299a26610189d911e36f64eb1d
3
+ size 1925063607
1/checkpoint-100/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
+ }
1/checkpoint-100/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f77a1172dee199d266b3c3c0624e631fa34d4bcad64000acfdff24213111834d
3
+ size 14575
1/checkpoint-100/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:23b5570fa0a948cd54b0d1f92b51a79341aa485dfe5365d6dee8c298b40ef4fe
3
+ size 627
1/checkpoint-100/trainer_state.json ADDED
@@ -0,0 +1,52 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": 58.07742554282218,
3
+ "best_model_checkpoint": "/media/makhataei/Backups/Whisper-Small-Common-Voice/checkpoint-100",
4
+ "epoch": 0.3861003861003861,
5
+ "eval_steps": 100,
6
+ "global_step": 100,
7
+ "is_hyper_param_search": false,
8
+ "is_local_process_zero": true,
9
+ "is_world_process_zero": true,
10
+ "log_history": [
11
+ {
12
+ "epoch": 0.1,
13
+ "learning_rate": 4.600000000000001e-06,
14
+ "loss": 3.1025,
15
+ "step": 25
16
+ },
17
+ {
18
+ "epoch": 0.19,
19
+ "learning_rate": 9.600000000000001e-06,
20
+ "loss": 1.3525,
21
+ "step": 50
22
+ },
23
+ {
24
+ "epoch": 0.29,
25
+ "learning_rate": 9.976884422110554e-06,
26
+ "loss": 0.9564,
27
+ "step": 75
28
+ },
29
+ {
30
+ "epoch": 0.39,
31
+ "learning_rate": 9.95175879396985e-06,
32
+ "loss": 0.7495,
33
+ "step": 100
34
+ },
35
+ {
36
+ "epoch": 0.39,
37
+ "eval_loss": 0.860159158706665,
38
+ "eval_runtime": 1426.2695,
39
+ "eval_samples_per_second": 6.062,
40
+ "eval_steps_per_second": 0.606,
41
+ "eval_wer": 58.07742554282218,
42
+ "step": 100
43
+ }
44
+ ],
45
+ "logging_steps": 25,
46
+ "max_steps": 10000,
47
+ "num_train_epochs": 39,
48
+ "save_steps": 100,
49
+ "total_flos": 1.616078241792e+18,
50
+ "trial_name": null,
51
+ "trial_params": null
52
+ }
1/checkpoint-100/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d286c349fc7a4c7419632d8b055a552e2835b4138f3ff268d56552451405f53b
3
+ size 4347
1/checkpoint-1000/config.json ADDED
@@ -0,0 +1,52 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "openai/whisper-small",
3
+ "activation_dropout": 0.0,
4
+ "activation_function": "gelu",
5
+ "apply_spec_augment": false,
6
+ "architectures": [
7
+ "WhisperForConditionalGeneration"
8
+ ],
9
+ "attention_dropout": 0.0,
10
+ "begin_suppress_tokens": [
11
+ 220,
12
+ 50257
13
+ ],
14
+ "bos_token_id": 50257,
15
+ "classifier_proj_size": 256,
16
+ "d_model": 768,
17
+ "decoder_attention_heads": 12,
18
+ "decoder_ffn_dim": 3072,
19
+ "decoder_layerdrop": 0.0,
20
+ "decoder_layers": 12,
21
+ "decoder_start_token_id": 50258,
22
+ "dropout": 0.0,
23
+ "encoder_attention_heads": 12,
24
+ "encoder_ffn_dim": 3072,
25
+ "encoder_layerdrop": 0.0,
26
+ "encoder_layers": 12,
27
+ "eos_token_id": 50257,
28
+ "forced_decoder_ids": null,
29
+ "init_std": 0.02,
30
+ "is_encoder_decoder": true,
31
+ "mask_feature_length": 10,
32
+ "mask_feature_min_masks": 0,
33
+ "mask_feature_prob": 0.0,
34
+ "mask_time_length": 10,
35
+ "mask_time_min_masks": 2,
36
+ "mask_time_prob": 0.05,
37
+ "max_length": 448,
38
+ "max_source_positions": 1500,
39
+ "max_target_positions": 448,
40
+ "median_filter_width": 7,
41
+ "model_type": "whisper",
42
+ "num_hidden_layers": 12,
43
+ "num_mel_bins": 80,
44
+ "pad_token_id": 50257,
45
+ "scale_embedding": false,
46
+ "suppress_tokens": [],
47
+ "torch_dtype": "float32",
48
+ "transformers_version": "4.35.0",
49
+ "use_cache": true,
50
+ "use_weighted_layer_sum": false,
51
+ "vocab_size": 51865
52
+ }
1/checkpoint-1000/generation_config.json ADDED
@@ -0,0 +1,263 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "alignment_heads": [
3
+ [
4
+ 5,
5
+ 3
6
+ ],
7
+ [
8
+ 5,
9
+ 9
10
+ ],
11
+ [
12
+ 8,
13
+ 0
14
+ ],
15
+ [
16
+ 8,
17
+ 4
18
+ ],
19
+ [
20
+ 8,
21
+ 7
22
+ ],
23
+ [
24
+ 8,
25
+ 8
26
+ ],
27
+ [
28
+ 9,
29
+ 0
30
+ ],
31
+ [
32
+ 9,
33
+ 7
34
+ ],
35
+ [
36
+ 9,
37
+ 9
38
+ ],
39
+ [
40
+ 10,
41
+ 5
42
+ ]
43
+ ],
44
+ "begin_suppress_tokens": [
45
+ 220,
46
+ 50257
47
+ ],
48
+ "bos_token_id": 50257,
49
+ "decoder_start_token_id": 50258,
50
+ "eos_token_id": 50257,
51
+ "forced_decoder_ids": [
52
+ [
53
+ 1,
54
+ null
55
+ ],
56
+ [
57
+ 2,
58
+ 50359
59
+ ]
60
+ ],
61
+ "is_multilingual": true,
62
+ "lang_to_id": {
63
+ "<|af|>": 50327,
64
+ "<|am|>": 50334,
65
+ "<|ar|>": 50272,
66
+ "<|as|>": 50350,
67
+ "<|az|>": 50304,
68
+ "<|ba|>": 50355,
69
+ "<|be|>": 50330,
70
+ "<|bg|>": 50292,
71
+ "<|bn|>": 50302,
72
+ "<|bo|>": 50347,
73
+ "<|br|>": 50309,
74
+ "<|bs|>": 50315,
75
+ "<|ca|>": 50270,
76
+ "<|cs|>": 50283,
77
+ "<|cy|>": 50297,
78
+ "<|da|>": 50285,
79
+ "<|de|>": 50261,
80
+ "<|el|>": 50281,
81
+ "<|en|>": 50259,
82
+ "<|es|>": 50262,
83
+ "<|et|>": 50307,
84
+ "<|eu|>": 50310,
85
+ "<|fa|>": 50300,
86
+ "<|fi|>": 50277,
87
+ "<|fo|>": 50338,
88
+ "<|fr|>": 50265,
89
+ "<|gl|>": 50319,
90
+ "<|gu|>": 50333,
91
+ "<|haw|>": 50352,
92
+ "<|ha|>": 50354,
93
+ "<|he|>": 50279,
94
+ "<|hi|>": 50276,
95
+ "<|hr|>": 50291,
96
+ "<|ht|>": 50339,
97
+ "<|hu|>": 50286,
98
+ "<|hy|>": 50312,
99
+ "<|id|>": 50275,
100
+ "<|is|>": 50311,
101
+ "<|it|>": 50274,
102
+ "<|ja|>": 50266,
103
+ "<|jw|>": 50356,
104
+ "<|ka|>": 50329,
105
+ "<|kk|>": 50316,
106
+ "<|km|>": 50323,
107
+ "<|kn|>": 50306,
108
+ "<|ko|>": 50264,
109
+ "<|la|>": 50294,
110
+ "<|lb|>": 50345,
111
+ "<|ln|>": 50353,
112
+ "<|lo|>": 50336,
113
+ "<|lt|>": 50293,
114
+ "<|lv|>": 50301,
115
+ "<|mg|>": 50349,
116
+ "<|mi|>": 50295,
117
+ "<|mk|>": 50308,
118
+ "<|ml|>": 50296,
119
+ "<|mn|>": 50314,
120
+ "<|mr|>": 50320,
121
+ "<|ms|>": 50282,
122
+ "<|mt|>": 50343,
123
+ "<|my|>": 50346,
124
+ "<|ne|>": 50313,
125
+ "<|nl|>": 50271,
126
+ "<|nn|>": 50342,
127
+ "<|no|>": 50288,
128
+ "<|oc|>": 50328,
129
+ "<|pa|>": 50321,
130
+ "<|pl|>": 50269,
131
+ "<|ps|>": 50340,
132
+ "<|pt|>": 50267,
133
+ "<|ro|>": 50284,
134
+ "<|ru|>": 50263,
135
+ "<|sa|>": 50344,
136
+ "<|sd|>": 50332,
137
+ "<|si|>": 50322,
138
+ "<|sk|>": 50298,
139
+ "<|sl|>": 50305,
140
+ "<|sn|>": 50324,
141
+ "<|so|>": 50326,
142
+ "<|sq|>": 50317,
143
+ "<|sr|>": 50303,
144
+ "<|su|>": 50357,
145
+ "<|sv|>": 50273,
146
+ "<|sw|>": 50318,
147
+ "<|ta|>": 50287,
148
+ "<|te|>": 50299,
149
+ "<|tg|>": 50331,
150
+ "<|th|>": 50289,
151
+ "<|tk|>": 50341,
152
+ "<|tl|>": 50348,
153
+ "<|tr|>": 50268,
154
+ "<|tt|>": 50351,
155
+ "<|uk|>": 50280,
156
+ "<|ur|>": 50290,
157
+ "<|uz|>": 50337,
158
+ "<|vi|>": 50278,
159
+ "<|yi|>": 50335,
160
+ "<|yo|>": 50325,
161
+ "<|zh|>": 50260
162
+ },
163
+ "max_initial_timestamp_index": 1,
164
+ "max_length": 448,
165
+ "no_timestamps_token_id": 50363,
166
+ "pad_token_id": 50257,
167
+ "return_timestamps": false,
168
+ "suppress_tokens": [
169
+ 1,
170
+ 2,
171
+ 7,
172
+ 8,
173
+ 9,
174
+ 10,
175
+ 14,
176
+ 25,
177
+ 26,
178
+ 27,
179
+ 28,
180
+ 29,
181
+ 31,
182
+ 58,
183
+ 59,
184
+ 60,
185
+ 61,
186
+ 62,
187
+ 63,
188
+ 90,
189
+ 91,
190
+ 92,
191
+ 93,
192
+ 359,
193
+ 503,
194
+ 522,
195
+ 542,
196
+ 873,
197
+ 893,
198
+ 902,
199
+ 918,
200
+ 922,
201
+ 931,
202
+ 1350,
203
+ 1853,
204
+ 1982,
205
+ 2460,
206
+ 2627,
207
+ 3246,
208
+ 3253,
209
+ 3268,
210
+ 3536,
211
+ 3846,
212
+ 3961,
213
+ 4183,
214
+ 4667,
215
+ 6585,
216
+ 6647,
217
+ 7273,
218
+ 9061,
219
+ 9383,
220
+ 10428,
221
+ 10929,
222
+ 11938,
223
+ 12033,
224
+ 12331,
225
+ 12562,
226
+ 13793,
227
+ 14157,
228
+ 14635,
229
+ 15265,
230
+ 15618,
231
+ 16553,
232
+ 16604,
233
+ 18362,
234
+ 18956,
235
+ 20075,
236
+ 21675,
237
+ 22520,
238
+ 26130,
239
+ 26161,
240
+ 26435,
241
+ 28279,
242
+ 29464,
243
+ 31650,
244
+ 32302,
245
+ 32470,
246
+ 36865,
247
+ 42863,
248
+ 47425,
249
+ 49870,
250
+ 50254,
251
+ 50258,
252
+ 50358,
253
+ 50359,
254
+ 50360,
255
+ 50361,
256
+ 50362
257
+ ],
258
+ "task_to_id": {
259
+ "transcribe": 50359,
260
+ "translate": 50358
261
+ },
262
+ "transformers_version": "4.35.0"
263
+ }
1/checkpoint-1000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5a998a900bd5736f988809531518496becb09ad1d9ef5fcce01828a8085fad74
3
+ size 966995080
1/checkpoint-1000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6203658dd395f90aabf25b0c2caa71050e61dcb36b48c57573496da2a152c08c
3
+ size 1925063607
1/checkpoint-1000/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
+ }
1/checkpoint-1000/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:32b4ce27efd07479bf346bd3f690d602c835e6ad5edea80a5f639b8451132ec3
3
+ size 14575
1/checkpoint-1000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bd34d3550383c621d310773592dca6a7c40a2035e8646bee311fa3164e3ef513
3
+ size 627
1/checkpoint-1000/trainer_state.json ADDED
@@ -0,0 +1,349 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": 46.55123943708554,
3
+ "best_model_checkpoint": "/media/makhataei/Backups/Whisper-Small-Common-Voice/checkpoint-500",
4
+ "epoch": 3.861003861003861,
5
+ "eval_steps": 100,
6
+ "global_step": 1000,
7
+ "is_hyper_param_search": false,
8
+ "is_local_process_zero": true,
9
+ "is_world_process_zero": true,
10
+ "log_history": [
11
+ {
12
+ "epoch": 0.1,
13
+ "learning_rate": 4.600000000000001e-06,
14
+ "loss": 3.1025,
15
+ "step": 25
16
+ },
17
+ {
18
+ "epoch": 0.19,
19
+ "learning_rate": 9.600000000000001e-06,
20
+ "loss": 1.3525,
21
+ "step": 50
22
+ },
23
+ {
24
+ "epoch": 0.29,
25
+ "learning_rate": 9.976884422110554e-06,
26
+ "loss": 0.9564,
27
+ "step": 75
28
+ },
29
+ {
30
+ "epoch": 0.39,
31
+ "learning_rate": 9.95175879396985e-06,
32
+ "loss": 0.7495,
33
+ "step": 100
34
+ },
35
+ {
36
+ "epoch": 0.39,
37
+ "eval_loss": 0.860159158706665,
38
+ "eval_runtime": 1426.2695,
39
+ "eval_samples_per_second": 6.062,
40
+ "eval_steps_per_second": 0.606,
41
+ "eval_wer": 58.07742554282218,
42
+ "step": 100
43
+ },
44
+ {
45
+ "epoch": 0.48,
46
+ "learning_rate": 9.926633165829147e-06,
47
+ "loss": 0.5073,
48
+ "step": 125
49
+ },
50
+ {
51
+ "epoch": 0.58,
52
+ "learning_rate": 9.901507537688444e-06,
53
+ "loss": 0.3749,
54
+ "step": 150
55
+ },
56
+ {
57
+ "epoch": 0.68,
58
+ "learning_rate": 9.87638190954774e-06,
59
+ "loss": 0.3247,
60
+ "step": 175
61
+ },
62
+ {
63
+ "epoch": 0.77,
64
+ "learning_rate": 9.851256281407035e-06,
65
+ "loss": 0.3059,
66
+ "step": 200
67
+ },
68
+ {
69
+ "epoch": 0.77,
70
+ "eval_loss": 0.5538379549980164,
71
+ "eval_runtime": 1589.2333,
72
+ "eval_samples_per_second": 5.44,
73
+ "eval_steps_per_second": 0.544,
74
+ "eval_wer": 55.22481850804518,
75
+ "step": 200
76
+ },
77
+ {
78
+ "epoch": 0.87,
79
+ "learning_rate": 9.826130653266333e-06,
80
+ "loss": 0.2963,
81
+ "step": 225
82
+ },
83
+ {
84
+ "epoch": 0.97,
85
+ "learning_rate": 9.80100502512563e-06,
86
+ "loss": 0.2882,
87
+ "step": 250
88
+ },
89
+ {
90
+ "epoch": 1.06,
91
+ "learning_rate": 9.775879396984925e-06,
92
+ "loss": 0.2327,
93
+ "step": 275
94
+ },
95
+ {
96
+ "epoch": 1.16,
97
+ "learning_rate": 9.750753768844222e-06,
98
+ "loss": 0.2002,
99
+ "step": 300
100
+ },
101
+ {
102
+ "epoch": 1.16,
103
+ "eval_loss": 0.5136818885803223,
104
+ "eval_runtime": 1646.1357,
105
+ "eval_samples_per_second": 5.252,
106
+ "eval_steps_per_second": 0.525,
107
+ "eval_wer": 52.46151047609599,
108
+ "step": 300
109
+ },
110
+ {
111
+ "epoch": 1.25,
112
+ "learning_rate": 9.725628140703518e-06,
113
+ "loss": 0.1893,
114
+ "step": 325
115
+ },
116
+ {
117
+ "epoch": 1.35,
118
+ "learning_rate": 9.700502512562815e-06,
119
+ "loss": 0.185,
120
+ "step": 350
121
+ },
122
+ {
123
+ "epoch": 1.45,
124
+ "learning_rate": 9.675376884422111e-06,
125
+ "loss": 0.1847,
126
+ "step": 375
127
+ },
128
+ {
129
+ "epoch": 1.54,
130
+ "learning_rate": 9.650251256281408e-06,
131
+ "loss": 0.1798,
132
+ "step": 400
133
+ },
134
+ {
135
+ "epoch": 1.54,
136
+ "eval_loss": 0.5102902054786682,
137
+ "eval_runtime": 1650.3734,
138
+ "eval_samples_per_second": 5.239,
139
+ "eval_steps_per_second": 0.524,
140
+ "eval_wer": 51.65781945064576,
141
+ "step": 400
142
+ },
143
+ {
144
+ "epoch": 1.64,
145
+ "learning_rate": 9.625125628140705e-06,
146
+ "loss": 0.1804,
147
+ "step": 425
148
+ },
149
+ {
150
+ "epoch": 1.74,
151
+ "learning_rate": 9.600000000000001e-06,
152
+ "loss": 0.1807,
153
+ "step": 450
154
+ },
155
+ {
156
+ "epoch": 1.83,
157
+ "learning_rate": 9.574874371859298e-06,
158
+ "loss": 0.1714,
159
+ "step": 475
160
+ },
161
+ {
162
+ "epoch": 1.93,
163
+ "learning_rate": 9.549748743718595e-06,
164
+ "loss": 0.1678,
165
+ "step": 500
166
+ },
167
+ {
168
+ "epoch": 1.93,
169
+ "eval_loss": 0.48063144087791443,
170
+ "eval_runtime": 1482.5018,
171
+ "eval_samples_per_second": 5.832,
172
+ "eval_steps_per_second": 0.583,
173
+ "eval_wer": 46.55123943708554,
174
+ "step": 500
175
+ },
176
+ {
177
+ "epoch": 2.03,
178
+ "learning_rate": 9.52462311557789e-06,
179
+ "loss": 0.1542,
180
+ "step": 525
181
+ },
182
+ {
183
+ "epoch": 2.12,
184
+ "learning_rate": 9.499497487437188e-06,
185
+ "loss": 0.0929,
186
+ "step": 550
187
+ },
188
+ {
189
+ "epoch": 2.22,
190
+ "learning_rate": 9.474371859296483e-06,
191
+ "loss": 0.0936,
192
+ "step": 575
193
+ },
194
+ {
195
+ "epoch": 2.32,
196
+ "learning_rate": 9.44924623115578e-06,
197
+ "loss": 0.0932,
198
+ "step": 600
199
+ },
200
+ {
201
+ "epoch": 2.32,
202
+ "eval_loss": 0.49363330006599426,
203
+ "eval_runtime": 1993.4903,
204
+ "eval_samples_per_second": 4.337,
205
+ "eval_steps_per_second": 0.434,
206
+ "eval_wer": 47.49549370772767,
207
+ "step": 600
208
+ },
209
+ {
210
+ "epoch": 2.41,
211
+ "learning_rate": 9.424120603015076e-06,
212
+ "loss": 0.0936,
213
+ "step": 625
214
+ },
215
+ {
216
+ "epoch": 2.51,
217
+ "learning_rate": 9.398994974874373e-06,
218
+ "loss": 0.0955,
219
+ "step": 650
220
+ },
221
+ {
222
+ "epoch": 2.61,
223
+ "learning_rate": 9.373869346733669e-06,
224
+ "loss": 0.0871,
225
+ "step": 675
226
+ },
227
+ {
228
+ "epoch": 2.7,
229
+ "learning_rate": 9.348743718592966e-06,
230
+ "loss": 0.0944,
231
+ "step": 700
232
+ },
233
+ {
234
+ "epoch": 2.7,
235
+ "eval_loss": 0.5119881629943848,
236
+ "eval_runtime": 1791.3192,
237
+ "eval_samples_per_second": 4.827,
238
+ "eval_steps_per_second": 0.483,
239
+ "eval_wer": 48.91600932678474,
240
+ "step": 700
241
+ },
242
+ {
243
+ "epoch": 2.8,
244
+ "learning_rate": 9.323618090452262e-06,
245
+ "loss": 0.0927,
246
+ "step": 725
247
+ },
248
+ {
249
+ "epoch": 2.9,
250
+ "learning_rate": 9.298492462311557e-06,
251
+ "loss": 0.0964,
252
+ "step": 750
253
+ },
254
+ {
255
+ "epoch": 2.99,
256
+ "learning_rate": 9.273366834170856e-06,
257
+ "loss": 0.1038,
258
+ "step": 775
259
+ },
260
+ {
261
+ "epoch": 3.09,
262
+ "learning_rate": 9.248241206030152e-06,
263
+ "loss": 0.0493,
264
+ "step": 800
265
+ },
266
+ {
267
+ "epoch": 3.09,
268
+ "eval_loss": 0.5198288559913635,
269
+ "eval_runtime": 1878.4077,
270
+ "eval_samples_per_second": 4.603,
271
+ "eval_steps_per_second": 0.46,
272
+ "eval_wer": 48.26114997271419,
273
+ "step": 800
274
+ },
275
+ {
276
+ "epoch": 3.19,
277
+ "learning_rate": 9.223115577889447e-06,
278
+ "loss": 0.0501,
279
+ "step": 825
280
+ },
281
+ {
282
+ "epoch": 3.28,
283
+ "learning_rate": 9.197989949748744e-06,
284
+ "loss": 0.0458,
285
+ "step": 850
286
+ },
287
+ {
288
+ "epoch": 3.38,
289
+ "learning_rate": 9.17286432160804e-06,
290
+ "loss": 0.0492,
291
+ "step": 875
292
+ },
293
+ {
294
+ "epoch": 3.47,
295
+ "learning_rate": 9.147738693467337e-06,
296
+ "loss": 0.0506,
297
+ "step": 900
298
+ },
299
+ {
300
+ "epoch": 3.47,
301
+ "eval_loss": 0.5227376222610474,
302
+ "eval_runtime": 1759.968,
303
+ "eval_samples_per_second": 4.913,
304
+ "eval_steps_per_second": 0.491,
305
+ "eval_wer": 47.41280944585008,
306
+ "step": 900
307
+ },
308
+ {
309
+ "epoch": 3.57,
310
+ "learning_rate": 9.122613065326634e-06,
311
+ "loss": 0.0494,
312
+ "step": 925
313
+ },
314
+ {
315
+ "epoch": 3.67,
316
+ "learning_rate": 9.09748743718593e-06,
317
+ "loss": 0.051,
318
+ "step": 950
319
+ },
320
+ {
321
+ "epoch": 3.76,
322
+ "learning_rate": 9.072361809045227e-06,
323
+ "loss": 0.0466,
324
+ "step": 975
325
+ },
326
+ {
327
+ "epoch": 3.86,
328
+ "learning_rate": 9.047236180904524e-06,
329
+ "loss": 0.0482,
330
+ "step": 1000
331
+ },
332
+ {
333
+ "epoch": 3.86,
334
+ "eval_loss": 0.5255833268165588,
335
+ "eval_runtime": 1701.3645,
336
+ "eval_samples_per_second": 5.082,
337
+ "eval_steps_per_second": 0.508,
338
+ "eval_wer": 46.94977757933555,
339
+ "step": 1000
340
+ }
341
+ ],
342
+ "logging_steps": 25,
343
+ "max_steps": 10000,
344
+ "num_train_epochs": 39,
345
+ "save_steps": 100,
346
+ "total_flos": 1.61564536369152e+19,
347
+ "trial_name": null,
348
+ "trial_params": null
349
+ }
1/checkpoint-1000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d286c349fc7a4c7419632d8b055a552e2835b4138f3ff268d56552451405f53b
3
+ size 4347
1/checkpoint-10000/config.json ADDED
@@ -0,0 +1,52 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "openai/whisper-small",
3
+ "activation_dropout": 0.0,
4
+ "activation_function": "gelu",
5
+ "apply_spec_augment": false,
6
+ "architectures": [
7
+ "WhisperForConditionalGeneration"
8
+ ],
9
+ "attention_dropout": 0.0,
10
+ "begin_suppress_tokens": [
11
+ 220,
12
+ 50257
13
+ ],
14
+ "bos_token_id": 50257,
15
+ "classifier_proj_size": 256,
16
+ "d_model": 768,
17
+ "decoder_attention_heads": 12,
18
+ "decoder_ffn_dim": 3072,
19
+ "decoder_layerdrop": 0.0,
20
+ "decoder_layers": 12,
21
+ "decoder_start_token_id": 50258,
22
+ "dropout": 0.0,
23
+ "encoder_attention_heads": 12,
24
+ "encoder_ffn_dim": 3072,
25
+ "encoder_layerdrop": 0.0,
26
+ "encoder_layers": 12,
27
+ "eos_token_id": 50257,
28
+ "forced_decoder_ids": null,
29
+ "init_std": 0.02,
30
+ "is_encoder_decoder": true,
31
+ "mask_feature_length": 10,
32
+ "mask_feature_min_masks": 0,
33
+ "mask_feature_prob": 0.0,
34
+ "mask_time_length": 10,
35
+ "mask_time_min_masks": 2,
36
+ "mask_time_prob": 0.05,
37
+ "max_length": 448,
38
+ "max_source_positions": 1500,
39
+ "max_target_positions": 448,
40
+ "median_filter_width": 7,
41
+ "model_type": "whisper",
42
+ "num_hidden_layers": 12,
43
+ "num_mel_bins": 80,
44
+ "pad_token_id": 50257,
45
+ "scale_embedding": false,
46
+ "suppress_tokens": [],
47
+ "torch_dtype": "float32",
48
+ "transformers_version": "4.35.0",
49
+ "use_cache": true,
50
+ "use_weighted_layer_sum": false,
51
+ "vocab_size": 51865
52
+ }
1/checkpoint-10000/generation_config.json ADDED
@@ -0,0 +1,263 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "alignment_heads": [
3
+ [
4
+ 5,
5
+ 3
6
+ ],
7
+ [
8
+ 5,
9
+ 9
10
+ ],
11
+ [
12
+ 8,
13
+ 0
14
+ ],
15
+ [
16
+ 8,
17
+ 4
18
+ ],
19
+ [
20
+ 8,
21
+ 7
22
+ ],
23
+ [
24
+ 8,
25
+ 8
26
+ ],
27
+ [
28
+ 9,
29
+ 0
30
+ ],
31
+ [
32
+ 9,
33
+ 7
34
+ ],
35
+ [
36
+ 9,
37
+ 9
38
+ ],
39
+ [
40
+ 10,
41
+ 5
42
+ ]
43
+ ],
44
+ "begin_suppress_tokens": [
45
+ 220,
46
+ 50257
47
+ ],
48
+ "bos_token_id": 50257,
49
+ "decoder_start_token_id": 50258,
50
+ "eos_token_id": 50257,
51
+ "forced_decoder_ids": [
52
+ [
53
+ 1,
54
+ null
55
+ ],
56
+ [
57
+ 2,
58
+ 50359
59
+ ]
60
+ ],
61
+ "is_multilingual": true,
62
+ "lang_to_id": {
63
+ "<|af|>": 50327,
64
+ "<|am|>": 50334,
65
+ "<|ar|>": 50272,
66
+ "<|as|>": 50350,
67
+ "<|az|>": 50304,
68
+ "<|ba|>": 50355,
69
+ "<|be|>": 50330,
70
+ "<|bg|>": 50292,
71
+ "<|bn|>": 50302,
72
+ "<|bo|>": 50347,
73
+ "<|br|>": 50309,
74
+ "<|bs|>": 50315,
75
+ "<|ca|>": 50270,
76
+ "<|cs|>": 50283,
77
+ "<|cy|>": 50297,
78
+ "<|da|>": 50285,
79
+ "<|de|>": 50261,
80
+ "<|el|>": 50281,
81
+ "<|en|>": 50259,
82
+ "<|es|>": 50262,
83
+ "<|et|>": 50307,
84
+ "<|eu|>": 50310,
85
+ "<|fa|>": 50300,
86
+ "<|fi|>": 50277,
87
+ "<|fo|>": 50338,
88
+ "<|fr|>": 50265,
89
+ "<|gl|>": 50319,
90
+ "<|gu|>": 50333,
91
+ "<|haw|>": 50352,
92
+ "<|ha|>": 50354,
93
+ "<|he|>": 50279,
94
+ "<|hi|>": 50276,
95
+ "<|hr|>": 50291,
96
+ "<|ht|>": 50339,
97
+ "<|hu|>": 50286,
98
+ "<|hy|>": 50312,
99
+ "<|id|>": 50275,
100
+ "<|is|>": 50311,
101
+ "<|it|>": 50274,
102
+ "<|ja|>": 50266,
103
+ "<|jw|>": 50356,
104
+ "<|ka|>": 50329,
105
+ "<|kk|>": 50316,
106
+ "<|km|>": 50323,
107
+ "<|kn|>": 50306,
108
+ "<|ko|>": 50264,
109
+ "<|la|>": 50294,
110
+ "<|lb|>": 50345,
111
+ "<|ln|>": 50353,
112
+ "<|lo|>": 50336,
113
+ "<|lt|>": 50293,
114
+ "<|lv|>": 50301,
115
+ "<|mg|>": 50349,
116
+ "<|mi|>": 50295,
117
+ "<|mk|>": 50308,
118
+ "<|ml|>": 50296,
119
+ "<|mn|>": 50314,
120
+ "<|mr|>": 50320,
121
+ "<|ms|>": 50282,
122
+ "<|mt|>": 50343,
123
+ "<|my|>": 50346,
124
+ "<|ne|>": 50313,
125
+ "<|nl|>": 50271,
126
+ "<|nn|>": 50342,
127
+ "<|no|>": 50288,
128
+ "<|oc|>": 50328,
129
+ "<|pa|>": 50321,
130
+ "<|pl|>": 50269,
131
+ "<|ps|>": 50340,
132
+ "<|pt|>": 50267,
133
+ "<|ro|>": 50284,
134
+ "<|ru|>": 50263,
135
+ "<|sa|>": 50344,
136
+ "<|sd|>": 50332,
137
+ "<|si|>": 50322,
138
+ "<|sk|>": 50298,
139
+ "<|sl|>": 50305,
140
+ "<|sn|>": 50324,
141
+ "<|so|>": 50326,
142
+ "<|sq|>": 50317,
143
+ "<|sr|>": 50303,
144
+ "<|su|>": 50357,
145
+ "<|sv|>": 50273,
146
+ "<|sw|>": 50318,
147
+ "<|ta|>": 50287,
148
+ "<|te|>": 50299,
149
+ "<|tg|>": 50331,
150
+ "<|th|>": 50289,
151
+ "<|tk|>": 50341,
152
+ "<|tl|>": 50348,
153
+ "<|tr|>": 50268,
154
+ "<|tt|>": 50351,
155
+ "<|uk|>": 50280,
156
+ "<|ur|>": 50290,
157
+ "<|uz|>": 50337,
158
+ "<|vi|>": 50278,
159
+ "<|yi|>": 50335,
160
+ "<|yo|>": 50325,
161
+ "<|zh|>": 50260
162
+ },
163
+ "max_initial_timestamp_index": 1,
164
+ "max_length": 448,
165
+ "no_timestamps_token_id": 50363,
166
+ "pad_token_id": 50257,
167
+ "return_timestamps": false,
168
+ "suppress_tokens": [
169
+ 1,
170
+ 2,
171
+ 7,
172
+ 8,
173
+ 9,
174
+ 10,
175
+ 14,
176
+ 25,
177
+ 26,
178
+ 27,
179
+ 28,
180
+ 29,
181
+ 31,
182
+ 58,
183
+ 59,
184
+ 60,
185
+ 61,
186
+ 62,
187
+ 63,
188
+ 90,
189
+ 91,
190
+ 92,
191
+ 93,
192
+ 359,
193
+ 503,
194
+ 522,
195
+ 542,
196
+ 873,
197
+ 893,
198
+ 902,
199
+ 918,
200
+ 922,
201
+ 931,
202
+ 1350,
203
+ 1853,
204
+ 1982,
205
+ 2460,
206
+ 2627,
207
+ 3246,
208
+ 3253,
209
+ 3268,
210
+ 3536,
211
+ 3846,
212
+ 3961,
213
+ 4183,
214
+ 4667,
215
+ 6585,
216
+ 6647,
217
+ 7273,
218
+ 9061,
219
+ 9383,
220
+ 10428,
221
+ 10929,
222
+ 11938,
223
+ 12033,
224
+ 12331,
225
+ 12562,
226
+ 13793,
227
+ 14157,
228
+ 14635,
229
+ 15265,
230
+ 15618,
231
+ 16553,
232
+ 16604,
233
+ 18362,
234
+ 18956,
235
+ 20075,
236
+ 21675,
237
+ 22520,
238
+ 26130,
239
+ 26161,
240
+ 26435,
241
+ 28279,
242
+ 29464,
243
+ 31650,
244
+ 32302,
245
+ 32470,
246
+ 36865,
247
+ 42863,
248
+ 47425,
249
+ 49870,
250
+ 50254,
251
+ 50258,
252
+ 50358,
253
+ 50359,
254
+ 50360,
255
+ 50361,
256
+ 50362
257
+ ],
258
+ "task_to_id": {
259
+ "transcribe": 50359,
260
+ "translate": 50358
261
+ },
262
+ "transformers_version": "4.35.0"
263
+ }
1/checkpoint-10000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:db9f54d47086b50ab47a1a940d82c330373eb42c69fe71086309412a14f7c6e1
3
+ size 966995080
1/checkpoint-10000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c2fd40b5660a0fd37b41a4a22a15ccae85a8c88fac1944241f9baf7f0d2f2528
3
+ size 1925063607
1/checkpoint-10000/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
+ }
1/checkpoint-10000/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:80f4b0ceeeac9fe9ccd8b3441e974861b710db5cc3db22f4aa638f4a927b1279
3
+ size 14575
1/checkpoint-10000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ff12ceee7a94c2b97b487b8bb01785e98be89ac05275cf630ed4d6b88ac367c6
3
+ size 627
1/checkpoint-10000/trainer_state.json ADDED
@@ -0,0 +1,3319 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": 45.965834862992175,
3
+ "best_model_checkpoint": "/media/makhataei/Backups/Whisper-Small-Common-Voice/checkpoint-6200",
4
+ "epoch": 38.61003861003861,
5
+ "eval_steps": 100,
6
+ "global_step": 10000,
7
+ "is_hyper_param_search": false,
8
+ "is_local_process_zero": true,
9
+ "is_world_process_zero": true,
10
+ "log_history": [
11
+ {
12
+ "epoch": 0.1,
13
+ "learning_rate": 4.600000000000001e-06,
14
+ "loss": 3.1025,
15
+ "step": 25
16
+ },
17
+ {
18
+ "epoch": 0.19,
19
+ "learning_rate": 9.600000000000001e-06,
20
+ "loss": 1.3525,
21
+ "step": 50
22
+ },
23
+ {
24
+ "epoch": 0.29,
25
+ "learning_rate": 9.976884422110554e-06,
26
+ "loss": 0.9564,
27
+ "step": 75
28
+ },
29
+ {
30
+ "epoch": 0.39,
31
+ "learning_rate": 9.95175879396985e-06,
32
+ "loss": 0.7495,
33
+ "step": 100
34
+ },
35
+ {
36
+ "epoch": 0.39,
37
+ "eval_loss": 0.860159158706665,
38
+ "eval_runtime": 1426.2695,
39
+ "eval_samples_per_second": 6.062,
40
+ "eval_steps_per_second": 0.606,
41
+ "eval_wer": 58.07742554282218,
42
+ "step": 100
43
+ },
44
+ {
45
+ "epoch": 0.48,
46
+ "learning_rate": 9.926633165829147e-06,
47
+ "loss": 0.5073,
48
+ "step": 125
49
+ },
50
+ {
51
+ "epoch": 0.58,
52
+ "learning_rate": 9.901507537688444e-06,
53
+ "loss": 0.3749,
54
+ "step": 150
55
+ },
56
+ {
57
+ "epoch": 0.68,
58
+ "learning_rate": 9.87638190954774e-06,
59
+ "loss": 0.3247,
60
+ "step": 175
61
+ },
62
+ {
63
+ "epoch": 0.77,
64
+ "learning_rate": 9.851256281407035e-06,
65
+ "loss": 0.3059,
66
+ "step": 200
67
+ },
68
+ {
69
+ "epoch": 0.77,
70
+ "eval_loss": 0.5538379549980164,
71
+ "eval_runtime": 1589.2333,
72
+ "eval_samples_per_second": 5.44,
73
+ "eval_steps_per_second": 0.544,
74
+ "eval_wer": 55.22481850804518,
75
+ "step": 200
76
+ },
77
+ {
78
+ "epoch": 0.87,
79
+ "learning_rate": 9.826130653266333e-06,
80
+ "loss": 0.2963,
81
+ "step": 225
82
+ },
83
+ {
84
+ "epoch": 0.97,
85
+ "learning_rate": 9.80100502512563e-06,
86
+ "loss": 0.2882,
87
+ "step": 250
88
+ },
89
+ {
90
+ "epoch": 1.06,
91
+ "learning_rate": 9.775879396984925e-06,
92
+ "loss": 0.2327,
93
+ "step": 275
94
+ },
95
+ {
96
+ "epoch": 1.16,
97
+ "learning_rate": 9.750753768844222e-06,
98
+ "loss": 0.2002,
99
+ "step": 300
100
+ },
101
+ {
102
+ "epoch": 1.16,
103
+ "eval_loss": 0.5136818885803223,
104
+ "eval_runtime": 1646.1357,
105
+ "eval_samples_per_second": 5.252,
106
+ "eval_steps_per_second": 0.525,
107
+ "eval_wer": 52.46151047609599,
108
+ "step": 300
109
+ },
110
+ {
111
+ "epoch": 1.25,
112
+ "learning_rate": 9.725628140703518e-06,
113
+ "loss": 0.1893,
114
+ "step": 325
115
+ },
116
+ {
117
+ "epoch": 1.35,
118
+ "learning_rate": 9.700502512562815e-06,
119
+ "loss": 0.185,
120
+ "step": 350
121
+ },
122
+ {
123
+ "epoch": 1.45,
124
+ "learning_rate": 9.675376884422111e-06,
125
+ "loss": 0.1847,
126
+ "step": 375
127
+ },
128
+ {
129
+ "epoch": 1.54,
130
+ "learning_rate": 9.650251256281408e-06,
131
+ "loss": 0.1798,
132
+ "step": 400
133
+ },
134
+ {
135
+ "epoch": 1.54,
136
+ "eval_loss": 0.5102902054786682,
137
+ "eval_runtime": 1650.3734,
138
+ "eval_samples_per_second": 5.239,
139
+ "eval_steps_per_second": 0.524,
140
+ "eval_wer": 51.65781945064576,
141
+ "step": 400
142
+ },
143
+ {
144
+ "epoch": 1.64,
145
+ "learning_rate": 9.625125628140705e-06,
146
+ "loss": 0.1804,
147
+ "step": 425
148
+ },
149
+ {
150
+ "epoch": 1.74,
151
+ "learning_rate": 9.600000000000001e-06,
152
+ "loss": 0.1807,
153
+ "step": 450
154
+ },
155
+ {
156
+ "epoch": 1.83,
157
+ "learning_rate": 9.574874371859298e-06,
158
+ "loss": 0.1714,
159
+ "step": 475
160
+ },
161
+ {
162
+ "epoch": 1.93,
163
+ "learning_rate": 9.549748743718595e-06,
164
+ "loss": 0.1678,
165
+ "step": 500
166
+ },
167
+ {
168
+ "epoch": 1.93,
169
+ "eval_loss": 0.48063144087791443,
170
+ "eval_runtime": 1482.5018,
171
+ "eval_samples_per_second": 5.832,
172
+ "eval_steps_per_second": 0.583,
173
+ "eval_wer": 46.55123943708554,
174
+ "step": 500
175
+ },
176
+ {
177
+ "epoch": 2.03,
178
+ "learning_rate": 9.52462311557789e-06,
179
+ "loss": 0.1542,
180
+ "step": 525
181
+ },
182
+ {
183
+ "epoch": 2.12,
184
+ "learning_rate": 9.499497487437188e-06,
185
+ "loss": 0.0929,
186
+ "step": 550
187
+ },
188
+ {
189
+ "epoch": 2.22,
190
+ "learning_rate": 9.474371859296483e-06,
191
+ "loss": 0.0936,
192
+ "step": 575
193
+ },
194
+ {
195
+ "epoch": 2.32,
196
+ "learning_rate": 9.44924623115578e-06,
197
+ "loss": 0.0932,
198
+ "step": 600
199
+ },
200
+ {
201
+ "epoch": 2.32,
202
+ "eval_loss": 0.49363330006599426,
203
+ "eval_runtime": 1993.4903,
204
+ "eval_samples_per_second": 4.337,
205
+ "eval_steps_per_second": 0.434,
206
+ "eval_wer": 47.49549370772767,
207
+ "step": 600
208
+ },
209
+ {
210
+ "epoch": 2.41,
211
+ "learning_rate": 9.424120603015076e-06,
212
+ "loss": 0.0936,
213
+ "step": 625
214
+ },
215
+ {
216
+ "epoch": 2.51,
217
+ "learning_rate": 9.398994974874373e-06,
218
+ "loss": 0.0955,
219
+ "step": 650
220
+ },
221
+ {
222
+ "epoch": 2.61,
223
+ "learning_rate": 9.373869346733669e-06,
224
+ "loss": 0.0871,
225
+ "step": 675
226
+ },
227
+ {
228
+ "epoch": 2.7,
229
+ "learning_rate": 9.348743718592966e-06,
230
+ "loss": 0.0944,
231
+ "step": 700
232
+ },
233
+ {
234
+ "epoch": 2.7,
235
+ "eval_loss": 0.5119881629943848,
236
+ "eval_runtime": 1791.3192,
237
+ "eval_samples_per_second": 4.827,
238
+ "eval_steps_per_second": 0.483,
239
+ "eval_wer": 48.91600932678474,
240
+ "step": 700
241
+ },
242
+ {
243
+ "epoch": 2.8,
244
+ "learning_rate": 9.323618090452262e-06,
245
+ "loss": 0.0927,
246
+ "step": 725
247
+ },
248
+ {
249
+ "epoch": 2.9,
250
+ "learning_rate": 9.298492462311557e-06,
251
+ "loss": 0.0964,
252
+ "step": 750
253
+ },
254
+ {
255
+ "epoch": 2.99,
256
+ "learning_rate": 9.273366834170856e-06,
257
+ "loss": 0.1038,
258
+ "step": 775
259
+ },
260
+ {
261
+ "epoch": 3.09,
262
+ "learning_rate": 9.248241206030152e-06,
263
+ "loss": 0.0493,
264
+ "step": 800
265
+ },
266
+ {
267
+ "epoch": 3.09,
268
+ "eval_loss": 0.5198288559913635,
269
+ "eval_runtime": 1878.4077,
270
+ "eval_samples_per_second": 4.603,
271
+ "eval_steps_per_second": 0.46,
272
+ "eval_wer": 48.26114997271419,
273
+ "step": 800
274
+ },
275
+ {
276
+ "epoch": 3.19,
277
+ "learning_rate": 9.223115577889447e-06,
278
+ "loss": 0.0501,
279
+ "step": 825
280
+ },
281
+ {
282
+ "epoch": 3.28,
283
+ "learning_rate": 9.197989949748744e-06,
284
+ "loss": 0.0458,
285
+ "step": 850
286
+ },
287
+ {
288
+ "epoch": 3.38,
289
+ "learning_rate": 9.17286432160804e-06,
290
+ "loss": 0.0492,
291
+ "step": 875
292
+ },
293
+ {
294
+ "epoch": 3.47,
295
+ "learning_rate": 9.147738693467337e-06,
296
+ "loss": 0.0506,
297
+ "step": 900
298
+ },
299
+ {
300
+ "epoch": 3.47,
301
+ "eval_loss": 0.5227376222610474,
302
+ "eval_runtime": 1759.968,
303
+ "eval_samples_per_second": 4.913,
304
+ "eval_steps_per_second": 0.491,
305
+ "eval_wer": 47.41280944585008,
306
+ "step": 900
307
+ },
308
+ {
309
+ "epoch": 3.57,
310
+ "learning_rate": 9.122613065326634e-06,
311
+ "loss": 0.0494,
312
+ "step": 925
313
+ },
314
+ {
315
+ "epoch": 3.67,
316
+ "learning_rate": 9.09748743718593e-06,
317
+ "loss": 0.051,
318
+ "step": 950
319
+ },
320
+ {
321
+ "epoch": 3.76,
322
+ "learning_rate": 9.072361809045227e-06,
323
+ "loss": 0.0466,
324
+ "step": 975
325
+ },
326
+ {
327
+ "epoch": 3.86,
328
+ "learning_rate": 9.047236180904524e-06,
329
+ "loss": 0.0482,
330
+ "step": 1000
331
+ },
332
+ {
333
+ "epoch": 3.86,
334
+ "eval_loss": 0.5255833268165588,
335
+ "eval_runtime": 1701.3645,
336
+ "eval_samples_per_second": 5.082,
337
+ "eval_steps_per_second": 0.508,
338
+ "eval_wer": 46.94977757933555,
339
+ "step": 1000
340
+ },
341
+ {
342
+ "epoch": 3.96,
343
+ "learning_rate": 9.02211055276382e-06,
344
+ "loss": 0.046,
345
+ "step": 1025
346
+ },
347
+ {
348
+ "epoch": 4.05,
349
+ "learning_rate": 8.996984924623117e-06,
350
+ "loss": 0.0368,
351
+ "step": 1050
352
+ },
353
+ {
354
+ "epoch": 4.15,
355
+ "learning_rate": 8.971859296482412e-06,
356
+ "loss": 0.0233,
357
+ "step": 1075
358
+ },
359
+ {
360
+ "epoch": 4.25,
361
+ "learning_rate": 8.94673366834171e-06,
362
+ "loss": 0.0244,
363
+ "step": 1100
364
+ },
365
+ {
366
+ "epoch": 4.25,
367
+ "eval_loss": 0.5491495132446289,
368
+ "eval_runtime": 1619.0836,
369
+ "eval_samples_per_second": 5.34,
370
+ "eval_steps_per_second": 0.534,
371
+ "eval_wer": 46.084900200095916,
372
+ "step": 1100
373
+ },
374
+ {
375
+ "epoch": 4.34,
376
+ "learning_rate": 8.921608040201005e-06,
377
+ "loss": 0.0258,
378
+ "step": 1125
379
+ },
380
+ {
381
+ "epoch": 4.44,
382
+ "learning_rate": 8.896482412060302e-06,
383
+ "loss": 0.0248,
384
+ "step": 1150
385
+ },
386
+ {
387
+ "epoch": 4.54,
388
+ "learning_rate": 8.8713567839196e-06,
389
+ "loss": 0.0241,
390
+ "step": 1175
391
+ },
392
+ {
393
+ "epoch": 4.63,
394
+ "learning_rate": 8.846231155778895e-06,
395
+ "loss": 0.0272,
396
+ "step": 1200
397
+ },
398
+ {
399
+ "epoch": 4.63,
400
+ "eval_loss": 0.5620735883712769,
401
+ "eval_runtime": 1837.7585,
402
+ "eval_samples_per_second": 4.705,
403
+ "eval_steps_per_second": 0.471,
404
+ "eval_wer": 46.916703874584506,
405
+ "step": 1200
406
+ },
407
+ {
408
+ "epoch": 4.73,
409
+ "learning_rate": 8.821105527638191e-06,
410
+ "loss": 0.0258,
411
+ "step": 1225
412
+ },
413
+ {
414
+ "epoch": 4.83,
415
+ "learning_rate": 8.795979899497488e-06,
416
+ "loss": 0.0237,
417
+ "step": 1250
418
+ },
419
+ {
420
+ "epoch": 4.92,
421
+ "learning_rate": 8.770854271356785e-06,
422
+ "loss": 0.0266,
423
+ "step": 1275
424
+ },
425
+ {
426
+ "epoch": 5.02,
427
+ "learning_rate": 8.745728643216081e-06,
428
+ "loss": 0.0246,
429
+ "step": 1300
430
+ },
431
+ {
432
+ "epoch": 5.02,
433
+ "eval_loss": 0.5855526328086853,
434
+ "eval_runtime": 1908.9589,
435
+ "eval_samples_per_second": 4.529,
436
+ "eval_steps_per_second": 0.453,
437
+ "eval_wer": 48.80355873063122,
438
+ "step": 1300
439
+ },
440
+ {
441
+ "epoch": 5.12,
442
+ "learning_rate": 8.720603015075378e-06,
443
+ "loss": 0.0142,
444
+ "step": 1325
445
+ },
446
+ {
447
+ "epoch": 5.21,
448
+ "learning_rate": 8.695477386934675e-06,
449
+ "loss": 0.014,
450
+ "step": 1350
451
+ },
452
+ {
453
+ "epoch": 5.31,
454
+ "learning_rate": 8.67035175879397e-06,
455
+ "loss": 0.0126,
456
+ "step": 1375
457
+ },
458
+ {
459
+ "epoch": 5.41,
460
+ "learning_rate": 8.645226130653268e-06,
461
+ "loss": 0.0159,
462
+ "step": 1400
463
+ },
464
+ {
465
+ "epoch": 5.41,
466
+ "eval_loss": 0.6145116090774536,
467
+ "eval_runtime": 1843.4778,
468
+ "eval_samples_per_second": 4.69,
469
+ "eval_steps_per_second": 0.469,
470
+ "eval_wer": 48.30414578889054,
471
+ "step": 1400
472
+ },
473
+ {
474
+ "epoch": 5.5,
475
+ "learning_rate": 8.620100502512564e-06,
476
+ "loss": 0.0144,
477
+ "step": 1425
478
+ },
479
+ {
480
+ "epoch": 5.6,
481
+ "learning_rate": 8.59497487437186e-06,
482
+ "loss": 0.0133,
483
+ "step": 1450
484
+ },
485
+ {
486
+ "epoch": 5.69,
487
+ "learning_rate": 8.569849246231156e-06,
488
+ "loss": 0.0151,
489
+ "step": 1475
490
+ },
491
+ {
492
+ "epoch": 5.79,
493
+ "learning_rate": 8.544723618090453e-06,
494
+ "loss": 0.0152,
495
+ "step": 1500
496
+ },
497
+ {
498
+ "epoch": 5.79,
499
+ "eval_loss": 0.6177700757980347,
500
+ "eval_runtime": 1839.895,
501
+ "eval_samples_per_second": 4.699,
502
+ "eval_steps_per_second": 0.47,
503
+ "eval_wer": 47.756775975260865,
504
+ "step": 1500
505
+ },
506
+ {
507
+ "epoch": 5.89,
508
+ "learning_rate": 8.51959798994975e-06,
509
+ "loss": 0.014,
510
+ "step": 1525
511
+ },
512
+ {
513
+ "epoch": 5.98,
514
+ "learning_rate": 8.494472361809046e-06,
515
+ "loss": 0.016,
516
+ "step": 1550
517
+ },
518
+ {
519
+ "epoch": 6.08,
520
+ "learning_rate": 8.469346733668342e-06,
521
+ "loss": 0.0092,
522
+ "step": 1575
523
+ },
524
+ {
525
+ "epoch": 6.18,
526
+ "learning_rate": 8.444221105527639e-06,
527
+ "loss": 0.008,
528
+ "step": 1600
529
+ },
530
+ {
531
+ "epoch": 6.18,
532
+ "eval_loss": 0.6190542578697205,
533
+ "eval_runtime": 1730.2546,
534
+ "eval_samples_per_second": 4.997,
535
+ "eval_steps_per_second": 0.5,
536
+ "eval_wer": 48.06105405897041,
537
+ "step": 1600
538
+ },
539
+ {
540
+ "epoch": 6.27,
541
+ "learning_rate": 8.419095477386936e-06,
542
+ "loss": 0.0079,
543
+ "step": 1625
544
+ },
545
+ {
546
+ "epoch": 6.37,
547
+ "learning_rate": 8.393969849246232e-06,
548
+ "loss": 0.0087,
549
+ "step": 1650
550
+ },
551
+ {
552
+ "epoch": 6.47,
553
+ "learning_rate": 8.368844221105529e-06,
554
+ "loss": 0.0082,
555
+ "step": 1675
556
+ },
557
+ {
558
+ "epoch": 6.56,
559
+ "learning_rate": 8.343718592964824e-06,
560
+ "loss": 0.0077,
561
+ "step": 1700
562
+ },
563
+ {
564
+ "epoch": 6.56,
565
+ "eval_loss": 0.6308984160423279,
566
+ "eval_runtime": 1686.6544,
567
+ "eval_samples_per_second": 5.126,
568
+ "eval_steps_per_second": 0.513,
569
+ "eval_wer": 46.35941194952953,
570
+ "step": 1700
571
+ },
572
+ {
573
+ "epoch": 6.66,
574
+ "learning_rate": 8.318592964824122e-06,
575
+ "loss": 0.0076,
576
+ "step": 1725
577
+ },
578
+ {
579
+ "epoch": 6.76,
580
+ "learning_rate": 8.293467336683417e-06,
581
+ "loss": 0.0095,
582
+ "step": 1750
583
+ },
584
+ {
585
+ "epoch": 6.85,
586
+ "learning_rate": 8.268341708542714e-06,
587
+ "loss": 0.0087,
588
+ "step": 1775
589
+ },
590
+ {
591
+ "epoch": 6.95,
592
+ "learning_rate": 8.24321608040201e-06,
593
+ "loss": 0.0097,
594
+ "step": 1800
595
+ },
596
+ {
597
+ "epoch": 6.95,
598
+ "eval_loss": 0.6272400617599487,
599
+ "eval_runtime": 1856.6423,
600
+ "eval_samples_per_second": 4.657,
601
+ "eval_steps_per_second": 0.466,
602
+ "eval_wer": 47.9138760728283,
603
+ "step": 1800
604
+ },
605
+ {
606
+ "epoch": 7.05,
607
+ "learning_rate": 8.218090452261307e-06,
608
+ "loss": 0.009,
609
+ "step": 1825
610
+ },
611
+ {
612
+ "epoch": 7.14,
613
+ "learning_rate": 8.192964824120604e-06,
614
+ "loss": 0.0053,
615
+ "step": 1850
616
+ },
617
+ {
618
+ "epoch": 7.24,
619
+ "learning_rate": 8.1678391959799e-06,
620
+ "loss": 0.0051,
621
+ "step": 1875
622
+ },
623
+ {
624
+ "epoch": 7.34,
625
+ "learning_rate": 8.142713567839197e-06,
626
+ "loss": 0.0056,
627
+ "step": 1900
628
+ },
629
+ {
630
+ "epoch": 7.34,
631
+ "eval_loss": 0.6594315767288208,
632
+ "eval_runtime": 1711.6251,
633
+ "eval_samples_per_second": 5.051,
634
+ "eval_steps_per_second": 0.505,
635
+ "eval_wer": 46.5562004927982,
636
+ "step": 1900
637
+ },
638
+ {
639
+ "epoch": 7.43,
640
+ "learning_rate": 8.117587939698493e-06,
641
+ "loss": 0.0056,
642
+ "step": 1925
643
+ },
644
+ {
645
+ "epoch": 7.53,
646
+ "learning_rate": 8.09246231155779e-06,
647
+ "loss": 0.0058,
648
+ "step": 1950
649
+ },
650
+ {
651
+ "epoch": 7.63,
652
+ "learning_rate": 8.067336683417087e-06,
653
+ "loss": 0.007,
654
+ "step": 1975
655
+ },
656
+ {
657
+ "epoch": 7.72,
658
+ "learning_rate": 8.042211055276382e-06,
659
+ "loss": 0.0049,
660
+ "step": 2000
661
+ },
662
+ {
663
+ "epoch": 7.72,
664
+ "eval_loss": 0.6581406593322754,
665
+ "eval_runtime": 1917.4042,
666
+ "eval_samples_per_second": 4.509,
667
+ "eval_steps_per_second": 0.451,
668
+ "eval_wer": 47.58313902531792,
669
+ "step": 2000
670
+ },
671
+ {
672
+ "epoch": 7.82,
673
+ "learning_rate": 8.01708542713568e-06,
674
+ "loss": 0.0064,
675
+ "step": 2025
676
+ },
677
+ {
678
+ "epoch": 7.92,
679
+ "learning_rate": 7.991959798994977e-06,
680
+ "loss": 0.0055,
681
+ "step": 2050
682
+ },
683
+ {
684
+ "epoch": 8.01,
685
+ "learning_rate": 7.966834170854271e-06,
686
+ "loss": 0.0055,
687
+ "step": 2075
688
+ },
689
+ {
690
+ "epoch": 8.11,
691
+ "learning_rate": 7.941708542713568e-06,
692
+ "loss": 0.0042,
693
+ "step": 2100
694
+ },
695
+ {
696
+ "epoch": 8.11,
697
+ "eval_loss": 0.695335865020752,
698
+ "eval_runtime": 1836.7482,
699
+ "eval_samples_per_second": 4.707,
700
+ "eval_steps_per_second": 0.471,
701
+ "eval_wer": 48.092474078483896,
702
+ "step": 2100
703
+ },
704
+ {
705
+ "epoch": 8.2,
706
+ "learning_rate": 7.916582914572865e-06,
707
+ "loss": 0.004,
708
+ "step": 2125
709
+ },
710
+ {
711
+ "epoch": 8.3,
712
+ "learning_rate": 7.891457286432161e-06,
713
+ "loss": 0.0046,
714
+ "step": 2150
715
+ },
716
+ {
717
+ "epoch": 8.4,
718
+ "learning_rate": 7.866331658291458e-06,
719
+ "loss": 0.004,
720
+ "step": 2175
721
+ },
722
+ {
723
+ "epoch": 8.49,
724
+ "learning_rate": 7.841206030150755e-06,
725
+ "loss": 0.0038,
726
+ "step": 2200
727
+ },
728
+ {
729
+ "epoch": 8.49,
730
+ "eval_loss": 0.6856936812400818,
731
+ "eval_runtime": 1910.3433,
732
+ "eval_samples_per_second": 4.526,
733
+ "eval_steps_per_second": 0.453,
734
+ "eval_wer": 48.371946883630166,
735
+ "step": 2200
736
+ },
737
+ {
738
+ "epoch": 8.59,
739
+ "learning_rate": 7.816080402010051e-06,
740
+ "loss": 0.0036,
741
+ "step": 2225
742
+ },
743
+ {
744
+ "epoch": 8.69,
745
+ "learning_rate": 7.790954773869348e-06,
746
+ "loss": 0.0038,
747
+ "step": 2250
748
+ },
749
+ {
750
+ "epoch": 8.78,
751
+ "learning_rate": 7.765829145728644e-06,
752
+ "loss": 0.0033,
753
+ "step": 2275
754
+ },
755
+ {
756
+ "epoch": 8.88,
757
+ "learning_rate": 7.740703517587941e-06,
758
+ "loss": 0.0033,
759
+ "step": 2300
760
+ },
761
+ {
762
+ "epoch": 8.88,
763
+ "eval_loss": 0.6982905268669128,
764
+ "eval_runtime": 1950.8321,
765
+ "eval_samples_per_second": 4.432,
766
+ "eval_steps_per_second": 0.443,
767
+ "eval_wer": 49.69654875890923,
768
+ "step": 2300
769
+ },
770
+ {
771
+ "epoch": 8.98,
772
+ "learning_rate": 7.715577889447236e-06,
773
+ "loss": 0.0036,
774
+ "step": 2325
775
+ },
776
+ {
777
+ "epoch": 9.07,
778
+ "learning_rate": 7.690452261306534e-06,
779
+ "loss": 0.0029,
780
+ "step": 2350
781
+ },
782
+ {
783
+ "epoch": 9.17,
784
+ "learning_rate": 7.66532663316583e-06,
785
+ "loss": 0.0027,
786
+ "step": 2375
787
+ },
788
+ {
789
+ "epoch": 9.27,
790
+ "learning_rate": 7.640201005025126e-06,
791
+ "loss": 0.003,
792
+ "step": 2400
793
+ },
794
+ {
795
+ "epoch": 9.27,
796
+ "eval_loss": 0.7109295129776001,
797
+ "eval_runtime": 1863.5651,
798
+ "eval_samples_per_second": 4.639,
799
+ "eval_steps_per_second": 0.464,
800
+ "eval_wer": 48.11727935704718,
801
+ "step": 2400
802
+ },
803
+ {
804
+ "epoch": 9.36,
805
+ "learning_rate": 7.615075376884423e-06,
806
+ "loss": 0.003,
807
+ "step": 2425
808
+ },
809
+ {
810
+ "epoch": 9.46,
811
+ "learning_rate": 7.589949748743719e-06,
812
+ "loss": 0.0025,
813
+ "step": 2450
814
+ },
815
+ {
816
+ "epoch": 9.56,
817
+ "learning_rate": 7.564824120603016e-06,
818
+ "loss": 0.0027,
819
+ "step": 2475
820
+ },
821
+ {
822
+ "epoch": 9.65,
823
+ "learning_rate": 7.5396984924623115e-06,
824
+ "loss": 0.0033,
825
+ "step": 2500
826
+ },
827
+ {
828
+ "epoch": 9.65,
829
+ "eval_loss": 0.6899213194847107,
830
+ "eval_runtime": 1934.8822,
831
+ "eval_samples_per_second": 4.468,
832
+ "eval_steps_per_second": 0.447,
833
+ "eval_wer": 48.236344694150915,
834
+ "step": 2500
835
+ },
836
+ {
837
+ "epoch": 9.75,
838
+ "learning_rate": 7.514572864321609e-06,
839
+ "loss": 0.0025,
840
+ "step": 2525
841
+ },
842
+ {
843
+ "epoch": 9.85,
844
+ "learning_rate": 7.4894472361809056e-06,
845
+ "loss": 0.002,
846
+ "step": 2550
847
+ },
848
+ {
849
+ "epoch": 9.94,
850
+ "learning_rate": 7.464321608040201e-06,
851
+ "loss": 0.0031,
852
+ "step": 2575
853
+ },
854
+ {
855
+ "epoch": 10.04,
856
+ "learning_rate": 7.439195979899499e-06,
857
+ "loss": 0.0027,
858
+ "step": 2600
859
+ },
860
+ {
861
+ "epoch": 10.04,
862
+ "eval_loss": 0.7074257135391235,
863
+ "eval_runtime": 1945.5025,
864
+ "eval_samples_per_second": 4.444,
865
+ "eval_steps_per_second": 0.445,
866
+ "eval_wer": 48.34052686411668,
867
+ "step": 2600
868
+ },
869
+ {
870
+ "epoch": 10.14,
871
+ "learning_rate": 7.4140703517587946e-06,
872
+ "loss": 0.0027,
873
+ "step": 2625
874
+ },
875
+ {
876
+ "epoch": 10.23,
877
+ "learning_rate": 7.388944723618091e-06,
878
+ "loss": 0.0029,
879
+ "step": 2650
880
+ },
881
+ {
882
+ "epoch": 10.33,
883
+ "learning_rate": 7.363819095477388e-06,
884
+ "loss": 0.0032,
885
+ "step": 2675
886
+ },
887
+ {
888
+ "epoch": 10.42,
889
+ "learning_rate": 7.338693467336684e-06,
890
+ "loss": 0.0035,
891
+ "step": 2700
892
+ },
893
+ {
894
+ "epoch": 10.42,
895
+ "eval_loss": 0.7017741799354553,
896
+ "eval_runtime": 1835.0488,
897
+ "eval_samples_per_second": 4.712,
898
+ "eval_steps_per_second": 0.471,
899
+ "eval_wer": 47.10026293595277,
900
+ "step": 2700
901
+ },
902
+ {
903
+ "epoch": 10.52,
904
+ "learning_rate": 7.313567839195981e-06,
905
+ "loss": 0.0024,
906
+ "step": 2725
907
+ },
908
+ {
909
+ "epoch": 10.62,
910
+ "learning_rate": 7.288442211055277e-06,
911
+ "loss": 0.0024,
912
+ "step": 2750
913
+ },
914
+ {
915
+ "epoch": 10.71,
916
+ "learning_rate": 7.2633165829145734e-06,
917
+ "loss": 0.0031,
918
+ "step": 2775
919
+ },
920
+ {
921
+ "epoch": 10.81,
922
+ "learning_rate": 7.23819095477387e-06,
923
+ "loss": 0.0026,
924
+ "step": 2800
925
+ },
926
+ {
927
+ "epoch": 10.81,
928
+ "eval_loss": 0.7197884917259216,
929
+ "eval_runtime": 1907.1753,
930
+ "eval_samples_per_second": 4.533,
931
+ "eval_steps_per_second": 0.454,
932
+ "eval_wer": 47.12506821451605,
933
+ "step": 2800
934
+ },
935
+ {
936
+ "epoch": 10.91,
937
+ "learning_rate": 7.213065326633167e-06,
938
+ "loss": 0.0022,
939
+ "step": 2825
940
+ },
941
+ {
942
+ "epoch": 11.0,
943
+ "learning_rate": 7.187939698492463e-06,
944
+ "loss": 0.0027,
945
+ "step": 2850
946
+ },
947
+ {
948
+ "epoch": 11.1,
949
+ "learning_rate": 7.162814070351759e-06,
950
+ "loss": 0.0022,
951
+ "step": 2875
952
+ },
953
+ {
954
+ "epoch": 11.2,
955
+ "learning_rate": 7.137688442211056e-06,
956
+ "loss": 0.0027,
957
+ "step": 2900
958
+ },
959
+ {
960
+ "epoch": 11.2,
961
+ "eval_loss": 0.7459501624107361,
962
+ "eval_runtime": 2001.7998,
963
+ "eval_samples_per_second": 4.319,
964
+ "eval_steps_per_second": 0.432,
965
+ "eval_wer": 48.27603313985216,
966
+ "step": 2900
967
+ },
968
+ {
969
+ "epoch": 11.29,
970
+ "learning_rate": 7.112562814070353e-06,
971
+ "loss": 0.0024,
972
+ "step": 2925
973
+ },
974
+ {
975
+ "epoch": 11.39,
976
+ "learning_rate": 7.087437185929649e-06,
977
+ "loss": 0.0028,
978
+ "step": 2950
979
+ },
980
+ {
981
+ "epoch": 11.49,
982
+ "learning_rate": 7.0623115577889456e-06,
983
+ "loss": 0.0022,
984
+ "step": 2975
985
+ },
986
+ {
987
+ "epoch": 11.58,
988
+ "learning_rate": 7.037185929648241e-06,
989
+ "loss": 0.0023,
990
+ "step": 3000
991
+ },
992
+ {
993
+ "epoch": 11.58,
994
+ "eval_loss": 0.7347891926765442,
995
+ "eval_runtime": 1876.4701,
996
+ "eval_samples_per_second": 4.608,
997
+ "eval_steps_per_second": 0.461,
998
+ "eval_wer": 47.71543384432207,
999
+ "step": 3000
1000
+ },
1001
+ {
1002
+ "epoch": 11.68,
1003
+ "learning_rate": 7.012060301507538e-06,
1004
+ "loss": 0.0022,
1005
+ "step": 3025
1006
+ },
1007
+ {
1008
+ "epoch": 11.78,
1009
+ "learning_rate": 6.9869346733668354e-06,
1010
+ "loss": 0.0028,
1011
+ "step": 3050
1012
+ },
1013
+ {
1014
+ "epoch": 11.87,
1015
+ "learning_rate": 6.961809045226131e-06,
1016
+ "loss": 0.0027,
1017
+ "step": 3075
1018
+ },
1019
+ {
1020
+ "epoch": 11.97,
1021
+ "learning_rate": 6.936683417085428e-06,
1022
+ "loss": 0.0023,
1023
+ "step": 3100
1024
+ },
1025
+ {
1026
+ "epoch": 11.97,
1027
+ "eval_loss": 0.7387125492095947,
1028
+ "eval_runtime": 2084.696,
1029
+ "eval_samples_per_second": 4.147,
1030
+ "eval_steps_per_second": 0.415,
1031
+ "eval_wer": 48.00813613136876,
1032
+ "step": 3100
1033
+ },
1034
+ {
1035
+ "epoch": 12.07,
1036
+ "learning_rate": 6.911557788944724e-06,
1037
+ "loss": 0.002,
1038
+ "step": 3125
1039
+ },
1040
+ {
1041
+ "epoch": 12.16,
1042
+ "learning_rate": 6.886432160804021e-06,
1043
+ "loss": 0.0019,
1044
+ "step": 3150
1045
+ },
1046
+ {
1047
+ "epoch": 12.26,
1048
+ "learning_rate": 6.861306532663318e-06,
1049
+ "loss": 0.0028,
1050
+ "step": 3175
1051
+ },
1052
+ {
1053
+ "epoch": 12.36,
1054
+ "learning_rate": 6.8361809045226135e-06,
1055
+ "loss": 0.0024,
1056
+ "step": 3200
1057
+ },
1058
+ {
1059
+ "epoch": 12.36,
1060
+ "eval_loss": 0.7199321985244751,
1061
+ "eval_runtime": 1992.5899,
1062
+ "eval_samples_per_second": 4.339,
1063
+ "eval_steps_per_second": 0.434,
1064
+ "eval_wer": 46.86874700269551,
1065
+ "step": 3200
1066
+ },
1067
+ {
1068
+ "epoch": 12.45,
1069
+ "learning_rate": 6.81105527638191e-06,
1070
+ "loss": 0.0024,
1071
+ "step": 3225
1072
+ },
1073
+ {
1074
+ "epoch": 12.55,
1075
+ "learning_rate": 6.785929648241206e-06,
1076
+ "loss": 0.0019,
1077
+ "step": 3250
1078
+ },
1079
+ {
1080
+ "epoch": 12.64,
1081
+ "learning_rate": 6.760804020100503e-06,
1082
+ "loss": 0.0024,
1083
+ "step": 3275
1084
+ },
1085
+ {
1086
+ "epoch": 12.74,
1087
+ "learning_rate": 6.735678391959799e-06,
1088
+ "loss": 0.0033,
1089
+ "step": 3300
1090
+ },
1091
+ {
1092
+ "epoch": 12.74,
1093
+ "eval_loss": 0.7249542474746704,
1094
+ "eval_runtime": 2175.7785,
1095
+ "eval_samples_per_second": 3.974,
1096
+ "eval_steps_per_second": 0.398,
1097
+ "eval_wer": 48.68945444924013,
1098
+ "step": 3300
1099
+ },
1100
+ {
1101
+ "epoch": 12.84,
1102
+ "learning_rate": 6.710552763819096e-06,
1103
+ "loss": 0.0021,
1104
+ "step": 3325
1105
+ },
1106
+ {
1107
+ "epoch": 12.93,
1108
+ "learning_rate": 6.685427135678393e-06,
1109
+ "loss": 0.0014,
1110
+ "step": 3350
1111
+ },
1112
+ {
1113
+ "epoch": 13.03,
1114
+ "learning_rate": 6.660301507537689e-06,
1115
+ "loss": 0.0011,
1116
+ "step": 3375
1117
+ },
1118
+ {
1119
+ "epoch": 13.13,
1120
+ "learning_rate": 6.6351758793969856e-06,
1121
+ "loss": 0.0017,
1122
+ "step": 3400
1123
+ },
1124
+ {
1125
+ "epoch": 13.13,
1126
+ "eval_loss": 0.7241615653038025,
1127
+ "eval_runtime": 2420.7565,
1128
+ "eval_samples_per_second": 3.572,
1129
+ "eval_steps_per_second": 0.357,
1130
+ "eval_wer": 49.044996775313784,
1131
+ "step": 3400
1132
+ },
1133
+ {
1134
+ "epoch": 13.22,
1135
+ "learning_rate": 6.610050251256281e-06,
1136
+ "loss": 0.0019,
1137
+ "step": 3425
1138
+ },
1139
+ {
1140
+ "epoch": 13.32,
1141
+ "learning_rate": 6.584924623115578e-06,
1142
+ "loss": 0.0014,
1143
+ "step": 3450
1144
+ },
1145
+ {
1146
+ "epoch": 13.42,
1147
+ "learning_rate": 6.5597989949748754e-06,
1148
+ "loss": 0.0011,
1149
+ "step": 3475
1150
+ },
1151
+ {
1152
+ "epoch": 13.51,
1153
+ "learning_rate": 6.534673366834171e-06,
1154
+ "loss": 0.0016,
1155
+ "step": 3500
1156
+ },
1157
+ {
1158
+ "epoch": 13.51,
1159
+ "eval_loss": 0.7358579039573669,
1160
+ "eval_runtime": 2057.6253,
1161
+ "eval_samples_per_second": 4.202,
1162
+ "eval_steps_per_second": 0.42,
1163
+ "eval_wer": 48.70103024590299,
1164
+ "step": 3500
1165
+ },
1166
+ {
1167
+ "epoch": 13.61,
1168
+ "learning_rate": 6.509547738693468e-06,
1169
+ "loss": 0.002,
1170
+ "step": 3525
1171
+ },
1172
+ {
1173
+ "epoch": 13.71,
1174
+ "learning_rate": 6.484422110552764e-06,
1175
+ "loss": 0.0015,
1176
+ "step": 3550
1177
+ },
1178
+ {
1179
+ "epoch": 13.8,
1180
+ "learning_rate": 6.459296482412061e-06,
1181
+ "loss": 0.0024,
1182
+ "step": 3575
1183
+ },
1184
+ {
1185
+ "epoch": 13.9,
1186
+ "learning_rate": 6.434170854271358e-06,
1187
+ "loss": 0.0022,
1188
+ "step": 3600
1189
+ },
1190
+ {
1191
+ "epoch": 13.9,
1192
+ "eval_loss": 0.7220202088356018,
1193
+ "eval_runtime": 2243.6869,
1194
+ "eval_samples_per_second": 3.853,
1195
+ "eval_steps_per_second": 0.386,
1196
+ "eval_wer": 48.1371235798978,
1197
+ "step": 3600
1198
+ },
1199
+ {
1200
+ "epoch": 14.0,
1201
+ "learning_rate": 6.4090452261306535e-06,
1202
+ "loss": 0.0021,
1203
+ "step": 3625
1204
+ },
1205
+ {
1206
+ "epoch": 14.09,
1207
+ "learning_rate": 6.38391959798995e-06,
1208
+ "loss": 0.0013,
1209
+ "step": 3650
1210
+ },
1211
+ {
1212
+ "epoch": 14.19,
1213
+ "learning_rate": 6.358793969849246e-06,
1214
+ "loss": 0.0014,
1215
+ "step": 3675
1216
+ },
1217
+ {
1218
+ "epoch": 14.29,
1219
+ "learning_rate": 6.333668341708543e-06,
1220
+ "loss": 0.0016,
1221
+ "step": 3700
1222
+ },
1223
+ {
1224
+ "epoch": 14.29,
1225
+ "eval_loss": 0.7430799007415771,
1226
+ "eval_runtime": 1822.7961,
1227
+ "eval_samples_per_second": 4.743,
1228
+ "eval_steps_per_second": 0.475,
1229
+ "eval_wer": 46.75629640654198,
1230
+ "step": 3700
1231
+ },
1232
+ {
1233
+ "epoch": 14.38,
1234
+ "learning_rate": 6.30854271356784e-06,
1235
+ "loss": 0.0008,
1236
+ "step": 3725
1237
+ },
1238
+ {
1239
+ "epoch": 14.48,
1240
+ "learning_rate": 6.283417085427136e-06,
1241
+ "loss": 0.0014,
1242
+ "step": 3750
1243
+ },
1244
+ {
1245
+ "epoch": 14.58,
1246
+ "learning_rate": 6.258291457286433e-06,
1247
+ "loss": 0.001,
1248
+ "step": 3775
1249
+ },
1250
+ {
1251
+ "epoch": 14.67,
1252
+ "learning_rate": 6.233165829145729e-06,
1253
+ "loss": 0.0012,
1254
+ "step": 3800
1255
+ },
1256
+ {
1257
+ "epoch": 14.67,
1258
+ "eval_loss": 0.7563586235046387,
1259
+ "eval_runtime": 2100.7282,
1260
+ "eval_samples_per_second": 4.116,
1261
+ "eval_steps_per_second": 0.412,
1262
+ "eval_wer": 47.014271303600076,
1263
+ "step": 3800
1264
+ },
1265
+ {
1266
+ "epoch": 14.77,
1267
+ "learning_rate": 6.208040201005026e-06,
1268
+ "loss": 0.0017,
1269
+ "step": 3825
1270
+ },
1271
+ {
1272
+ "epoch": 14.86,
1273
+ "learning_rate": 6.182914572864322e-06,
1274
+ "loss": 0.0015,
1275
+ "step": 3850
1276
+ },
1277
+ {
1278
+ "epoch": 14.96,
1279
+ "learning_rate": 6.157788944723618e-06,
1280
+ "loss": 0.0012,
1281
+ "step": 3875
1282
+ },
1283
+ {
1284
+ "epoch": 15.06,
1285
+ "learning_rate": 6.1326633165829155e-06,
1286
+ "loss": 0.0014,
1287
+ "step": 3900
1288
+ },
1289
+ {
1290
+ "epoch": 15.06,
1291
+ "eval_loss": 0.7769792079925537,
1292
+ "eval_runtime": 2049.0992,
1293
+ "eval_samples_per_second": 4.219,
1294
+ "eval_steps_per_second": 0.422,
1295
+ "eval_wer": 47.02419341502539,
1296
+ "step": 3900
1297
+ },
1298
+ {
1299
+ "epoch": 15.15,
1300
+ "learning_rate": 6.107537688442211e-06,
1301
+ "loss": 0.0011,
1302
+ "step": 3925
1303
+ },
1304
+ {
1305
+ "epoch": 15.25,
1306
+ "learning_rate": 6.082412060301508e-06,
1307
+ "loss": 0.001,
1308
+ "step": 3950
1309
+ },
1310
+ {
1311
+ "epoch": 15.35,
1312
+ "learning_rate": 6.057286432160805e-06,
1313
+ "loss": 0.0015,
1314
+ "step": 3975
1315
+ },
1316
+ {
1317
+ "epoch": 15.44,
1318
+ "learning_rate": 6.032160804020101e-06,
1319
+ "loss": 0.0008,
1320
+ "step": 4000
1321
+ },
1322
+ {
1323
+ "epoch": 15.44,
1324
+ "eval_loss": 0.8116338849067688,
1325
+ "eval_runtime": 2088.1418,
1326
+ "eval_samples_per_second": 4.141,
1327
+ "eval_steps_per_second": 0.414,
1328
+ "eval_wer": 48.46951431264573,
1329
+ "step": 4000
1330
+ },
1331
+ {
1332
+ "epoch": 15.54,
1333
+ "learning_rate": 6.007035175879398e-06,
1334
+ "loss": 0.0015,
1335
+ "step": 4025
1336
+ },
1337
+ {
1338
+ "epoch": 15.64,
1339
+ "learning_rate": 5.9819095477386935e-06,
1340
+ "loss": 0.0008,
1341
+ "step": 4050
1342
+ },
1343
+ {
1344
+ "epoch": 15.73,
1345
+ "learning_rate": 5.95678391959799e-06,
1346
+ "loss": 0.0016,
1347
+ "step": 4075
1348
+ },
1349
+ {
1350
+ "epoch": 15.83,
1351
+ "learning_rate": 5.9316582914572876e-06,
1352
+ "loss": 0.0015,
1353
+ "step": 4100
1354
+ },
1355
+ {
1356
+ "epoch": 15.83,
1357
+ "eval_loss": 0.7622503638267517,
1358
+ "eval_runtime": 2010.7151,
1359
+ "eval_samples_per_second": 4.3,
1360
+ "eval_steps_per_second": 0.43,
1361
+ "eval_wer": 48.092474078483896,
1362
+ "step": 4100
1363
+ },
1364
+ {
1365
+ "epoch": 15.93,
1366
+ "learning_rate": 5.906532663316583e-06,
1367
+ "loss": 0.0016,
1368
+ "step": 4125
1369
+ },
1370
+ {
1371
+ "epoch": 16.02,
1372
+ "learning_rate": 5.88140703517588e-06,
1373
+ "loss": 0.002,
1374
+ "step": 4150
1375
+ },
1376
+ {
1377
+ "epoch": 16.12,
1378
+ "learning_rate": 5.856281407035176e-06,
1379
+ "loss": 0.0015,
1380
+ "step": 4175
1381
+ },
1382
+ {
1383
+ "epoch": 16.22,
1384
+ "learning_rate": 5.831155778894473e-06,
1385
+ "loss": 0.002,
1386
+ "step": 4200
1387
+ },
1388
+ {
1389
+ "epoch": 16.22,
1390
+ "eval_loss": 0.7698885202407837,
1391
+ "eval_runtime": 1914.0687,
1392
+ "eval_samples_per_second": 4.517,
1393
+ "eval_steps_per_second": 0.452,
1394
+ "eval_wer": 47.72535595574738,
1395
+ "step": 4200
1396
+ },
1397
+ {
1398
+ "epoch": 16.31,
1399
+ "learning_rate": 5.80603015075377e-06,
1400
+ "loss": 0.0013,
1401
+ "step": 4225
1402
+ },
1403
+ {
1404
+ "epoch": 16.41,
1405
+ "learning_rate": 5.780904522613066e-06,
1406
+ "loss": 0.0017,
1407
+ "step": 4250
1408
+ },
1409
+ {
1410
+ "epoch": 16.51,
1411
+ "learning_rate": 5.755778894472362e-06,
1412
+ "loss": 0.0023,
1413
+ "step": 4275
1414
+ },
1415
+ {
1416
+ "epoch": 16.6,
1417
+ "learning_rate": 5.730653266331658e-06,
1418
+ "loss": 0.001,
1419
+ "step": 4300
1420
+ },
1421
+ {
1422
+ "epoch": 16.6,
1423
+ "eval_loss": 0.7630722522735596,
1424
+ "eval_runtime": 1923.0045,
1425
+ "eval_samples_per_second": 4.496,
1426
+ "eval_steps_per_second": 0.45,
1427
+ "eval_wer": 46.87536174364572,
1428
+ "step": 4300
1429
+ },
1430
+ {
1431
+ "epoch": 16.7,
1432
+ "learning_rate": 5.7055276381909555e-06,
1433
+ "loss": 0.0022,
1434
+ "step": 4325
1435
+ },
1436
+ {
1437
+ "epoch": 16.8,
1438
+ "learning_rate": 5.680402010050252e-06,
1439
+ "loss": 0.0012,
1440
+ "step": 4350
1441
+ },
1442
+ {
1443
+ "epoch": 16.89,
1444
+ "learning_rate": 5.655276381909548e-06,
1445
+ "loss": 0.0018,
1446
+ "step": 4375
1447
+ },
1448
+ {
1449
+ "epoch": 16.99,
1450
+ "learning_rate": 5.6301507537688445e-06,
1451
+ "loss": 0.0009,
1452
+ "step": 4400
1453
+ },
1454
+ {
1455
+ "epoch": 16.99,
1456
+ "eval_loss": 0.7591461539268494,
1457
+ "eval_runtime": 1922.4876,
1458
+ "eval_samples_per_second": 4.497,
1459
+ "eval_steps_per_second": 0.45,
1460
+ "eval_wer": 48.416596385044066,
1461
+ "step": 4400
1462
+ },
1463
+ {
1464
+ "epoch": 17.08,
1465
+ "learning_rate": 5.605025125628141e-06,
1466
+ "loss": 0.0012,
1467
+ "step": 4425
1468
+ },
1469
+ {
1470
+ "epoch": 17.18,
1471
+ "learning_rate": 5.579899497487438e-06,
1472
+ "loss": 0.0009,
1473
+ "step": 4450
1474
+ },
1475
+ {
1476
+ "epoch": 17.28,
1477
+ "learning_rate": 5.554773869346734e-06,
1478
+ "loss": 0.0006,
1479
+ "step": 4475
1480
+ },
1481
+ {
1482
+ "epoch": 17.37,
1483
+ "learning_rate": 5.52964824120603e-06,
1484
+ "loss": 0.0008,
1485
+ "step": 4500
1486
+ },
1487
+ {
1488
+ "epoch": 17.37,
1489
+ "eval_loss": 0.779658317565918,
1490
+ "eval_runtime": 1935.42,
1491
+ "eval_samples_per_second": 4.467,
1492
+ "eval_steps_per_second": 0.447,
1493
+ "eval_wer": 47.17798614211771,
1494
+ "step": 4500
1495
+ },
1496
+ {
1497
+ "epoch": 17.47,
1498
+ "learning_rate": 5.504522613065328e-06,
1499
+ "loss": 0.0022,
1500
+ "step": 4525
1501
+ },
1502
+ {
1503
+ "epoch": 17.57,
1504
+ "learning_rate": 5.479396984924623e-06,
1505
+ "loss": 0.0008,
1506
+ "step": 4550
1507
+ },
1508
+ {
1509
+ "epoch": 17.66,
1510
+ "learning_rate": 5.45427135678392e-06,
1511
+ "loss": 0.0016,
1512
+ "step": 4575
1513
+ },
1514
+ {
1515
+ "epoch": 17.76,
1516
+ "learning_rate": 5.429145728643217e-06,
1517
+ "loss": 0.0008,
1518
+ "step": 4600
1519
+ },
1520
+ {
1521
+ "epoch": 17.76,
1522
+ "eval_loss": 0.7850877046585083,
1523
+ "eval_runtime": 1838.7142,
1524
+ "eval_samples_per_second": 4.702,
1525
+ "eval_steps_per_second": 0.47,
1526
+ "eval_wer": 46.76621851796729,
1527
+ "step": 4600
1528
+ },
1529
+ {
1530
+ "epoch": 17.86,
1531
+ "learning_rate": 5.404020100502513e-06,
1532
+ "loss": 0.0011,
1533
+ "step": 4625
1534
+ },
1535
+ {
1536
+ "epoch": 17.95,
1537
+ "learning_rate": 5.37889447236181e-06,
1538
+ "loss": 0.0007,
1539
+ "step": 4650
1540
+ },
1541
+ {
1542
+ "epoch": 18.05,
1543
+ "learning_rate": 5.353768844221106e-06,
1544
+ "loss": 0.0015,
1545
+ "step": 4675
1546
+ },
1547
+ {
1548
+ "epoch": 18.15,
1549
+ "learning_rate": 5.328643216080402e-06,
1550
+ "loss": 0.0007,
1551
+ "step": 4700
1552
+ },
1553
+ {
1554
+ "epoch": 18.15,
1555
+ "eval_loss": 0.7897337079048157,
1556
+ "eval_runtime": 1943.7204,
1557
+ "eval_samples_per_second": 4.448,
1558
+ "eval_steps_per_second": 0.445,
1559
+ "eval_wer": 48.315721585553405,
1560
+ "step": 4700
1561
+ },
1562
+ {
1563
+ "epoch": 18.24,
1564
+ "learning_rate": 5.3035175879397e-06,
1565
+ "loss": 0.0007,
1566
+ "step": 4725
1567
+ },
1568
+ {
1569
+ "epoch": 18.34,
1570
+ "learning_rate": 5.2783919597989955e-06,
1571
+ "loss": 0.0006,
1572
+ "step": 4750
1573
+ },
1574
+ {
1575
+ "epoch": 18.44,
1576
+ "learning_rate": 5.253266331658292e-06,
1577
+ "loss": 0.0007,
1578
+ "step": 4775
1579
+ },
1580
+ {
1581
+ "epoch": 18.53,
1582
+ "learning_rate": 5.228140703517588e-06,
1583
+ "loss": 0.0006,
1584
+ "step": 4800
1585
+ },
1586
+ {
1587
+ "epoch": 18.53,
1588
+ "eval_loss": 0.7760252952575684,
1589
+ "eval_runtime": 1927.3698,
1590
+ "eval_samples_per_second": 4.486,
1591
+ "eval_steps_per_second": 0.449,
1592
+ "eval_wer": 48.806866101106316,
1593
+ "step": 4800
1594
+ },
1595
+ {
1596
+ "epoch": 18.63,
1597
+ "learning_rate": 5.2030150753768845e-06,
1598
+ "loss": 0.001,
1599
+ "step": 4825
1600
+ },
1601
+ {
1602
+ "epoch": 18.73,
1603
+ "learning_rate": 5.177889447236182e-06,
1604
+ "loss": 0.0008,
1605
+ "step": 4850
1606
+ },
1607
+ {
1608
+ "epoch": 18.82,
1609
+ "learning_rate": 5.152763819095478e-06,
1610
+ "loss": 0.0017,
1611
+ "step": 4875
1612
+ },
1613
+ {
1614
+ "epoch": 18.92,
1615
+ "learning_rate": 5.127638190954774e-06,
1616
+ "loss": 0.0016,
1617
+ "step": 4900
1618
+ },
1619
+ {
1620
+ "epoch": 18.92,
1621
+ "eval_loss": 0.7762993574142456,
1622
+ "eval_runtime": 2000.3946,
1623
+ "eval_samples_per_second": 4.322,
1624
+ "eval_steps_per_second": 0.432,
1625
+ "eval_wer": 47.97175505614261,
1626
+ "step": 4900
1627
+ },
1628
+ {
1629
+ "epoch": 19.02,
1630
+ "learning_rate": 5.10251256281407e-06,
1631
+ "loss": 0.0011,
1632
+ "step": 4925
1633
+ },
1634
+ {
1635
+ "epoch": 19.11,
1636
+ "learning_rate": 5.077386934673368e-06,
1637
+ "loss": 0.0006,
1638
+ "step": 4950
1639
+ },
1640
+ {
1641
+ "epoch": 19.21,
1642
+ "learning_rate": 5.052261306532664e-06,
1643
+ "loss": 0.0006,
1644
+ "step": 4975
1645
+ },
1646
+ {
1647
+ "epoch": 19.31,
1648
+ "learning_rate": 5.02713567839196e-06,
1649
+ "loss": 0.0009,
1650
+ "step": 5000
1651
+ },
1652
+ {
1653
+ "epoch": 19.31,
1654
+ "eval_loss": 0.8151038289070129,
1655
+ "eval_runtime": 1941.3748,
1656
+ "eval_samples_per_second": 4.454,
1657
+ "eval_steps_per_second": 0.446,
1658
+ "eval_wer": 48.158621487985975,
1659
+ "step": 5000
1660
+ },
1661
+ {
1662
+ "epoch": 19.4,
1663
+ "learning_rate": 5.002010050251257e-06,
1664
+ "loss": 0.0008,
1665
+ "step": 5025
1666
+ },
1667
+ {
1668
+ "epoch": 19.5,
1669
+ "learning_rate": 4.976884422110553e-06,
1670
+ "loss": 0.0012,
1671
+ "step": 5050
1672
+ },
1673
+ {
1674
+ "epoch": 19.59,
1675
+ "learning_rate": 4.95175879396985e-06,
1676
+ "loss": 0.001,
1677
+ "step": 5075
1678
+ },
1679
+ {
1680
+ "epoch": 19.69,
1681
+ "learning_rate": 4.9266331658291465e-06,
1682
+ "loss": 0.0004,
1683
+ "step": 5100
1684
+ },
1685
+ {
1686
+ "epoch": 19.69,
1687
+ "eval_loss": 0.7967365384101868,
1688
+ "eval_runtime": 1771.7869,
1689
+ "eval_samples_per_second": 4.88,
1690
+ "eval_steps_per_second": 0.488,
1691
+ "eval_wer": 47.55668006151709,
1692
+ "step": 5100
1693
+ },
1694
+ {
1695
+ "epoch": 19.79,
1696
+ "learning_rate": 4.901507537688442e-06,
1697
+ "loss": 0.0007,
1698
+ "step": 5125
1699
+ },
1700
+ {
1701
+ "epoch": 19.88,
1702
+ "learning_rate": 4.876381909547739e-06,
1703
+ "loss": 0.0007,
1704
+ "step": 5150
1705
+ },
1706
+ {
1707
+ "epoch": 19.98,
1708
+ "learning_rate": 4.8512562814070355e-06,
1709
+ "loss": 0.0009,
1710
+ "step": 5175
1711
+ },
1712
+ {
1713
+ "epoch": 20.08,
1714
+ "learning_rate": 4.826130653266332e-06,
1715
+ "loss": 0.0007,
1716
+ "step": 5200
1717
+ },
1718
+ {
1719
+ "epoch": 20.08,
1720
+ "eval_loss": 0.8094375729560852,
1721
+ "eval_runtime": 1764.7964,
1722
+ "eval_samples_per_second": 4.899,
1723
+ "eval_steps_per_second": 0.49,
1724
+ "eval_wer": 46.78606274081791,
1725
+ "step": 5200
1726
+ },
1727
+ {
1728
+ "epoch": 20.17,
1729
+ "learning_rate": 4.801005025125629e-06,
1730
+ "loss": 0.0007,
1731
+ "step": 5225
1732
+ },
1733
+ {
1734
+ "epoch": 20.27,
1735
+ "learning_rate": 4.7758793969849245e-06,
1736
+ "loss": 0.0012,
1737
+ "step": 5250
1738
+ },
1739
+ {
1740
+ "epoch": 20.37,
1741
+ "learning_rate": 4.750753768844221e-06,
1742
+ "loss": 0.0008,
1743
+ "step": 5275
1744
+ },
1745
+ {
1746
+ "epoch": 20.46,
1747
+ "learning_rate": 4.725628140703518e-06,
1748
+ "loss": 0.001,
1749
+ "step": 5300
1750
+ },
1751
+ {
1752
+ "epoch": 20.46,
1753
+ "eval_loss": 0.8205662369728088,
1754
+ "eval_runtime": 1833.049,
1755
+ "eval_samples_per_second": 4.717,
1756
+ "eval_steps_per_second": 0.472,
1757
+ "eval_wer": 47.15648823402954,
1758
+ "step": 5300
1759
+ },
1760
+ {
1761
+ "epoch": 20.56,
1762
+ "learning_rate": 4.700502512562814e-06,
1763
+ "loss": 0.0013,
1764
+ "step": 5325
1765
+ },
1766
+ {
1767
+ "epoch": 20.66,
1768
+ "learning_rate": 4.675376884422111e-06,
1769
+ "loss": 0.0007,
1770
+ "step": 5350
1771
+ },
1772
+ {
1773
+ "epoch": 20.75,
1774
+ "learning_rate": 4.650251256281408e-06,
1775
+ "loss": 0.0007,
1776
+ "step": 5375
1777
+ },
1778
+ {
1779
+ "epoch": 20.85,
1780
+ "learning_rate": 4.625125628140703e-06,
1781
+ "loss": 0.0008,
1782
+ "step": 5400
1783
+ },
1784
+ {
1785
+ "epoch": 20.85,
1786
+ "eval_loss": 0.8015366792678833,
1787
+ "eval_runtime": 1825.3401,
1788
+ "eval_samples_per_second": 4.737,
1789
+ "eval_steps_per_second": 0.474,
1790
+ "eval_wer": 47.621173785781615,
1791
+ "step": 5400
1792
+ },
1793
+ {
1794
+ "epoch": 20.95,
1795
+ "learning_rate": 4.600000000000001e-06,
1796
+ "loss": 0.0007,
1797
+ "step": 5425
1798
+ },
1799
+ {
1800
+ "epoch": 21.04,
1801
+ "learning_rate": 4.574874371859297e-06,
1802
+ "loss": 0.0006,
1803
+ "step": 5450
1804
+ },
1805
+ {
1806
+ "epoch": 21.14,
1807
+ "learning_rate": 4.549748743718593e-06,
1808
+ "loss": 0.0005,
1809
+ "step": 5475
1810
+ },
1811
+ {
1812
+ "epoch": 21.24,
1813
+ "learning_rate": 4.52462311557789e-06,
1814
+ "loss": 0.0004,
1815
+ "step": 5500
1816
+ },
1817
+ {
1818
+ "epoch": 21.24,
1819
+ "eval_loss": 0.8104465007781982,
1820
+ "eval_runtime": 1735.3093,
1821
+ "eval_samples_per_second": 4.982,
1822
+ "eval_steps_per_second": 0.498,
1823
+ "eval_wer": 47.220981958294054,
1824
+ "step": 5500
1825
+ },
1826
+ {
1827
+ "epoch": 21.33,
1828
+ "learning_rate": 4.499497487437186e-06,
1829
+ "loss": 0.0004,
1830
+ "step": 5525
1831
+ },
1832
+ {
1833
+ "epoch": 21.43,
1834
+ "learning_rate": 4.474371859296483e-06,
1835
+ "loss": 0.0005,
1836
+ "step": 5550
1837
+ },
1838
+ {
1839
+ "epoch": 21.53,
1840
+ "learning_rate": 4.44924623115578e-06,
1841
+ "loss": 0.001,
1842
+ "step": 5575
1843
+ },
1844
+ {
1845
+ "epoch": 21.62,
1846
+ "learning_rate": 4.4241206030150755e-06,
1847
+ "loss": 0.0003,
1848
+ "step": 5600
1849
+ },
1850
+ {
1851
+ "epoch": 21.62,
1852
+ "eval_loss": 0.8020026683807373,
1853
+ "eval_runtime": 1798.7419,
1854
+ "eval_samples_per_second": 4.807,
1855
+ "eval_steps_per_second": 0.481,
1856
+ "eval_wer": 48.145392006085565,
1857
+ "step": 5600
1858
+ },
1859
+ {
1860
+ "epoch": 21.72,
1861
+ "learning_rate": 4.398994974874372e-06,
1862
+ "loss": 0.0008,
1863
+ "step": 5625
1864
+ },
1865
+ {
1866
+ "epoch": 21.81,
1867
+ "learning_rate": 4.373869346733669e-06,
1868
+ "loss": 0.0004,
1869
+ "step": 5650
1870
+ },
1871
+ {
1872
+ "epoch": 21.91,
1873
+ "learning_rate": 4.348743718592965e-06,
1874
+ "loss": 0.0006,
1875
+ "step": 5675
1876
+ },
1877
+ {
1878
+ "epoch": 22.01,
1879
+ "learning_rate": 4.323618090452262e-06,
1880
+ "loss": 0.0004,
1881
+ "step": 5700
1882
+ },
1883
+ {
1884
+ "epoch": 22.01,
1885
+ "eval_loss": 0.8294846415519714,
1886
+ "eval_runtime": 1827.1151,
1887
+ "eval_samples_per_second": 4.732,
1888
+ "eval_steps_per_second": 0.473,
1889
+ "eval_wer": 46.94150915314779,
1890
+ "step": 5700
1891
+ },
1892
+ {
1893
+ "epoch": 22.1,
1894
+ "learning_rate": 4.298492462311558e-06,
1895
+ "loss": 0.0002,
1896
+ "step": 5725
1897
+ },
1898
+ {
1899
+ "epoch": 22.2,
1900
+ "learning_rate": 4.273366834170854e-06,
1901
+ "loss": 0.0003,
1902
+ "step": 5750
1903
+ },
1904
+ {
1905
+ "epoch": 22.3,
1906
+ "learning_rate": 4.248241206030151e-06,
1907
+ "loss": 0.0002,
1908
+ "step": 5775
1909
+ },
1910
+ {
1911
+ "epoch": 22.39,
1912
+ "learning_rate": 4.223115577889448e-06,
1913
+ "loss": 0.0004,
1914
+ "step": 5800
1915
+ },
1916
+ {
1917
+ "epoch": 22.39,
1918
+ "eval_loss": 0.822780191898346,
1919
+ "eval_runtime": 1778.1094,
1920
+ "eval_samples_per_second": 4.862,
1921
+ "eval_steps_per_second": 0.486,
1922
+ "eval_wer": 46.61077210563741,
1923
+ "step": 5800
1924
+ },
1925
+ {
1926
+ "epoch": 22.49,
1927
+ "learning_rate": 4.197989949748744e-06,
1928
+ "loss": 0.0003,
1929
+ "step": 5825
1930
+ },
1931
+ {
1932
+ "epoch": 22.59,
1933
+ "learning_rate": 4.172864321608041e-06,
1934
+ "loss": 0.0003,
1935
+ "step": 5850
1936
+ },
1937
+ {
1938
+ "epoch": 22.68,
1939
+ "learning_rate": 4.147738693467337e-06,
1940
+ "loss": 0.0004,
1941
+ "step": 5875
1942
+ },
1943
+ {
1944
+ "epoch": 22.78,
1945
+ "learning_rate": 4.122613065326633e-06,
1946
+ "loss": 0.0005,
1947
+ "step": 5900
1948
+ },
1949
+ {
1950
+ "epoch": 22.78,
1951
+ "eval_loss": 0.8386306166648865,
1952
+ "eval_runtime": 1804.5855,
1953
+ "eval_samples_per_second": 4.791,
1954
+ "eval_steps_per_second": 0.479,
1955
+ "eval_wer": 48.09743513419656,
1956
+ "step": 5900
1957
+ },
1958
+ {
1959
+ "epoch": 22.88,
1960
+ "learning_rate": 4.09748743718593e-06,
1961
+ "loss": 0.0007,
1962
+ "step": 5925
1963
+ },
1964
+ {
1965
+ "epoch": 22.97,
1966
+ "learning_rate": 4.0723618090452265e-06,
1967
+ "loss": 0.0013,
1968
+ "step": 5950
1969
+ },
1970
+ {
1971
+ "epoch": 23.07,
1972
+ "learning_rate": 4.047236180904523e-06,
1973
+ "loss": 0.0008,
1974
+ "step": 5975
1975
+ },
1976
+ {
1977
+ "epoch": 23.17,
1978
+ "learning_rate": 4.02211055276382e-06,
1979
+ "loss": 0.0007,
1980
+ "step": 6000
1981
+ },
1982
+ {
1983
+ "epoch": 23.17,
1984
+ "eval_loss": 0.8414955735206604,
1985
+ "eval_runtime": 1746.5835,
1986
+ "eval_samples_per_second": 4.95,
1987
+ "eval_steps_per_second": 0.495,
1988
+ "eval_wer": 46.61407947611252,
1989
+ "step": 6000
1990
+ },
1991
+ {
1992
+ "epoch": 23.26,
1993
+ "learning_rate": 3.997989949748744e-06,
1994
+ "loss": 0.0004,
1995
+ "step": 6025
1996
+ },
1997
+ {
1998
+ "epoch": 23.36,
1999
+ "learning_rate": 3.97286432160804e-06,
2000
+ "loss": 0.0004,
2001
+ "step": 6050
2002
+ },
2003
+ {
2004
+ "epoch": 23.46,
2005
+ "learning_rate": 3.947738693467337e-06,
2006
+ "loss": 0.0006,
2007
+ "step": 6075
2008
+ },
2009
+ {
2010
+ "epoch": 23.55,
2011
+ "learning_rate": 3.9226130653266335e-06,
2012
+ "loss": 0.0003,
2013
+ "step": 6100
2014
+ },
2015
+ {
2016
+ "epoch": 23.55,
2017
+ "eval_loss": 0.8283448815345764,
2018
+ "eval_runtime": 1775.2653,
2019
+ "eval_samples_per_second": 4.87,
2020
+ "eval_steps_per_second": 0.487,
2021
+ "eval_wer": 46.32633824477848,
2022
+ "step": 6100
2023
+ },
2024
+ {
2025
+ "epoch": 23.65,
2026
+ "learning_rate": 3.89748743718593e-06,
2027
+ "loss": 0.0003,
2028
+ "step": 6125
2029
+ },
2030
+ {
2031
+ "epoch": 23.75,
2032
+ "learning_rate": 3.872361809045227e-06,
2033
+ "loss": 0.0011,
2034
+ "step": 6150
2035
+ },
2036
+ {
2037
+ "epoch": 23.84,
2038
+ "learning_rate": 3.8472361809045225e-06,
2039
+ "loss": 0.0003,
2040
+ "step": 6175
2041
+ },
2042
+ {
2043
+ "epoch": 23.94,
2044
+ "learning_rate": 3.822110552763819e-06,
2045
+ "loss": 0.0005,
2046
+ "step": 6200
2047
+ },
2048
+ {
2049
+ "epoch": 23.94,
2050
+ "eval_loss": 0.8341825604438782,
2051
+ "eval_runtime": 1685.2846,
2052
+ "eval_samples_per_second": 5.13,
2053
+ "eval_steps_per_second": 0.513,
2054
+ "eval_wer": 45.965834862992175,
2055
+ "step": 6200
2056
+ },
2057
+ {
2058
+ "epoch": 24.03,
2059
+ "learning_rate": 3.796984924623116e-06,
2060
+ "loss": 0.0003,
2061
+ "step": 6225
2062
+ },
2063
+ {
2064
+ "epoch": 24.13,
2065
+ "learning_rate": 3.7718592964824124e-06,
2066
+ "loss": 0.0008,
2067
+ "step": 6250
2068
+ },
2069
+ {
2070
+ "epoch": 24.23,
2071
+ "learning_rate": 3.746733668341709e-06,
2072
+ "loss": 0.0003,
2073
+ "step": 6275
2074
+ },
2075
+ {
2076
+ "epoch": 24.32,
2077
+ "learning_rate": 3.721608040201005e-06,
2078
+ "loss": 0.0002,
2079
+ "step": 6300
2080
+ },
2081
+ {
2082
+ "epoch": 24.32,
2083
+ "eval_loss": 0.8378849029541016,
2084
+ "eval_runtime": 1806.9028,
2085
+ "eval_samples_per_second": 4.785,
2086
+ "eval_steps_per_second": 0.479,
2087
+ "eval_wer": 46.624001587537826,
2088
+ "step": 6300
2089
+ },
2090
+ {
2091
+ "epoch": 24.42,
2092
+ "learning_rate": 3.6964824120603014e-06,
2093
+ "loss": 0.0002,
2094
+ "step": 6325
2095
+ },
2096
+ {
2097
+ "epoch": 24.52,
2098
+ "learning_rate": 3.6713567839195984e-06,
2099
+ "loss": 0.0002,
2100
+ "step": 6350
2101
+ },
2102
+ {
2103
+ "epoch": 24.61,
2104
+ "learning_rate": 3.646231155778895e-06,
2105
+ "loss": 0.0004,
2106
+ "step": 6375
2107
+ },
2108
+ {
2109
+ "epoch": 24.71,
2110
+ "learning_rate": 3.6211055276381912e-06,
2111
+ "loss": 0.0005,
2112
+ "step": 6400
2113
+ },
2114
+ {
2115
+ "epoch": 24.71,
2116
+ "eval_loss": 0.8371052742004395,
2117
+ "eval_runtime": 1889.3529,
2118
+ "eval_samples_per_second": 4.576,
2119
+ "eval_steps_per_second": 0.458,
2120
+ "eval_wer": 48.080898281821035,
2121
+ "step": 6400
2122
+ },
2123
+ {
2124
+ "epoch": 24.81,
2125
+ "learning_rate": 3.5959798994974874e-06,
2126
+ "loss": 0.0005,
2127
+ "step": 6425
2128
+ },
2129
+ {
2130
+ "epoch": 24.9,
2131
+ "learning_rate": 3.570854271356784e-06,
2132
+ "loss": 0.0005,
2133
+ "step": 6450
2134
+ },
2135
+ {
2136
+ "epoch": 25.0,
2137
+ "learning_rate": 3.545728643216081e-06,
2138
+ "loss": 0.0002,
2139
+ "step": 6475
2140
+ },
2141
+ {
2142
+ "epoch": 25.1,
2143
+ "learning_rate": 3.5206030150753773e-06,
2144
+ "loss": 0.0002,
2145
+ "step": 6500
2146
+ },
2147
+ {
2148
+ "epoch": 25.1,
2149
+ "eval_loss": 0.8257761597633362,
2150
+ "eval_runtime": 1907.5906,
2151
+ "eval_samples_per_second": 4.532,
2152
+ "eval_steps_per_second": 0.453,
2153
+ "eval_wer": 47.586446395793025,
2154
+ "step": 6500
2155
+ },
2156
+ {
2157
+ "epoch": 25.19,
2158
+ "learning_rate": 3.4954773869346735e-06,
2159
+ "loss": 0.0003,
2160
+ "step": 6525
2161
+ },
2162
+ {
2163
+ "epoch": 25.29,
2164
+ "learning_rate": 3.47035175879397e-06,
2165
+ "loss": 0.0002,
2166
+ "step": 6550
2167
+ },
2168
+ {
2169
+ "epoch": 25.39,
2170
+ "learning_rate": 3.4452261306532663e-06,
2171
+ "loss": 0.0001,
2172
+ "step": 6575
2173
+ },
2174
+ {
2175
+ "epoch": 25.48,
2176
+ "learning_rate": 3.4201005025125634e-06,
2177
+ "loss": 0.0001,
2178
+ "step": 6600
2179
+ },
2180
+ {
2181
+ "epoch": 25.48,
2182
+ "eval_loss": 0.8396378755569458,
2183
+ "eval_runtime": 1829.8555,
2184
+ "eval_samples_per_second": 4.725,
2185
+ "eval_steps_per_second": 0.473,
2186
+ "eval_wer": 47.260670403995306,
2187
+ "step": 6600
2188
+ },
2189
+ {
2190
+ "epoch": 25.58,
2191
+ "learning_rate": 3.3949748743718596e-06,
2192
+ "loss": 0.0001,
2193
+ "step": 6625
2194
+ },
2195
+ {
2196
+ "epoch": 25.68,
2197
+ "learning_rate": 3.369849246231156e-06,
2198
+ "loss": 0.0003,
2199
+ "step": 6650
2200
+ },
2201
+ {
2202
+ "epoch": 25.77,
2203
+ "learning_rate": 3.3447236180904524e-06,
2204
+ "loss": 0.0001,
2205
+ "step": 6675
2206
+ },
2207
+ {
2208
+ "epoch": 25.87,
2209
+ "learning_rate": 3.319597989949749e-06,
2210
+ "loss": 0.0002,
2211
+ "step": 6700
2212
+ },
2213
+ {
2214
+ "epoch": 25.87,
2215
+ "eval_loss": 0.8490523099899292,
2216
+ "eval_runtime": 1855.9322,
2217
+ "eval_samples_per_second": 4.659,
2218
+ "eval_steps_per_second": 0.466,
2219
+ "eval_wer": 47.543450579616675,
2220
+ "step": 6700
2221
+ },
2222
+ {
2223
+ "epoch": 25.97,
2224
+ "learning_rate": 3.2944723618090456e-06,
2225
+ "loss": 0.0002,
2226
+ "step": 6725
2227
+ },
2228
+ {
2229
+ "epoch": 26.06,
2230
+ "learning_rate": 3.2693467336683422e-06,
2231
+ "loss": 0.0001,
2232
+ "step": 6750
2233
+ },
2234
+ {
2235
+ "epoch": 26.16,
2236
+ "learning_rate": 3.2442211055276384e-06,
2237
+ "loss": 0.0001,
2238
+ "step": 6775
2239
+ },
2240
+ {
2241
+ "epoch": 26.25,
2242
+ "learning_rate": 3.219095477386935e-06,
2243
+ "loss": 0.0005,
2244
+ "step": 6800
2245
+ },
2246
+ {
2247
+ "epoch": 26.25,
2248
+ "eval_loss": 0.8443611860275269,
2249
+ "eval_runtime": 1852.0247,
2250
+ "eval_samples_per_second": 4.668,
2251
+ "eval_steps_per_second": 0.467,
2252
+ "eval_wer": 48.99373253294968,
2253
+ "step": 6800
2254
+ },
2255
+ {
2256
+ "epoch": 26.35,
2257
+ "learning_rate": 3.1939698492462312e-06,
2258
+ "loss": 0.0001,
2259
+ "step": 6825
2260
+ },
2261
+ {
2262
+ "epoch": 26.45,
2263
+ "learning_rate": 3.1688442211055283e-06,
2264
+ "loss": 0.0001,
2265
+ "step": 6850
2266
+ },
2267
+ {
2268
+ "epoch": 26.54,
2269
+ "learning_rate": 3.1437185929648245e-06,
2270
+ "loss": 0.0001,
2271
+ "step": 6875
2272
+ },
2273
+ {
2274
+ "epoch": 26.64,
2275
+ "learning_rate": 3.1185929648241207e-06,
2276
+ "loss": 0.0001,
2277
+ "step": 6900
2278
+ },
2279
+ {
2280
+ "epoch": 26.64,
2281
+ "eval_loss": 0.8540186882019043,
2282
+ "eval_runtime": 1842.737,
2283
+ "eval_samples_per_second": 4.692,
2284
+ "eval_steps_per_second": 0.469,
2285
+ "eval_wer": 48.05443931802021,
2286
+ "step": 6900
2287
+ },
2288
+ {
2289
+ "epoch": 26.74,
2290
+ "learning_rate": 3.0934673366834173e-06,
2291
+ "loss": 0.0001,
2292
+ "step": 6925
2293
+ },
2294
+ {
2295
+ "epoch": 26.83,
2296
+ "learning_rate": 3.0683417085427135e-06,
2297
+ "loss": 0.0001,
2298
+ "step": 6950
2299
+ },
2300
+ {
2301
+ "epoch": 26.93,
2302
+ "learning_rate": 3.0432160804020105e-06,
2303
+ "loss": 0.0002,
2304
+ "step": 6975
2305
+ },
2306
+ {
2307
+ "epoch": 27.03,
2308
+ "learning_rate": 3.0180904522613067e-06,
2309
+ "loss": 0.0001,
2310
+ "step": 7000
2311
+ },
2312
+ {
2313
+ "epoch": 27.03,
2314
+ "eval_loss": 0.8605034947395325,
2315
+ "eval_runtime": 1849.0664,
2316
+ "eval_samples_per_second": 4.676,
2317
+ "eval_steps_per_second": 0.468,
2318
+ "eval_wer": 48.53896909262291,
2319
+ "step": 7000
2320
+ },
2321
+ {
2322
+ "epoch": 27.12,
2323
+ "learning_rate": 2.9929648241206034e-06,
2324
+ "loss": 0.0002,
2325
+ "step": 7025
2326
+ },
2327
+ {
2328
+ "epoch": 27.22,
2329
+ "learning_rate": 2.9678391959798996e-06,
2330
+ "loss": 0.0001,
2331
+ "step": 7050
2332
+ },
2333
+ {
2334
+ "epoch": 27.32,
2335
+ "learning_rate": 2.942713567839196e-06,
2336
+ "loss": 0.0001,
2337
+ "step": 7075
2338
+ },
2339
+ {
2340
+ "epoch": 27.41,
2341
+ "learning_rate": 2.917587939698493e-06,
2342
+ "loss": 0.0001,
2343
+ "step": 7100
2344
+ },
2345
+ {
2346
+ "epoch": 27.41,
2347
+ "eval_loss": 0.8613005876541138,
2348
+ "eval_runtime": 1885.8486,
2349
+ "eval_samples_per_second": 4.585,
2350
+ "eval_steps_per_second": 0.459,
2351
+ "eval_wer": 48.92262406773495,
2352
+ "step": 7100
2353
+ },
2354
+ {
2355
+ "epoch": 27.51,
2356
+ "learning_rate": 2.8924623115577894e-06,
2357
+ "loss": 0.0001,
2358
+ "step": 7125
2359
+ },
2360
+ {
2361
+ "epoch": 27.61,
2362
+ "learning_rate": 2.8673366834170856e-06,
2363
+ "loss": 0.0001,
2364
+ "step": 7150
2365
+ },
2366
+ {
2367
+ "epoch": 27.7,
2368
+ "learning_rate": 2.8422110552763822e-06,
2369
+ "loss": 0.0003,
2370
+ "step": 7175
2371
+ },
2372
+ {
2373
+ "epoch": 27.8,
2374
+ "learning_rate": 2.8170854271356784e-06,
2375
+ "loss": 0.0001,
2376
+ "step": 7200
2377
+ },
2378
+ {
2379
+ "epoch": 27.8,
2380
+ "eval_loss": 0.8672710061073303,
2381
+ "eval_runtime": 1858.2034,
2382
+ "eval_samples_per_second": 4.653,
2383
+ "eval_steps_per_second": 0.466,
2384
+ "eval_wer": 48.94908303153578,
2385
+ "step": 7200
2386
+ },
2387
+ {
2388
+ "epoch": 27.9,
2389
+ "learning_rate": 2.7919597989949755e-06,
2390
+ "loss": 0.0001,
2391
+ "step": 7225
2392
+ },
2393
+ {
2394
+ "epoch": 27.99,
2395
+ "learning_rate": 2.7668341708542717e-06,
2396
+ "loss": 0.0001,
2397
+ "step": 7250
2398
+ },
2399
+ {
2400
+ "epoch": 28.09,
2401
+ "learning_rate": 2.7417085427135683e-06,
2402
+ "loss": 0.0001,
2403
+ "step": 7275
2404
+ },
2405
+ {
2406
+ "epoch": 28.19,
2407
+ "learning_rate": 2.7165829145728645e-06,
2408
+ "loss": 0.0001,
2409
+ "step": 7300
2410
+ },
2411
+ {
2412
+ "epoch": 28.19,
2413
+ "eval_loss": 0.8688055872917175,
2414
+ "eval_runtime": 1883.3395,
2415
+ "eval_samples_per_second": 4.591,
2416
+ "eval_steps_per_second": 0.459,
2417
+ "eval_wer": 48.7936366192059,
2418
+ "step": 7300
2419
+ },
2420
+ {
2421
+ "epoch": 28.28,
2422
+ "learning_rate": 2.6914572864321607e-06,
2423
+ "loss": 0.0002,
2424
+ "step": 7325
2425
+ },
2426
+ {
2427
+ "epoch": 28.38,
2428
+ "learning_rate": 2.6663316582914573e-06,
2429
+ "loss": 0.0001,
2430
+ "step": 7350
2431
+ },
2432
+ {
2433
+ "epoch": 28.47,
2434
+ "learning_rate": 2.6412060301507544e-06,
2435
+ "loss": 0.0001,
2436
+ "step": 7375
2437
+ },
2438
+ {
2439
+ "epoch": 28.57,
2440
+ "learning_rate": 2.6160804020100506e-06,
2441
+ "loss": 0.0001,
2442
+ "step": 7400
2443
+ },
2444
+ {
2445
+ "epoch": 28.57,
2446
+ "eval_loss": 0.8710694313049316,
2447
+ "eval_runtime": 1878.6911,
2448
+ "eval_samples_per_second": 4.602,
2449
+ "eval_steps_per_second": 0.46,
2450
+ "eval_wer": 48.8349787501447,
2451
+ "step": 7400
2452
+ },
2453
+ {
2454
+ "epoch": 28.67,
2455
+ "learning_rate": 2.5909547738693468e-06,
2456
+ "loss": 0.0001,
2457
+ "step": 7425
2458
+ },
2459
+ {
2460
+ "epoch": 28.76,
2461
+ "learning_rate": 2.5658291457286434e-06,
2462
+ "loss": 0.0001,
2463
+ "step": 7450
2464
+ },
2465
+ {
2466
+ "epoch": 28.86,
2467
+ "learning_rate": 2.5407035175879396e-06,
2468
+ "loss": 0.0001,
2469
+ "step": 7475
2470
+ },
2471
+ {
2472
+ "epoch": 28.96,
2473
+ "learning_rate": 2.5155778894472366e-06,
2474
+ "loss": 0.0001,
2475
+ "step": 7500
2476
+ },
2477
+ {
2478
+ "epoch": 28.96,
2479
+ "eval_loss": 0.872825026512146,
2480
+ "eval_runtime": 1901.4507,
2481
+ "eval_samples_per_second": 4.547,
2482
+ "eval_steps_per_second": 0.455,
2483
+ "eval_wer": 48.85813034347042,
2484
+ "step": 7500
2485
+ },
2486
+ {
2487
+ "epoch": 29.05,
2488
+ "learning_rate": 2.490452261306533e-06,
2489
+ "loss": 0.0001,
2490
+ "step": 7525
2491
+ },
2492
+ {
2493
+ "epoch": 29.15,
2494
+ "learning_rate": 2.4653266331658294e-06,
2495
+ "loss": 0.0002,
2496
+ "step": 7550
2497
+ },
2498
+ {
2499
+ "epoch": 29.25,
2500
+ "learning_rate": 2.4402010050251256e-06,
2501
+ "loss": 0.0001,
2502
+ "step": 7575
2503
+ },
2504
+ {
2505
+ "epoch": 29.34,
2506
+ "learning_rate": 2.4150753768844223e-06,
2507
+ "loss": 0.0001,
2508
+ "step": 7600
2509
+ },
2510
+ {
2511
+ "epoch": 29.34,
2512
+ "eval_loss": 0.8735654354095459,
2513
+ "eval_runtime": 1920.4668,
2514
+ "eval_samples_per_second": 4.502,
2515
+ "eval_steps_per_second": 0.45,
2516
+ "eval_wer": 49.8453804302889,
2517
+ "step": 7600
2518
+ },
2519
+ {
2520
+ "epoch": 29.44,
2521
+ "learning_rate": 2.389949748743719e-06,
2522
+ "loss": 0.0001,
2523
+ "step": 7625
2524
+ },
2525
+ {
2526
+ "epoch": 29.54,
2527
+ "learning_rate": 2.3648241206030155e-06,
2528
+ "loss": 0.0001,
2529
+ "step": 7650
2530
+ },
2531
+ {
2532
+ "epoch": 29.63,
2533
+ "learning_rate": 2.3396984924623117e-06,
2534
+ "loss": 0.0001,
2535
+ "step": 7675
2536
+ },
2537
+ {
2538
+ "epoch": 29.73,
2539
+ "learning_rate": 2.314572864321608e-06,
2540
+ "loss": 0.0001,
2541
+ "step": 7700
2542
+ },
2543
+ {
2544
+ "epoch": 29.73,
2545
+ "eval_loss": 0.8759255409240723,
2546
+ "eval_runtime": 1934.3216,
2547
+ "eval_samples_per_second": 4.47,
2548
+ "eval_steps_per_second": 0.447,
2549
+ "eval_wer": 49.92475732169139,
2550
+ "step": 7700
2551
+ },
2552
+ {
2553
+ "epoch": 29.83,
2554
+ "learning_rate": 2.289447236180905e-06,
2555
+ "loss": 0.0001,
2556
+ "step": 7725
2557
+ },
2558
+ {
2559
+ "epoch": 29.92,
2560
+ "learning_rate": 2.264321608040201e-06,
2561
+ "loss": 0.0001,
2562
+ "step": 7750
2563
+ },
2564
+ {
2565
+ "epoch": 30.02,
2566
+ "learning_rate": 2.2391959798994978e-06,
2567
+ "loss": 0.0001,
2568
+ "step": 7775
2569
+ },
2570
+ {
2571
+ "epoch": 30.12,
2572
+ "learning_rate": 2.214070351758794e-06,
2573
+ "loss": 0.0001,
2574
+ "step": 7800
2575
+ },
2576
+ {
2577
+ "epoch": 30.12,
2578
+ "eval_loss": 0.8785723447799683,
2579
+ "eval_runtime": 1950.1663,
2580
+ "eval_samples_per_second": 4.433,
2581
+ "eval_steps_per_second": 0.444,
2582
+ "eval_wer": 49.693241388434124,
2583
+ "step": 7800
2584
+ },
2585
+ {
2586
+ "epoch": 30.21,
2587
+ "learning_rate": 2.1889447236180906e-06,
2588
+ "loss": 0.0001,
2589
+ "step": 7825
2590
+ },
2591
+ {
2592
+ "epoch": 30.31,
2593
+ "learning_rate": 2.163819095477387e-06,
2594
+ "loss": 0.0001,
2595
+ "step": 7850
2596
+ },
2597
+ {
2598
+ "epoch": 30.41,
2599
+ "learning_rate": 2.1386934673366834e-06,
2600
+ "loss": 0.0001,
2601
+ "step": 7875
2602
+ },
2603
+ {
2604
+ "epoch": 30.5,
2605
+ "learning_rate": 2.11356783919598e-06,
2606
+ "loss": 0.0001,
2607
+ "step": 7900
2608
+ },
2609
+ {
2610
+ "epoch": 30.5,
2611
+ "eval_loss": 0.8809360265731812,
2612
+ "eval_runtime": 1948.9202,
2613
+ "eval_samples_per_second": 4.436,
2614
+ "eval_steps_per_second": 0.444,
2615
+ "eval_wer": 49.74285194556069,
2616
+ "step": 7900
2617
+ },
2618
+ {
2619
+ "epoch": 30.6,
2620
+ "learning_rate": 2.0884422110552766e-06,
2621
+ "loss": 0.0001,
2622
+ "step": 7925
2623
+ },
2624
+ {
2625
+ "epoch": 30.69,
2626
+ "learning_rate": 2.063316582914573e-06,
2627
+ "loss": 0.0001,
2628
+ "step": 7950
2629
+ },
2630
+ {
2631
+ "epoch": 30.79,
2632
+ "learning_rate": 2.0381909547738695e-06,
2633
+ "loss": 0.0001,
2634
+ "step": 7975
2635
+ },
2636
+ {
2637
+ "epoch": 30.89,
2638
+ "learning_rate": 2.013065326633166e-06,
2639
+ "loss": 0.0001,
2640
+ "step": 8000
2641
+ },
2642
+ {
2643
+ "epoch": 30.89,
2644
+ "eval_loss": 0.8825691342353821,
2645
+ "eval_runtime": 1955.4266,
2646
+ "eval_samples_per_second": 4.422,
2647
+ "eval_steps_per_second": 0.442,
2648
+ "eval_wer": 49.91318152502853,
2649
+ "step": 8000
2650
+ },
2651
+ {
2652
+ "epoch": 30.98,
2653
+ "learning_rate": 1.9879396984924627e-06,
2654
+ "loss": 0.0001,
2655
+ "step": 8025
2656
+ },
2657
+ {
2658
+ "epoch": 31.08,
2659
+ "learning_rate": 1.962814070351759e-06,
2660
+ "loss": 0.0002,
2661
+ "step": 8050
2662
+ },
2663
+ {
2664
+ "epoch": 31.18,
2665
+ "learning_rate": 1.9376884422110555e-06,
2666
+ "loss": 0.0001,
2667
+ "step": 8075
2668
+ },
2669
+ {
2670
+ "epoch": 31.27,
2671
+ "learning_rate": 1.912562814070352e-06,
2672
+ "loss": 0.0001,
2673
+ "step": 8100
2674
+ },
2675
+ {
2676
+ "epoch": 31.27,
2677
+ "eval_loss": 0.8839845657348633,
2678
+ "eval_runtime": 1965.0139,
2679
+ "eval_samples_per_second": 4.4,
2680
+ "eval_steps_per_second": 0.44,
2681
+ "eval_wer": 50.45062922723289,
2682
+ "step": 8100
2683
+ },
2684
+ {
2685
+ "epoch": 31.37,
2686
+ "learning_rate": 1.8874371859296483e-06,
2687
+ "loss": 0.0001,
2688
+ "step": 8125
2689
+ },
2690
+ {
2691
+ "epoch": 31.47,
2692
+ "learning_rate": 1.862311557788945e-06,
2693
+ "loss": 0.0001,
2694
+ "step": 8150
2695
+ },
2696
+ {
2697
+ "epoch": 31.56,
2698
+ "learning_rate": 1.8371859296482414e-06,
2699
+ "loss": 0.0001,
2700
+ "step": 8175
2701
+ },
2702
+ {
2703
+ "epoch": 31.66,
2704
+ "learning_rate": 1.8120603015075378e-06,
2705
+ "loss": 0.0001,
2706
+ "step": 8200
2707
+ },
2708
+ {
2709
+ "epoch": 31.66,
2710
+ "eval_loss": 0.8858728408813477,
2711
+ "eval_runtime": 1964.1317,
2712
+ "eval_samples_per_second": 4.402,
2713
+ "eval_steps_per_second": 0.44,
2714
+ "eval_wer": 50.2918754444279,
2715
+ "step": 8200
2716
+ },
2717
+ {
2718
+ "epoch": 31.76,
2719
+ "learning_rate": 1.7869346733668344e-06,
2720
+ "loss": 0.0001,
2721
+ "step": 8225
2722
+ },
2723
+ {
2724
+ "epoch": 31.85,
2725
+ "learning_rate": 1.7618090452261308e-06,
2726
+ "loss": 0.0001,
2727
+ "step": 8250
2728
+ },
2729
+ {
2730
+ "epoch": 31.95,
2731
+ "learning_rate": 1.7366834170854272e-06,
2732
+ "loss": 0.0001,
2733
+ "step": 8275
2734
+ },
2735
+ {
2736
+ "epoch": 32.05,
2737
+ "learning_rate": 1.7115577889447238e-06,
2738
+ "loss": 0.0001,
2739
+ "step": 8300
2740
+ },
2741
+ {
2742
+ "epoch": 32.05,
2743
+ "eval_loss": 0.8887734413146973,
2744
+ "eval_runtime": 1974.3769,
2745
+ "eval_samples_per_second": 4.379,
2746
+ "eval_steps_per_second": 0.438,
2747
+ "eval_wer": 50.553157711961106,
2748
+ "step": 8300
2749
+ },
2750
+ {
2751
+ "epoch": 32.14,
2752
+ "learning_rate": 1.6864321608040202e-06,
2753
+ "loss": 0.0001,
2754
+ "step": 8325
2755
+ },
2756
+ {
2757
+ "epoch": 32.24,
2758
+ "learning_rate": 1.6613065326633169e-06,
2759
+ "loss": 0.0001,
2760
+ "step": 8350
2761
+ },
2762
+ {
2763
+ "epoch": 32.34,
2764
+ "learning_rate": 1.6361809045226133e-06,
2765
+ "loss": 0.0001,
2766
+ "step": 8375
2767
+ },
2768
+ {
2769
+ "epoch": 32.43,
2770
+ "learning_rate": 1.6110552763819095e-06,
2771
+ "loss": 0.0001,
2772
+ "step": 8400
2773
+ },
2774
+ {
2775
+ "epoch": 32.43,
2776
+ "eval_loss": 0.8924394845962524,
2777
+ "eval_runtime": 1954.9726,
2778
+ "eval_samples_per_second": 4.423,
2779
+ "eval_steps_per_second": 0.442,
2780
+ "eval_wer": 50.64907145573911,
2781
+ "step": 8400
2782
+ },
2783
+ {
2784
+ "epoch": 32.53,
2785
+ "learning_rate": 1.5859296482412063e-06,
2786
+ "loss": 0.0001,
2787
+ "step": 8425
2788
+ },
2789
+ {
2790
+ "epoch": 32.63,
2791
+ "learning_rate": 1.5608040201005025e-06,
2792
+ "loss": 0.0001,
2793
+ "step": 8450
2794
+ },
2795
+ {
2796
+ "epoch": 32.72,
2797
+ "learning_rate": 1.5356783919597993e-06,
2798
+ "loss": 0.0001,
2799
+ "step": 8475
2800
+ },
2801
+ {
2802
+ "epoch": 32.82,
2803
+ "learning_rate": 1.5105527638190955e-06,
2804
+ "loss": 0.0001,
2805
+ "step": 8500
2806
+ },
2807
+ {
2808
+ "epoch": 32.82,
2809
+ "eval_loss": 0.8927690982818604,
2810
+ "eval_runtime": 1969.9296,
2811
+ "eval_samples_per_second": 4.389,
2812
+ "eval_steps_per_second": 0.439,
2813
+ "eval_wer": 50.99965272610011,
2814
+ "step": 8500
2815
+ },
2816
+ {
2817
+ "epoch": 32.92,
2818
+ "learning_rate": 1.485427135678392e-06,
2819
+ "loss": 0.0001,
2820
+ "step": 8525
2821
+ },
2822
+ {
2823
+ "epoch": 33.01,
2824
+ "learning_rate": 1.4603015075376886e-06,
2825
+ "loss": 0.0001,
2826
+ "step": 8550
2827
+ },
2828
+ {
2829
+ "epoch": 33.11,
2830
+ "learning_rate": 1.435175879396985e-06,
2831
+ "loss": 0.0001,
2832
+ "step": 8575
2833
+ },
2834
+ {
2835
+ "epoch": 33.2,
2836
+ "learning_rate": 1.4100502512562816e-06,
2837
+ "loss": 0.0001,
2838
+ "step": 8600
2839
+ },
2840
+ {
2841
+ "epoch": 33.2,
2842
+ "eval_loss": 0.8920454382896423,
2843
+ "eval_runtime": 1993.7139,
2844
+ "eval_samples_per_second": 4.337,
2845
+ "eval_steps_per_second": 0.434,
2846
+ "eval_wer": 51.31881397694763,
2847
+ "step": 8600
2848
+ },
2849
+ {
2850
+ "epoch": 33.3,
2851
+ "learning_rate": 1.384924623115578e-06,
2852
+ "loss": 0.0001,
2853
+ "step": 8625
2854
+ },
2855
+ {
2856
+ "epoch": 33.4,
2857
+ "learning_rate": 1.3597989949748744e-06,
2858
+ "loss": 0.0001,
2859
+ "step": 8650
2860
+ },
2861
+ {
2862
+ "epoch": 33.49,
2863
+ "learning_rate": 1.334673366834171e-06,
2864
+ "loss": 0.0001,
2865
+ "step": 8675
2866
+ },
2867
+ {
2868
+ "epoch": 33.59,
2869
+ "learning_rate": 1.3095477386934674e-06,
2870
+ "loss": 0.0001,
2871
+ "step": 8700
2872
+ },
2873
+ {
2874
+ "epoch": 33.59,
2875
+ "eval_loss": 0.8945680856704712,
2876
+ "eval_runtime": 1991.4509,
2877
+ "eval_samples_per_second": 4.342,
2878
+ "eval_steps_per_second": 0.434,
2879
+ "eval_wer": 51.14186965652958,
2880
+ "step": 8700
2881
+ },
2882
+ {
2883
+ "epoch": 33.69,
2884
+ "learning_rate": 1.284422110552764e-06,
2885
+ "loss": 0.0001,
2886
+ "step": 8725
2887
+ },
2888
+ {
2889
+ "epoch": 33.78,
2890
+ "learning_rate": 1.2592964824120605e-06,
2891
+ "loss": 0.0001,
2892
+ "step": 8750
2893
+ },
2894
+ {
2895
+ "epoch": 33.88,
2896
+ "learning_rate": 1.2341708542713569e-06,
2897
+ "loss": 0.0001,
2898
+ "step": 8775
2899
+ },
2900
+ {
2901
+ "epoch": 33.98,
2902
+ "learning_rate": 1.2090452261306533e-06,
2903
+ "loss": 0.0001,
2904
+ "step": 8800
2905
+ },
2906
+ {
2907
+ "epoch": 33.98,
2908
+ "eval_loss": 0.8956067562103271,
2909
+ "eval_runtime": 1992.4579,
2910
+ "eval_samples_per_second": 4.339,
2911
+ "eval_steps_per_second": 0.434,
2912
+ "eval_wer": 51.155099138429996,
2913
+ "step": 8800
2914
+ },
2915
+ {
2916
+ "epoch": 34.07,
2917
+ "learning_rate": 1.18391959798995e-06,
2918
+ "loss": 0.0001,
2919
+ "step": 8825
2920
+ },
2921
+ {
2922
+ "epoch": 34.17,
2923
+ "learning_rate": 1.1587939698492463e-06,
2924
+ "loss": 0.0001,
2925
+ "step": 8850
2926
+ },
2927
+ {
2928
+ "epoch": 34.27,
2929
+ "learning_rate": 1.133668341708543e-06,
2930
+ "loss": 0.0001,
2931
+ "step": 8875
2932
+ },
2933
+ {
2934
+ "epoch": 34.36,
2935
+ "learning_rate": 1.1085427135678393e-06,
2936
+ "loss": 0.0001,
2937
+ "step": 8900
2938
+ },
2939
+ {
2940
+ "epoch": 34.36,
2941
+ "eval_loss": 0.8987666368484497,
2942
+ "eval_runtime": 1982.6531,
2943
+ "eval_samples_per_second": 4.361,
2944
+ "eval_steps_per_second": 0.436,
2945
+ "eval_wer": 50.74829256999222,
2946
+ "step": 8900
2947
+ },
2948
+ {
2949
+ "epoch": 34.46,
2950
+ "learning_rate": 1.0834170854271357e-06,
2951
+ "loss": 0.0001,
2952
+ "step": 8925
2953
+ },
2954
+ {
2955
+ "epoch": 34.56,
2956
+ "learning_rate": 1.0582914572864322e-06,
2957
+ "loss": 0.0001,
2958
+ "step": 8950
2959
+ },
2960
+ {
2961
+ "epoch": 34.65,
2962
+ "learning_rate": 1.0331658291457288e-06,
2963
+ "loss": 0.0001,
2964
+ "step": 8975
2965
+ },
2966
+ {
2967
+ "epoch": 34.75,
2968
+ "learning_rate": 1.0080402010050252e-06,
2969
+ "loss": 0.0001,
2970
+ "step": 9000
2971
+ },
2972
+ {
2973
+ "epoch": 34.75,
2974
+ "eval_loss": 0.9003340601921082,
2975
+ "eval_runtime": 1982.3624,
2976
+ "eval_samples_per_second": 4.361,
2977
+ "eval_steps_per_second": 0.436,
2978
+ "eval_wer": 50.888855815184144,
2979
+ "step": 9000
2980
+ },
2981
+ {
2982
+ "epoch": 34.85,
2983
+ "learning_rate": 9.829145728643216e-07,
2984
+ "loss": 0.0001,
2985
+ "step": 9025
2986
+ },
2987
+ {
2988
+ "epoch": 34.94,
2989
+ "learning_rate": 9.577889447236182e-07,
2990
+ "loss": 0.0001,
2991
+ "step": 9050
2992
+ },
2993
+ {
2994
+ "epoch": 35.04,
2995
+ "learning_rate": 9.326633165829146e-07,
2996
+ "loss": 0.0001,
2997
+ "step": 9075
2998
+ },
2999
+ {
3000
+ "epoch": 35.14,
3001
+ "learning_rate": 9.075376884422111e-07,
3002
+ "loss": 0.0001,
3003
+ "step": 9100
3004
+ },
3005
+ {
3006
+ "epoch": 35.14,
3007
+ "eval_loss": 0.9010794162750244,
3008
+ "eval_runtime": 1996.3748,
3009
+ "eval_samples_per_second": 4.331,
3010
+ "eval_steps_per_second": 0.433,
3011
+ "eval_wer": 50.93019794612294,
3012
+ "step": 9100
3013
+ },
3014
+ {
3015
+ "epoch": 35.23,
3016
+ "learning_rate": 8.824120603015077e-07,
3017
+ "loss": 0.0001,
3018
+ "step": 9125
3019
+ },
3020
+ {
3021
+ "epoch": 35.33,
3022
+ "learning_rate": 8.572864321608041e-07,
3023
+ "loss": 0.0001,
3024
+ "step": 9150
3025
+ },
3026
+ {
3027
+ "epoch": 35.42,
3028
+ "learning_rate": 8.321608040201006e-07,
3029
+ "loss": 0.0001,
3030
+ "step": 9175
3031
+ },
3032
+ {
3033
+ "epoch": 35.52,
3034
+ "learning_rate": 8.070351758793971e-07,
3035
+ "loss": 0.0001,
3036
+ "step": 9200
3037
+ },
3038
+ {
3039
+ "epoch": 35.52,
3040
+ "eval_loss": 0.9019636511802673,
3041
+ "eval_runtime": 1992.9594,
3042
+ "eval_samples_per_second": 4.338,
3043
+ "eval_steps_per_second": 0.434,
3044
+ "eval_wer": 50.953349539448666,
3045
+ "step": 9200
3046
+ },
3047
+ {
3048
+ "epoch": 35.62,
3049
+ "learning_rate": 7.819095477386936e-07,
3050
+ "loss": 0.0001,
3051
+ "step": 9225
3052
+ },
3053
+ {
3054
+ "epoch": 35.71,
3055
+ "learning_rate": 7.567839195979901e-07,
3056
+ "loss": 0.0001,
3057
+ "step": 9250
3058
+ },
3059
+ {
3060
+ "epoch": 35.81,
3061
+ "learning_rate": 7.316582914572864e-07,
3062
+ "loss": 0.0001,
3063
+ "step": 9275
3064
+ },
3065
+ {
3066
+ "epoch": 35.91,
3067
+ "learning_rate": 7.065326633165829e-07,
3068
+ "loss": 0.0001,
3069
+ "step": 9300
3070
+ },
3071
+ {
3072
+ "epoch": 35.91,
3073
+ "eval_loss": 0.901878297328949,
3074
+ "eval_runtime": 1999.423,
3075
+ "eval_samples_per_second": 4.324,
3076
+ "eval_steps_per_second": 0.433,
3077
+ "eval_wer": 51.560252021630205,
3078
+ "step": 9300
3079
+ },
3080
+ {
3081
+ "epoch": 36.0,
3082
+ "learning_rate": 6.814070351758795e-07,
3083
+ "loss": 0.0001,
3084
+ "step": 9325
3085
+ },
3086
+ {
3087
+ "epoch": 36.1,
3088
+ "learning_rate": 6.56281407035176e-07,
3089
+ "loss": 0.0001,
3090
+ "step": 9350
3091
+ },
3092
+ {
3093
+ "epoch": 36.2,
3094
+ "learning_rate": 6.311557788944724e-07,
3095
+ "loss": 0.0001,
3096
+ "step": 9375
3097
+ },
3098
+ {
3099
+ "epoch": 36.29,
3100
+ "learning_rate": 6.060301507537689e-07,
3101
+ "loss": 0.0001,
3102
+ "step": 9400
3103
+ },
3104
+ {
3105
+ "epoch": 36.29,
3106
+ "eval_loss": 0.9032866358757019,
3107
+ "eval_runtime": 2014.4202,
3108
+ "eval_samples_per_second": 4.292,
3109
+ "eval_steps_per_second": 0.429,
3110
+ "eval_wer": 51.64789733922045,
3111
+ "step": 9400
3112
+ },
3113
+ {
3114
+ "epoch": 36.39,
3115
+ "learning_rate": 5.809045226130654e-07,
3116
+ "loss": 0.0001,
3117
+ "step": 9425
3118
+ },
3119
+ {
3120
+ "epoch": 36.49,
3121
+ "learning_rate": 5.557788944723619e-07,
3122
+ "loss": 0.0001,
3123
+ "step": 9450
3124
+ },
3125
+ {
3126
+ "epoch": 36.58,
3127
+ "learning_rate": 5.306532663316583e-07,
3128
+ "loss": 0.0001,
3129
+ "step": 9475
3130
+ },
3131
+ {
3132
+ "epoch": 36.68,
3133
+ "learning_rate": 5.055276381909549e-07,
3134
+ "loss": 0.0001,
3135
+ "step": 9500
3136
+ },
3137
+ {
3138
+ "epoch": 36.68,
3139
+ "eval_loss": 0.9039894938468933,
3140
+ "eval_runtime": 2029.2995,
3141
+ "eval_samples_per_second": 4.261,
3142
+ "eval_steps_per_second": 0.426,
3143
+ "eval_wer": 51.68097104397149,
3144
+ "step": 9500
3145
+ },
3146
+ {
3147
+ "epoch": 36.78,
3148
+ "learning_rate": 4.804020100502513e-07,
3149
+ "loss": 0.0001,
3150
+ "step": 9525
3151
+ },
3152
+ {
3153
+ "epoch": 36.87,
3154
+ "learning_rate": 4.552763819095478e-07,
3155
+ "loss": 0.0001,
3156
+ "step": 9550
3157
+ },
3158
+ {
3159
+ "epoch": 36.97,
3160
+ "learning_rate": 4.3015075376884424e-07,
3161
+ "loss": 0.0001,
3162
+ "step": 9575
3163
+ },
3164
+ {
3165
+ "epoch": 37.07,
3166
+ "learning_rate": 4.0502512562814075e-07,
3167
+ "loss": 0.0001,
3168
+ "step": 9600
3169
+ },
3170
+ {
3171
+ "epoch": 37.07,
3172
+ "eval_loss": 0.9048639535903931,
3173
+ "eval_runtime": 2059.954,
3174
+ "eval_samples_per_second": 4.197,
3175
+ "eval_steps_per_second": 0.42,
3176
+ "eval_wer": 51.674356303021284,
3177
+ "step": 9600
3178
+ },
3179
+ {
3180
+ "epoch": 37.16,
3181
+ "learning_rate": 3.7989949748743727e-07,
3182
+ "loss": 0.0001,
3183
+ "step": 9625
3184
+ },
3185
+ {
3186
+ "epoch": 37.26,
3187
+ "learning_rate": 3.547738693467337e-07,
3188
+ "loss": 0.0001,
3189
+ "step": 9650
3190
+ },
3191
+ {
3192
+ "epoch": 37.36,
3193
+ "learning_rate": 3.296482412060302e-07,
3194
+ "loss": 0.0001,
3195
+ "step": 9675
3196
+ },
3197
+ {
3198
+ "epoch": 37.45,
3199
+ "learning_rate": 3.0452261306532665e-07,
3200
+ "loss": 0.0001,
3201
+ "step": 9700
3202
+ },
3203
+ {
3204
+ "epoch": 37.45,
3205
+ "eval_loss": 0.905308187007904,
3206
+ "eval_runtime": 1999.2162,
3207
+ "eval_samples_per_second": 4.325,
3208
+ "eval_steps_per_second": 0.433,
3209
+ "eval_wer": 51.6429362835078,
3210
+ "step": 9700
3211
+ },
3212
+ {
3213
+ "epoch": 37.55,
3214
+ "learning_rate": 2.7939698492462317e-07,
3215
+ "loss": 0.0001,
3216
+ "step": 9725
3217
+ },
3218
+ {
3219
+ "epoch": 37.64,
3220
+ "learning_rate": 2.5427135678391963e-07,
3221
+ "loss": 0.0001,
3222
+ "step": 9750
3223
+ },
3224
+ {
3225
+ "epoch": 37.74,
3226
+ "learning_rate": 2.291457286432161e-07,
3227
+ "loss": 0.0001,
3228
+ "step": 9775
3229
+ },
3230
+ {
3231
+ "epoch": 37.84,
3232
+ "learning_rate": 2.0402010050251258e-07,
3233
+ "loss": 0.0001,
3234
+ "step": 9800
3235
+ },
3236
+ {
3237
+ "epoch": 37.84,
3238
+ "eval_loss": 0.9057276248931885,
3239
+ "eval_runtime": 2081.4562,
3240
+ "eval_samples_per_second": 4.154,
3241
+ "eval_steps_per_second": 0.416,
3242
+ "eval_wer": 51.59663309685635,
3243
+ "step": 9800
3244
+ },
3245
+ {
3246
+ "epoch": 37.93,
3247
+ "learning_rate": 1.7889447236180904e-07,
3248
+ "loss": 0.0,
3249
+ "step": 9825
3250
+ },
3251
+ {
3252
+ "epoch": 38.03,
3253
+ "learning_rate": 1.5376884422110556e-07,
3254
+ "loss": 0.0001,
3255
+ "step": 9850
3256
+ },
3257
+ {
3258
+ "epoch": 38.13,
3259
+ "learning_rate": 1.2864321608040202e-07,
3260
+ "loss": 0.0001,
3261
+ "step": 9875
3262
+ },
3263
+ {
3264
+ "epoch": 38.22,
3265
+ "learning_rate": 1.035175879396985e-07,
3266
+ "loss": 0.0001,
3267
+ "step": 9900
3268
+ },
3269
+ {
3270
+ "epoch": 38.22,
3271
+ "eval_loss": 0.9060160517692566,
3272
+ "eval_runtime": 2006.0349,
3273
+ "eval_samples_per_second": 4.31,
3274
+ "eval_steps_per_second": 0.431,
3275
+ "eval_wer": 51.73719634204825,
3276
+ "step": 9900
3277
+ },
3278
+ {
3279
+ "epoch": 38.32,
3280
+ "learning_rate": 7.839195979899497e-08,
3281
+ "loss": 0.0001,
3282
+ "step": 9925
3283
+ },
3284
+ {
3285
+ "epoch": 38.42,
3286
+ "learning_rate": 5.326633165829146e-08,
3287
+ "loss": 0.0001,
3288
+ "step": 9950
3289
+ },
3290
+ {
3291
+ "epoch": 38.51,
3292
+ "learning_rate": 2.8140703517587944e-08,
3293
+ "loss": 0.0001,
3294
+ "step": 9975
3295
+ },
3296
+ {
3297
+ "epoch": 38.61,
3298
+ "learning_rate": 3.0150753768844226e-09,
3299
+ "loss": 0.0001,
3300
+ "step": 10000
3301
+ },
3302
+ {
3303
+ "epoch": 38.61,
3304
+ "eval_loss": 0.9061161279678345,
3305
+ "eval_runtime": 2160.7023,
3306
+ "eval_samples_per_second": 4.001,
3307
+ "eval_steps_per_second": 0.4,
3308
+ "eval_wer": 51.37007821931173,
3309
+ "step": 10000
3310
+ }
3311
+ ],
3312
+ "logging_steps": 25,
3313
+ "max_steps": 10000,
3314
+ "num_train_epochs": 39,
3315
+ "save_steps": 100,
3316
+ "total_flos": 1.615529929531392e+20,
3317
+ "trial_name": null,
3318
+ "trial_params": null
3319
+ }
1/checkpoint-10000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d286c349fc7a4c7419632d8b055a552e2835b4138f3ff268d56552451405f53b
3
+ size 4347
1/checkpoint-1100/config.json ADDED
@@ -0,0 +1,52 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "openai/whisper-small",
3
+ "activation_dropout": 0.0,
4
+ "activation_function": "gelu",
5
+ "apply_spec_augment": false,
6
+ "architectures": [
7
+ "WhisperForConditionalGeneration"
8
+ ],
9
+ "attention_dropout": 0.0,
10
+ "begin_suppress_tokens": [
11
+ 220,
12
+ 50257
13
+ ],
14
+ "bos_token_id": 50257,
15
+ "classifier_proj_size": 256,
16
+ "d_model": 768,
17
+ "decoder_attention_heads": 12,
18
+ "decoder_ffn_dim": 3072,
19
+ "decoder_layerdrop": 0.0,
20
+ "decoder_layers": 12,
21
+ "decoder_start_token_id": 50258,
22
+ "dropout": 0.0,
23
+ "encoder_attention_heads": 12,
24
+ "encoder_ffn_dim": 3072,
25
+ "encoder_layerdrop": 0.0,
26
+ "encoder_layers": 12,
27
+ "eos_token_id": 50257,
28
+ "forced_decoder_ids": null,
29
+ "init_std": 0.02,
30
+ "is_encoder_decoder": true,
31
+ "mask_feature_length": 10,
32
+ "mask_feature_min_masks": 0,
33
+ "mask_feature_prob": 0.0,
34
+ "mask_time_length": 10,
35
+ "mask_time_min_masks": 2,
36
+ "mask_time_prob": 0.05,
37
+ "max_length": 448,
38
+ "max_source_positions": 1500,
39
+ "max_target_positions": 448,
40
+ "median_filter_width": 7,
41
+ "model_type": "whisper",
42
+ "num_hidden_layers": 12,
43
+ "num_mel_bins": 80,
44
+ "pad_token_id": 50257,
45
+ "scale_embedding": false,
46
+ "suppress_tokens": [],
47
+ "torch_dtype": "float32",
48
+ "transformers_version": "4.35.0",
49
+ "use_cache": true,
50
+ "use_weighted_layer_sum": false,
51
+ "vocab_size": 51865
52
+ }
1/checkpoint-1100/generation_config.json ADDED
@@ -0,0 +1,263 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "alignment_heads": [
3
+ [
4
+ 5,
5
+ 3
6
+ ],
7
+ [
8
+ 5,
9
+ 9
10
+ ],
11
+ [
12
+ 8,
13
+ 0
14
+ ],
15
+ [
16
+ 8,
17
+ 4
18
+ ],
19
+ [
20
+ 8,
21
+ 7
22
+ ],
23
+ [
24
+ 8,
25
+ 8
26
+ ],
27
+ [
28
+ 9,
29
+ 0
30
+ ],
31
+ [
32
+ 9,
33
+ 7
34
+ ],
35
+ [
36
+ 9,
37
+ 9
38
+ ],
39
+ [
40
+ 10,
41
+ 5
42
+ ]
43
+ ],
44
+ "begin_suppress_tokens": [
45
+ 220,
46
+ 50257
47
+ ],
48
+ "bos_token_id": 50257,
49
+ "decoder_start_token_id": 50258,
50
+ "eos_token_id": 50257,
51
+ "forced_decoder_ids": [
52
+ [
53
+ 1,
54
+ null
55
+ ],
56
+ [
57
+ 2,
58
+ 50359
59
+ ]
60
+ ],
61
+ "is_multilingual": true,
62
+ "lang_to_id": {
63
+ "<|af|>": 50327,
64
+ "<|am|>": 50334,
65
+ "<|ar|>": 50272,
66
+ "<|as|>": 50350,
67
+ "<|az|>": 50304,
68
+ "<|ba|>": 50355,
69
+ "<|be|>": 50330,
70
+ "<|bg|>": 50292,
71
+ "<|bn|>": 50302,
72
+ "<|bo|>": 50347,
73
+ "<|br|>": 50309,
74
+ "<|bs|>": 50315,
75
+ "<|ca|>": 50270,
76
+ "<|cs|>": 50283,
77
+ "<|cy|>": 50297,
78
+ "<|da|>": 50285,
79
+ "<|de|>": 50261,
80
+ "<|el|>": 50281,
81
+ "<|en|>": 50259,
82
+ "<|es|>": 50262,
83
+ "<|et|>": 50307,
84
+ "<|eu|>": 50310,
85
+ "<|fa|>": 50300,
86
+ "<|fi|>": 50277,
87
+ "<|fo|>": 50338,
88
+ "<|fr|>": 50265,
89
+ "<|gl|>": 50319,
90
+ "<|gu|>": 50333,
91
+ "<|haw|>": 50352,
92
+ "<|ha|>": 50354,
93
+ "<|he|>": 50279,
94
+ "<|hi|>": 50276,
95
+ "<|hr|>": 50291,
96
+ "<|ht|>": 50339,
97
+ "<|hu|>": 50286,
98
+ "<|hy|>": 50312,
99
+ "<|id|>": 50275,
100
+ "<|is|>": 50311,
101
+ "<|it|>": 50274,
102
+ "<|ja|>": 50266,
103
+ "<|jw|>": 50356,
104
+ "<|ka|>": 50329,
105
+ "<|kk|>": 50316,
106
+ "<|km|>": 50323,
107
+ "<|kn|>": 50306,
108
+ "<|ko|>": 50264,
109
+ "<|la|>": 50294,
110
+ "<|lb|>": 50345,
111
+ "<|ln|>": 50353,
112
+ "<|lo|>": 50336,
113
+ "<|lt|>": 50293,
114
+ "<|lv|>": 50301,
115
+ "<|mg|>": 50349,
116
+ "<|mi|>": 50295,
117
+ "<|mk|>": 50308,
118
+ "<|ml|>": 50296,
119
+ "<|mn|>": 50314,
120
+ "<|mr|>": 50320,
121
+ "<|ms|>": 50282,
122
+ "<|mt|>": 50343,
123
+ "<|my|>": 50346,
124
+ "<|ne|>": 50313,
125
+ "<|nl|>": 50271,
126
+ "<|nn|>": 50342,
127
+ "<|no|>": 50288,
128
+ "<|oc|>": 50328,
129
+ "<|pa|>": 50321,
130
+ "<|pl|>": 50269,
131
+ "<|ps|>": 50340,
132
+ "<|pt|>": 50267,
133
+ "<|ro|>": 50284,
134
+ "<|ru|>": 50263,
135
+ "<|sa|>": 50344,
136
+ "<|sd|>": 50332,
137
+ "<|si|>": 50322,
138
+ "<|sk|>": 50298,
139
+ "<|sl|>": 50305,
140
+ "<|sn|>": 50324,
141
+ "<|so|>": 50326,
142
+ "<|sq|>": 50317,
143
+ "<|sr|>": 50303,
144
+ "<|su|>": 50357,
145
+ "<|sv|>": 50273,
146
+ "<|sw|>": 50318,
147
+ "<|ta|>": 50287,
148
+ "<|te|>": 50299,
149
+ "<|tg|>": 50331,
150
+ "<|th|>": 50289,
151
+ "<|tk|>": 50341,
152
+ "<|tl|>": 50348,
153
+ "<|tr|>": 50268,
154
+ "<|tt|>": 50351,
155
+ "<|uk|>": 50280,
156
+ "<|ur|>": 50290,
157
+ "<|uz|>": 50337,
158
+ "<|vi|>": 50278,
159
+ "<|yi|>": 50335,
160
+ "<|yo|>": 50325,
161
+ "<|zh|>": 50260
162
+ },
163
+ "max_initial_timestamp_index": 1,
164
+ "max_length": 448,
165
+ "no_timestamps_token_id": 50363,
166
+ "pad_token_id": 50257,
167
+ "return_timestamps": false,
168
+ "suppress_tokens": [
169
+ 1,
170
+ 2,
171
+ 7,
172
+ 8,
173
+ 9,
174
+ 10,
175
+ 14,
176
+ 25,
177
+ 26,
178
+ 27,
179
+ 28,
180
+ 29,
181
+ 31,
182
+ 58,
183
+ 59,
184
+ 60,
185
+ 61,
186
+ 62,
187
+ 63,
188
+ 90,
189
+ 91,
190
+ 92,
191
+ 93,
192
+ 359,
193
+ 503,
194
+ 522,
195
+ 542,
196
+ 873,
197
+ 893,
198
+ 902,
199
+ 918,
200
+ 922,
201
+ 931,
202
+ 1350,
203
+ 1853,
204
+ 1982,
205
+ 2460,
206
+ 2627,
207
+ 3246,
208
+ 3253,
209
+ 3268,
210
+ 3536,
211
+ 3846,
212
+ 3961,
213
+ 4183,
214
+ 4667,
215
+ 6585,
216
+ 6647,
217
+ 7273,
218
+ 9061,
219
+ 9383,
220
+ 10428,
221
+ 10929,
222
+ 11938,
223
+ 12033,
224
+ 12331,
225
+ 12562,
226
+ 13793,
227
+ 14157,
228
+ 14635,
229
+ 15265,
230
+ 15618,
231
+ 16553,
232
+ 16604,
233
+ 18362,
234
+ 18956,
235
+ 20075,
236
+ 21675,
237
+ 22520,
238
+ 26130,
239
+ 26161,
240
+ 26435,
241
+ 28279,
242
+ 29464,
243
+ 31650,
244
+ 32302,
245
+ 32470,
246
+ 36865,
247
+ 42863,
248
+ 47425,
249
+ 49870,
250
+ 50254,
251
+ 50258,
252
+ 50358,
253
+ 50359,
254
+ 50360,
255
+ 50361,
256
+ 50362
257
+ ],
258
+ "task_to_id": {
259
+ "transcribe": 50359,
260
+ "translate": 50358
261
+ },
262
+ "transformers_version": "4.35.0"
263
+ }
1/checkpoint-1100/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2f8de38a65e8b63dd10f7cc2ce1ce47e367c615071f6f5393f8b28e2ddd2f2b4
3
+ size 966995080
1/checkpoint-1100/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f0473580520da16304229a3ad7395bfb95a8a2f0cbfefb8418a1f0f38841d2d4
3
+ size 1925063607
1/checkpoint-1100/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
+ }
1/checkpoint-1100/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f799a7692bc62cf4d69fb5867bcee7d7ad3f469cd7775707ea6d287e27a5db40
3
+ size 14575
1/checkpoint-1100/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1e6612c399471f3114e2eb6ea051ef81c32a3e11cba700ec3591833029061e0b
3
+ size 627
1/checkpoint-1100/trainer_state.json ADDED
@@ -0,0 +1,382 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": 46.084900200095916,
3
+ "best_model_checkpoint": "/media/makhataei/Backups/Whisper-Small-Common-Voice/checkpoint-1100",
4
+ "epoch": 4.2471042471042475,
5
+ "eval_steps": 100,
6
+ "global_step": 1100,
7
+ "is_hyper_param_search": false,
8
+ "is_local_process_zero": true,
9
+ "is_world_process_zero": true,
10
+ "log_history": [
11
+ {
12
+ "epoch": 0.1,
13
+ "learning_rate": 4.600000000000001e-06,
14
+ "loss": 3.1025,
15
+ "step": 25
16
+ },
17
+ {
18
+ "epoch": 0.19,
19
+ "learning_rate": 9.600000000000001e-06,
20
+ "loss": 1.3525,
21
+ "step": 50
22
+ },
23
+ {
24
+ "epoch": 0.29,
25
+ "learning_rate": 9.976884422110554e-06,
26
+ "loss": 0.9564,
27
+ "step": 75
28
+ },
29
+ {
30
+ "epoch": 0.39,
31
+ "learning_rate": 9.95175879396985e-06,
32
+ "loss": 0.7495,
33
+ "step": 100
34
+ },
35
+ {
36
+ "epoch": 0.39,
37
+ "eval_loss": 0.860159158706665,
38
+ "eval_runtime": 1426.2695,
39
+ "eval_samples_per_second": 6.062,
40
+ "eval_steps_per_second": 0.606,
41
+ "eval_wer": 58.07742554282218,
42
+ "step": 100
43
+ },
44
+ {
45
+ "epoch": 0.48,
46
+ "learning_rate": 9.926633165829147e-06,
47
+ "loss": 0.5073,
48
+ "step": 125
49
+ },
50
+ {
51
+ "epoch": 0.58,
52
+ "learning_rate": 9.901507537688444e-06,
53
+ "loss": 0.3749,
54
+ "step": 150
55
+ },
56
+ {
57
+ "epoch": 0.68,
58
+ "learning_rate": 9.87638190954774e-06,
59
+ "loss": 0.3247,
60
+ "step": 175
61
+ },
62
+ {
63
+ "epoch": 0.77,
64
+ "learning_rate": 9.851256281407035e-06,
65
+ "loss": 0.3059,
66
+ "step": 200
67
+ },
68
+ {
69
+ "epoch": 0.77,
70
+ "eval_loss": 0.5538379549980164,
71
+ "eval_runtime": 1589.2333,
72
+ "eval_samples_per_second": 5.44,
73
+ "eval_steps_per_second": 0.544,
74
+ "eval_wer": 55.22481850804518,
75
+ "step": 200
76
+ },
77
+ {
78
+ "epoch": 0.87,
79
+ "learning_rate": 9.826130653266333e-06,
80
+ "loss": 0.2963,
81
+ "step": 225
82
+ },
83
+ {
84
+ "epoch": 0.97,
85
+ "learning_rate": 9.80100502512563e-06,
86
+ "loss": 0.2882,
87
+ "step": 250
88
+ },
89
+ {
90
+ "epoch": 1.06,
91
+ "learning_rate": 9.775879396984925e-06,
92
+ "loss": 0.2327,
93
+ "step": 275
94
+ },
95
+ {
96
+ "epoch": 1.16,
97
+ "learning_rate": 9.750753768844222e-06,
98
+ "loss": 0.2002,
99
+ "step": 300
100
+ },
101
+ {
102
+ "epoch": 1.16,
103
+ "eval_loss": 0.5136818885803223,
104
+ "eval_runtime": 1646.1357,
105
+ "eval_samples_per_second": 5.252,
106
+ "eval_steps_per_second": 0.525,
107
+ "eval_wer": 52.46151047609599,
108
+ "step": 300
109
+ },
110
+ {
111
+ "epoch": 1.25,
112
+ "learning_rate": 9.725628140703518e-06,
113
+ "loss": 0.1893,
114
+ "step": 325
115
+ },
116
+ {
117
+ "epoch": 1.35,
118
+ "learning_rate": 9.700502512562815e-06,
119
+ "loss": 0.185,
120
+ "step": 350
121
+ },
122
+ {
123
+ "epoch": 1.45,
124
+ "learning_rate": 9.675376884422111e-06,
125
+ "loss": 0.1847,
126
+ "step": 375
127
+ },
128
+ {
129
+ "epoch": 1.54,
130
+ "learning_rate": 9.650251256281408e-06,
131
+ "loss": 0.1798,
132
+ "step": 400
133
+ },
134
+ {
135
+ "epoch": 1.54,
136
+ "eval_loss": 0.5102902054786682,
137
+ "eval_runtime": 1650.3734,
138
+ "eval_samples_per_second": 5.239,
139
+ "eval_steps_per_second": 0.524,
140
+ "eval_wer": 51.65781945064576,
141
+ "step": 400
142
+ },
143
+ {
144
+ "epoch": 1.64,
145
+ "learning_rate": 9.625125628140705e-06,
146
+ "loss": 0.1804,
147
+ "step": 425
148
+ },
149
+ {
150
+ "epoch": 1.74,
151
+ "learning_rate": 9.600000000000001e-06,
152
+ "loss": 0.1807,
153
+ "step": 450
154
+ },
155
+ {
156
+ "epoch": 1.83,
157
+ "learning_rate": 9.574874371859298e-06,
158
+ "loss": 0.1714,
159
+ "step": 475
160
+ },
161
+ {
162
+ "epoch": 1.93,
163
+ "learning_rate": 9.549748743718595e-06,
164
+ "loss": 0.1678,
165
+ "step": 500
166
+ },
167
+ {
168
+ "epoch": 1.93,
169
+ "eval_loss": 0.48063144087791443,
170
+ "eval_runtime": 1482.5018,
171
+ "eval_samples_per_second": 5.832,
172
+ "eval_steps_per_second": 0.583,
173
+ "eval_wer": 46.55123943708554,
174
+ "step": 500
175
+ },
176
+ {
177
+ "epoch": 2.03,
178
+ "learning_rate": 9.52462311557789e-06,
179
+ "loss": 0.1542,
180
+ "step": 525
181
+ },
182
+ {
183
+ "epoch": 2.12,
184
+ "learning_rate": 9.499497487437188e-06,
185
+ "loss": 0.0929,
186
+ "step": 550
187
+ },
188
+ {
189
+ "epoch": 2.22,
190
+ "learning_rate": 9.474371859296483e-06,
191
+ "loss": 0.0936,
192
+ "step": 575
193
+ },
194
+ {
195
+ "epoch": 2.32,
196
+ "learning_rate": 9.44924623115578e-06,
197
+ "loss": 0.0932,
198
+ "step": 600
199
+ },
200
+ {
201
+ "epoch": 2.32,
202
+ "eval_loss": 0.49363330006599426,
203
+ "eval_runtime": 1993.4903,
204
+ "eval_samples_per_second": 4.337,
205
+ "eval_steps_per_second": 0.434,
206
+ "eval_wer": 47.49549370772767,
207
+ "step": 600
208
+ },
209
+ {
210
+ "epoch": 2.41,
211
+ "learning_rate": 9.424120603015076e-06,
212
+ "loss": 0.0936,
213
+ "step": 625
214
+ },
215
+ {
216
+ "epoch": 2.51,
217
+ "learning_rate": 9.398994974874373e-06,
218
+ "loss": 0.0955,
219
+ "step": 650
220
+ },
221
+ {
222
+ "epoch": 2.61,
223
+ "learning_rate": 9.373869346733669e-06,
224
+ "loss": 0.0871,
225
+ "step": 675
226
+ },
227
+ {
228
+ "epoch": 2.7,
229
+ "learning_rate": 9.348743718592966e-06,
230
+ "loss": 0.0944,
231
+ "step": 700
232
+ },
233
+ {
234
+ "epoch": 2.7,
235
+ "eval_loss": 0.5119881629943848,
236
+ "eval_runtime": 1791.3192,
237
+ "eval_samples_per_second": 4.827,
238
+ "eval_steps_per_second": 0.483,
239
+ "eval_wer": 48.91600932678474,
240
+ "step": 700
241
+ },
242
+ {
243
+ "epoch": 2.8,
244
+ "learning_rate": 9.323618090452262e-06,
245
+ "loss": 0.0927,
246
+ "step": 725
247
+ },
248
+ {
249
+ "epoch": 2.9,
250
+ "learning_rate": 9.298492462311557e-06,
251
+ "loss": 0.0964,
252
+ "step": 750
253
+ },
254
+ {
255
+ "epoch": 2.99,
256
+ "learning_rate": 9.273366834170856e-06,
257
+ "loss": 0.1038,
258
+ "step": 775
259
+ },
260
+ {
261
+ "epoch": 3.09,
262
+ "learning_rate": 9.248241206030152e-06,
263
+ "loss": 0.0493,
264
+ "step": 800
265
+ },
266
+ {
267
+ "epoch": 3.09,
268
+ "eval_loss": 0.5198288559913635,
269
+ "eval_runtime": 1878.4077,
270
+ "eval_samples_per_second": 4.603,
271
+ "eval_steps_per_second": 0.46,
272
+ "eval_wer": 48.26114997271419,
273
+ "step": 800
274
+ },
275
+ {
276
+ "epoch": 3.19,
277
+ "learning_rate": 9.223115577889447e-06,
278
+ "loss": 0.0501,
279
+ "step": 825
280
+ },
281
+ {
282
+ "epoch": 3.28,
283
+ "learning_rate": 9.197989949748744e-06,
284
+ "loss": 0.0458,
285
+ "step": 850
286
+ },
287
+ {
288
+ "epoch": 3.38,
289
+ "learning_rate": 9.17286432160804e-06,
290
+ "loss": 0.0492,
291
+ "step": 875
292
+ },
293
+ {
294
+ "epoch": 3.47,
295
+ "learning_rate": 9.147738693467337e-06,
296
+ "loss": 0.0506,
297
+ "step": 900
298
+ },
299
+ {
300
+ "epoch": 3.47,
301
+ "eval_loss": 0.5227376222610474,
302
+ "eval_runtime": 1759.968,
303
+ "eval_samples_per_second": 4.913,
304
+ "eval_steps_per_second": 0.491,
305
+ "eval_wer": 47.41280944585008,
306
+ "step": 900
307
+ },
308
+ {
309
+ "epoch": 3.57,
310
+ "learning_rate": 9.122613065326634e-06,
311
+ "loss": 0.0494,
312
+ "step": 925
313
+ },
314
+ {
315
+ "epoch": 3.67,
316
+ "learning_rate": 9.09748743718593e-06,
317
+ "loss": 0.051,
318
+ "step": 950
319
+ },
320
+ {
321
+ "epoch": 3.76,
322
+ "learning_rate": 9.072361809045227e-06,
323
+ "loss": 0.0466,
324
+ "step": 975
325
+ },
326
+ {
327
+ "epoch": 3.86,
328
+ "learning_rate": 9.047236180904524e-06,
329
+ "loss": 0.0482,
330
+ "step": 1000
331
+ },
332
+ {
333
+ "epoch": 3.86,
334
+ "eval_loss": 0.5255833268165588,
335
+ "eval_runtime": 1701.3645,
336
+ "eval_samples_per_second": 5.082,
337
+ "eval_steps_per_second": 0.508,
338
+ "eval_wer": 46.94977757933555,
339
+ "step": 1000
340
+ },
341
+ {
342
+ "epoch": 3.96,
343
+ "learning_rate": 9.02211055276382e-06,
344
+ "loss": 0.046,
345
+ "step": 1025
346
+ },
347
+ {
348
+ "epoch": 4.05,
349
+ "learning_rate": 8.996984924623117e-06,
350
+ "loss": 0.0368,
351
+ "step": 1050
352
+ },
353
+ {
354
+ "epoch": 4.15,
355
+ "learning_rate": 8.971859296482412e-06,
356
+ "loss": 0.0233,
357
+ "step": 1075
358
+ },
359
+ {
360
+ "epoch": 4.25,
361
+ "learning_rate": 8.94673366834171e-06,
362
+ "loss": 0.0244,
363
+ "step": 1100
364
+ },
365
+ {
366
+ "epoch": 4.25,
367
+ "eval_loss": 0.5491495132446289,
368
+ "eval_runtime": 1619.0836,
369
+ "eval_samples_per_second": 5.34,
370
+ "eval_steps_per_second": 0.534,
371
+ "eval_wer": 46.084900200095916,
372
+ "step": 1100
373
+ }
374
+ ],
375
+ "logging_steps": 25,
376
+ "max_steps": 10000,
377
+ "num_train_epochs": 39,
378
+ "save_steps": 100,
379
+ "total_flos": 1.77710889517056e+19,
380
+ "trial_name": null,
381
+ "trial_params": null
382
+ }
1/checkpoint-1100/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d286c349fc7a4c7419632d8b055a552e2835b4138f3ff268d56552451405f53b
3
+ size 4347
1/checkpoint-1200/config.json ADDED
@@ -0,0 +1,52 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "openai/whisper-small",
3
+ "activation_dropout": 0.0,
4
+ "activation_function": "gelu",
5
+ "apply_spec_augment": false,
6
+ "architectures": [
7
+ "WhisperForConditionalGeneration"
8
+ ],
9
+ "attention_dropout": 0.0,
10
+ "begin_suppress_tokens": [
11
+ 220,
12
+ 50257
13
+ ],
14
+ "bos_token_id": 50257,
15
+ "classifier_proj_size": 256,
16
+ "d_model": 768,
17
+ "decoder_attention_heads": 12,
18
+ "decoder_ffn_dim": 3072,
19
+ "decoder_layerdrop": 0.0,
20
+ "decoder_layers": 12,
21
+ "decoder_start_token_id": 50258,
22
+ "dropout": 0.0,
23
+ "encoder_attention_heads": 12,
24
+ "encoder_ffn_dim": 3072,
25
+ "encoder_layerdrop": 0.0,
26
+ "encoder_layers": 12,
27
+ "eos_token_id": 50257,
28
+ "forced_decoder_ids": null,
29
+ "init_std": 0.02,
30
+ "is_encoder_decoder": true,
31
+ "mask_feature_length": 10,
32
+ "mask_feature_min_masks": 0,
33
+ "mask_feature_prob": 0.0,
34
+ "mask_time_length": 10,
35
+ "mask_time_min_masks": 2,
36
+ "mask_time_prob": 0.05,
37
+ "max_length": 448,
38
+ "max_source_positions": 1500,
39
+ "max_target_positions": 448,
40
+ "median_filter_width": 7,
41
+ "model_type": "whisper",
42
+ "num_hidden_layers": 12,
43
+ "num_mel_bins": 80,
44
+ "pad_token_id": 50257,
45
+ "scale_embedding": false,
46
+ "suppress_tokens": [],
47
+ "torch_dtype": "float32",
48
+ "transformers_version": "4.35.0",
49
+ "use_cache": true,
50
+ "use_weighted_layer_sum": false,
51
+ "vocab_size": 51865
52
+ }
1/checkpoint-1200/generation_config.json ADDED
@@ -0,0 +1,263 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "alignment_heads": [
3
+ [
4
+ 5,
5
+ 3
6
+ ],
7
+ [
8
+ 5,
9
+ 9
10
+ ],
11
+ [
12
+ 8,
13
+ 0
14
+ ],
15
+ [
16
+ 8,
17
+ 4
18
+ ],
19
+ [
20
+ 8,
21
+ 7
22
+ ],
23
+ [
24
+ 8,
25
+ 8
26
+ ],
27
+ [
28
+ 9,
29
+ 0
30
+ ],
31
+ [
32
+ 9,
33
+ 7
34
+ ],
35
+ [
36
+ 9,
37
+ 9
38
+ ],
39
+ [
40
+ 10,
41
+ 5
42
+ ]
43
+ ],
44
+ "begin_suppress_tokens": [
45
+ 220,
46
+ 50257
47
+ ],
48
+ "bos_token_id": 50257,
49
+ "decoder_start_token_id": 50258,
50
+ "eos_token_id": 50257,
51
+ "forced_decoder_ids": [
52
+ [
53
+ 1,
54
+ null
55
+ ],
56
+ [
57
+ 2,
58
+ 50359
59
+ ]
60
+ ],
61
+ "is_multilingual": true,
62
+ "lang_to_id": {
63
+ "<|af|>": 50327,
64
+ "<|am|>": 50334,
65
+ "<|ar|>": 50272,
66
+ "<|as|>": 50350,
67
+ "<|az|>": 50304,
68
+ "<|ba|>": 50355,
69
+ "<|be|>": 50330,
70
+ "<|bg|>": 50292,
71
+ "<|bn|>": 50302,
72
+ "<|bo|>": 50347,
73
+ "<|br|>": 50309,
74
+ "<|bs|>": 50315,
75
+ "<|ca|>": 50270,
76
+ "<|cs|>": 50283,
77
+ "<|cy|>": 50297,
78
+ "<|da|>": 50285,
79
+ "<|de|>": 50261,
80
+ "<|el|>": 50281,
81
+ "<|en|>": 50259,
82
+ "<|es|>": 50262,
83
+ "<|et|>": 50307,
84
+ "<|eu|>": 50310,
85
+ "<|fa|>": 50300,
86
+ "<|fi|>": 50277,
87
+ "<|fo|>": 50338,
88
+ "<|fr|>": 50265,
89
+ "<|gl|>": 50319,
90
+ "<|gu|>": 50333,
91
+ "<|haw|>": 50352,
92
+ "<|ha|>": 50354,
93
+ "<|he|>": 50279,
94
+ "<|hi|>": 50276,
95
+ "<|hr|>": 50291,
96
+ "<|ht|>": 50339,
97
+ "<|hu|>": 50286,
98
+ "<|hy|>": 50312,
99
+ "<|id|>": 50275,
100
+ "<|is|>": 50311,
101
+ "<|it|>": 50274,
102
+ "<|ja|>": 50266,
103
+ "<|jw|>": 50356,
104
+ "<|ka|>": 50329,
105
+ "<|kk|>": 50316,
106
+ "<|km|>": 50323,
107
+ "<|kn|>": 50306,
108
+ "<|ko|>": 50264,
109
+ "<|la|>": 50294,
110
+ "<|lb|>": 50345,
111
+ "<|ln|>": 50353,
112
+ "<|lo|>": 50336,
113
+ "<|lt|>": 50293,
114
+ "<|lv|>": 50301,
115
+ "<|mg|>": 50349,
116
+ "<|mi|>": 50295,
117
+ "<|mk|>": 50308,
118
+ "<|ml|>": 50296,
119
+ "<|mn|>": 50314,
120
+ "<|mr|>": 50320,
121
+ "<|ms|>": 50282,
122
+ "<|mt|>": 50343,
123
+ "<|my|>": 50346,
124
+ "<|ne|>": 50313,
125
+ "<|nl|>": 50271,
126
+ "<|nn|>": 50342,
127
+ "<|no|>": 50288,
128
+ "<|oc|>": 50328,
129
+ "<|pa|>": 50321,
130
+ "<|pl|>": 50269,
131
+ "<|ps|>": 50340,
132
+ "<|pt|>": 50267,
133
+ "<|ro|>": 50284,
134
+ "<|ru|>": 50263,
135
+ "<|sa|>": 50344,
136
+ "<|sd|>": 50332,
137
+ "<|si|>": 50322,
138
+ "<|sk|>": 50298,
139
+ "<|sl|>": 50305,
140
+ "<|sn|>": 50324,
141
+ "<|so|>": 50326,
142
+ "<|sq|>": 50317,
143
+ "<|sr|>": 50303,
144
+ "<|su|>": 50357,
145
+ "<|sv|>": 50273,
146
+ "<|sw|>": 50318,
147
+ "<|ta|>": 50287,
148
+ "<|te|>": 50299,
149
+ "<|tg|>": 50331,
150
+ "<|th|>": 50289,
151
+ "<|tk|>": 50341,
152
+ "<|tl|>": 50348,
153
+ "<|tr|>": 50268,
154
+ "<|tt|>": 50351,
155
+ "<|uk|>": 50280,
156
+ "<|ur|>": 50290,
157
+ "<|uz|>": 50337,
158
+ "<|vi|>": 50278,
159
+ "<|yi|>": 50335,
160
+ "<|yo|>": 50325,
161
+ "<|zh|>": 50260
162
+ },
163
+ "max_initial_timestamp_index": 1,
164
+ "max_length": 448,
165
+ "no_timestamps_token_id": 50363,
166
+ "pad_token_id": 50257,
167
+ "return_timestamps": false,
168
+ "suppress_tokens": [
169
+ 1,
170
+ 2,
171
+ 7,
172
+ 8,
173
+ 9,
174
+ 10,
175
+ 14,
176
+ 25,
177
+ 26,
178
+ 27,
179
+ 28,
180
+ 29,
181
+ 31,
182
+ 58,
183
+ 59,
184
+ 60,
185
+ 61,
186
+ 62,
187
+ 63,
188
+ 90,
189
+ 91,
190
+ 92,
191
+ 93,
192
+ 359,
193
+ 503,
194
+ 522,
195
+ 542,
196
+ 873,
197
+ 893,
198
+ 902,
199
+ 918,
200
+ 922,
201
+ 931,
202
+ 1350,
203
+ 1853,
204
+ 1982,
205
+ 2460,
206
+ 2627,
207
+ 3246,
208
+ 3253,
209
+ 3268,
210
+ 3536,
211
+ 3846,
212
+ 3961,
213
+ 4183,
214
+ 4667,
215
+ 6585,
216
+ 6647,
217
+ 7273,
218
+ 9061,
219
+ 9383,
220
+ 10428,
221
+ 10929,
222
+ 11938,
223
+ 12033,
224
+ 12331,
225
+ 12562,
226
+ 13793,
227
+ 14157,
228
+ 14635,
229
+ 15265,
230
+ 15618,
231
+ 16553,
232
+ 16604,
233
+ 18362,
234
+ 18956,
235
+ 20075,
236
+ 21675,
237
+ 22520,
238
+ 26130,
239
+ 26161,
240
+ 26435,
241
+ 28279,
242
+ 29464,
243
+ 31650,
244
+ 32302,
245
+ 32470,
246
+ 36865,
247
+ 42863,
248
+ 47425,
249
+ 49870,
250
+ 50254,
251
+ 50258,
252
+ 50358,
253
+ 50359,
254
+ 50360,
255
+ 50361,
256
+ 50362
257
+ ],
258
+ "task_to_id": {
259
+ "transcribe": 50359,
260
+ "translate": 50358
261
+ },
262
+ "transformers_version": "4.35.0"
263
+ }
1/checkpoint-1200/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:97a5dc5df3a851c021fd0f8d2ae5188c14751e1bad89d586aca0df26d18ba7bd
3
+ size 966995080
1/checkpoint-1200/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cff392fe179b368e85bc9921bc1aa01a47a051db275fb95a7297349d383fc988
3
+ size 1925063607
1/checkpoint-1200/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
+ }
1/checkpoint-1200/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1ccfd66f234dbfa70dd48e3bbe4c6d7e1749387f8160290226bb31dbadeaebf4
3
+ size 14575
1/checkpoint-1200/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8ddf2281e531ea7d559da2b8886e658468b1eb6200d0e7bf4ffcd70761ac2800
3
+ size 627
1/checkpoint-1200/trainer_state.json ADDED
@@ -0,0 +1,415 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": 46.084900200095916,
3
+ "best_model_checkpoint": "/media/makhataei/Backups/Whisper-Small-Common-Voice/checkpoint-1100",
4
+ "epoch": 4.633204633204633,
5
+ "eval_steps": 100,
6
+ "global_step": 1200,
7
+ "is_hyper_param_search": false,
8
+ "is_local_process_zero": true,
9
+ "is_world_process_zero": true,
10
+ "log_history": [
11
+ {
12
+ "epoch": 0.1,
13
+ "learning_rate": 4.600000000000001e-06,
14
+ "loss": 3.1025,
15
+ "step": 25
16
+ },
17
+ {
18
+ "epoch": 0.19,
19
+ "learning_rate": 9.600000000000001e-06,
20
+ "loss": 1.3525,
21
+ "step": 50
22
+ },
23
+ {
24
+ "epoch": 0.29,
25
+ "learning_rate": 9.976884422110554e-06,
26
+ "loss": 0.9564,
27
+ "step": 75
28
+ },
29
+ {
30
+ "epoch": 0.39,
31
+ "learning_rate": 9.95175879396985e-06,
32
+ "loss": 0.7495,
33
+ "step": 100
34
+ },
35
+ {
36
+ "epoch": 0.39,
37
+ "eval_loss": 0.860159158706665,
38
+ "eval_runtime": 1426.2695,
39
+ "eval_samples_per_second": 6.062,
40
+ "eval_steps_per_second": 0.606,
41
+ "eval_wer": 58.07742554282218,
42
+ "step": 100
43
+ },
44
+ {
45
+ "epoch": 0.48,
46
+ "learning_rate": 9.926633165829147e-06,
47
+ "loss": 0.5073,
48
+ "step": 125
49
+ },
50
+ {
51
+ "epoch": 0.58,
52
+ "learning_rate": 9.901507537688444e-06,
53
+ "loss": 0.3749,
54
+ "step": 150
55
+ },
56
+ {
57
+ "epoch": 0.68,
58
+ "learning_rate": 9.87638190954774e-06,
59
+ "loss": 0.3247,
60
+ "step": 175
61
+ },
62
+ {
63
+ "epoch": 0.77,
64
+ "learning_rate": 9.851256281407035e-06,
65
+ "loss": 0.3059,
66
+ "step": 200
67
+ },
68
+ {
69
+ "epoch": 0.77,
70
+ "eval_loss": 0.5538379549980164,
71
+ "eval_runtime": 1589.2333,
72
+ "eval_samples_per_second": 5.44,
73
+ "eval_steps_per_second": 0.544,
74
+ "eval_wer": 55.22481850804518,
75
+ "step": 200
76
+ },
77
+ {
78
+ "epoch": 0.87,
79
+ "learning_rate": 9.826130653266333e-06,
80
+ "loss": 0.2963,
81
+ "step": 225
82
+ },
83
+ {
84
+ "epoch": 0.97,
85
+ "learning_rate": 9.80100502512563e-06,
86
+ "loss": 0.2882,
87
+ "step": 250
88
+ },
89
+ {
90
+ "epoch": 1.06,
91
+ "learning_rate": 9.775879396984925e-06,
92
+ "loss": 0.2327,
93
+ "step": 275
94
+ },
95
+ {
96
+ "epoch": 1.16,
97
+ "learning_rate": 9.750753768844222e-06,
98
+ "loss": 0.2002,
99
+ "step": 300
100
+ },
101
+ {
102
+ "epoch": 1.16,
103
+ "eval_loss": 0.5136818885803223,
104
+ "eval_runtime": 1646.1357,
105
+ "eval_samples_per_second": 5.252,
106
+ "eval_steps_per_second": 0.525,
107
+ "eval_wer": 52.46151047609599,
108
+ "step": 300
109
+ },
110
+ {
111
+ "epoch": 1.25,
112
+ "learning_rate": 9.725628140703518e-06,
113
+ "loss": 0.1893,
114
+ "step": 325
115
+ },
116
+ {
117
+ "epoch": 1.35,
118
+ "learning_rate": 9.700502512562815e-06,
119
+ "loss": 0.185,
120
+ "step": 350
121
+ },
122
+ {
123
+ "epoch": 1.45,
124
+ "learning_rate": 9.675376884422111e-06,
125
+ "loss": 0.1847,
126
+ "step": 375
127
+ },
128
+ {
129
+ "epoch": 1.54,
130
+ "learning_rate": 9.650251256281408e-06,
131
+ "loss": 0.1798,
132
+ "step": 400
133
+ },
134
+ {
135
+ "epoch": 1.54,
136
+ "eval_loss": 0.5102902054786682,
137
+ "eval_runtime": 1650.3734,
138
+ "eval_samples_per_second": 5.239,
139
+ "eval_steps_per_second": 0.524,
140
+ "eval_wer": 51.65781945064576,
141
+ "step": 400
142
+ },
143
+ {
144
+ "epoch": 1.64,
145
+ "learning_rate": 9.625125628140705e-06,
146
+ "loss": 0.1804,
147
+ "step": 425
148
+ },
149
+ {
150
+ "epoch": 1.74,
151
+ "learning_rate": 9.600000000000001e-06,
152
+ "loss": 0.1807,
153
+ "step": 450
154
+ },
155
+ {
156
+ "epoch": 1.83,
157
+ "learning_rate": 9.574874371859298e-06,
158
+ "loss": 0.1714,
159
+ "step": 475
160
+ },
161
+ {
162
+ "epoch": 1.93,
163
+ "learning_rate": 9.549748743718595e-06,
164
+ "loss": 0.1678,
165
+ "step": 500
166
+ },
167
+ {
168
+ "epoch": 1.93,
169
+ "eval_loss": 0.48063144087791443,
170
+ "eval_runtime": 1482.5018,
171
+ "eval_samples_per_second": 5.832,
172
+ "eval_steps_per_second": 0.583,
173
+ "eval_wer": 46.55123943708554,
174
+ "step": 500
175
+ },
176
+ {
177
+ "epoch": 2.03,
178
+ "learning_rate": 9.52462311557789e-06,
179
+ "loss": 0.1542,
180
+ "step": 525
181
+ },
182
+ {
183
+ "epoch": 2.12,
184
+ "learning_rate": 9.499497487437188e-06,
185
+ "loss": 0.0929,
186
+ "step": 550
187
+ },
188
+ {
189
+ "epoch": 2.22,
190
+ "learning_rate": 9.474371859296483e-06,
191
+ "loss": 0.0936,
192
+ "step": 575
193
+ },
194
+ {
195
+ "epoch": 2.32,
196
+ "learning_rate": 9.44924623115578e-06,
197
+ "loss": 0.0932,
198
+ "step": 600
199
+ },
200
+ {
201
+ "epoch": 2.32,
202
+ "eval_loss": 0.49363330006599426,
203
+ "eval_runtime": 1993.4903,
204
+ "eval_samples_per_second": 4.337,
205
+ "eval_steps_per_second": 0.434,
206
+ "eval_wer": 47.49549370772767,
207
+ "step": 600
208
+ },
209
+ {
210
+ "epoch": 2.41,
211
+ "learning_rate": 9.424120603015076e-06,
212
+ "loss": 0.0936,
213
+ "step": 625
214
+ },
215
+ {
216
+ "epoch": 2.51,
217
+ "learning_rate": 9.398994974874373e-06,
218
+ "loss": 0.0955,
219
+ "step": 650
220
+ },
221
+ {
222
+ "epoch": 2.61,
223
+ "learning_rate": 9.373869346733669e-06,
224
+ "loss": 0.0871,
225
+ "step": 675
226
+ },
227
+ {
228
+ "epoch": 2.7,
229
+ "learning_rate": 9.348743718592966e-06,
230
+ "loss": 0.0944,
231
+ "step": 700
232
+ },
233
+ {
234
+ "epoch": 2.7,
235
+ "eval_loss": 0.5119881629943848,
236
+ "eval_runtime": 1791.3192,
237
+ "eval_samples_per_second": 4.827,
238
+ "eval_steps_per_second": 0.483,
239
+ "eval_wer": 48.91600932678474,
240
+ "step": 700
241
+ },
242
+ {
243
+ "epoch": 2.8,
244
+ "learning_rate": 9.323618090452262e-06,
245
+ "loss": 0.0927,
246
+ "step": 725
247
+ },
248
+ {
249
+ "epoch": 2.9,
250
+ "learning_rate": 9.298492462311557e-06,
251
+ "loss": 0.0964,
252
+ "step": 750
253
+ },
254
+ {
255
+ "epoch": 2.99,
256
+ "learning_rate": 9.273366834170856e-06,
257
+ "loss": 0.1038,
258
+ "step": 775
259
+ },
260
+ {
261
+ "epoch": 3.09,
262
+ "learning_rate": 9.248241206030152e-06,
263
+ "loss": 0.0493,
264
+ "step": 800
265
+ },
266
+ {
267
+ "epoch": 3.09,
268
+ "eval_loss": 0.5198288559913635,
269
+ "eval_runtime": 1878.4077,
270
+ "eval_samples_per_second": 4.603,
271
+ "eval_steps_per_second": 0.46,
272
+ "eval_wer": 48.26114997271419,
273
+ "step": 800
274
+ },
275
+ {
276
+ "epoch": 3.19,
277
+ "learning_rate": 9.223115577889447e-06,
278
+ "loss": 0.0501,
279
+ "step": 825
280
+ },
281
+ {
282
+ "epoch": 3.28,
283
+ "learning_rate": 9.197989949748744e-06,
284
+ "loss": 0.0458,
285
+ "step": 850
286
+ },
287
+ {
288
+ "epoch": 3.38,
289
+ "learning_rate": 9.17286432160804e-06,
290
+ "loss": 0.0492,
291
+ "step": 875
292
+ },
293
+ {
294
+ "epoch": 3.47,
295
+ "learning_rate": 9.147738693467337e-06,
296
+ "loss": 0.0506,
297
+ "step": 900
298
+ },
299
+ {
300
+ "epoch": 3.47,
301
+ "eval_loss": 0.5227376222610474,
302
+ "eval_runtime": 1759.968,
303
+ "eval_samples_per_second": 4.913,
304
+ "eval_steps_per_second": 0.491,
305
+ "eval_wer": 47.41280944585008,
306
+ "step": 900
307
+ },
308
+ {
309
+ "epoch": 3.57,
310
+ "learning_rate": 9.122613065326634e-06,
311
+ "loss": 0.0494,
312
+ "step": 925
313
+ },
314
+ {
315
+ "epoch": 3.67,
316
+ "learning_rate": 9.09748743718593e-06,
317
+ "loss": 0.051,
318
+ "step": 950
319
+ },
320
+ {
321
+ "epoch": 3.76,
322
+ "learning_rate": 9.072361809045227e-06,
323
+ "loss": 0.0466,
324
+ "step": 975
325
+ },
326
+ {
327
+ "epoch": 3.86,
328
+ "learning_rate": 9.047236180904524e-06,
329
+ "loss": 0.0482,
330
+ "step": 1000
331
+ },
332
+ {
333
+ "epoch": 3.86,
334
+ "eval_loss": 0.5255833268165588,
335
+ "eval_runtime": 1701.3645,
336
+ "eval_samples_per_second": 5.082,
337
+ "eval_steps_per_second": 0.508,
338
+ "eval_wer": 46.94977757933555,
339
+ "step": 1000
340
+ },
341
+ {
342
+ "epoch": 3.96,
343
+ "learning_rate": 9.02211055276382e-06,
344
+ "loss": 0.046,
345
+ "step": 1025
346
+ },
347
+ {
348
+ "epoch": 4.05,
349
+ "learning_rate": 8.996984924623117e-06,
350
+ "loss": 0.0368,
351
+ "step": 1050
352
+ },
353
+ {
354
+ "epoch": 4.15,
355
+ "learning_rate": 8.971859296482412e-06,
356
+ "loss": 0.0233,
357
+ "step": 1075
358
+ },
359
+ {
360
+ "epoch": 4.25,
361
+ "learning_rate": 8.94673366834171e-06,
362
+ "loss": 0.0244,
363
+ "step": 1100
364
+ },
365
+ {
366
+ "epoch": 4.25,
367
+ "eval_loss": 0.5491495132446289,
368
+ "eval_runtime": 1619.0836,
369
+ "eval_samples_per_second": 5.34,
370
+ "eval_steps_per_second": 0.534,
371
+ "eval_wer": 46.084900200095916,
372
+ "step": 1100
373
+ },
374
+ {
375
+ "epoch": 4.34,
376
+ "learning_rate": 8.921608040201005e-06,
377
+ "loss": 0.0258,
378
+ "step": 1125
379
+ },
380
+ {
381
+ "epoch": 4.44,
382
+ "learning_rate": 8.896482412060302e-06,
383
+ "loss": 0.0248,
384
+ "step": 1150
385
+ },
386
+ {
387
+ "epoch": 4.54,
388
+ "learning_rate": 8.8713567839196e-06,
389
+ "loss": 0.0241,
390
+ "step": 1175
391
+ },
392
+ {
393
+ "epoch": 4.63,
394
+ "learning_rate": 8.846231155778895e-06,
395
+ "loss": 0.0272,
396
+ "step": 1200
397
+ },
398
+ {
399
+ "epoch": 4.63,
400
+ "eval_loss": 0.5620735883712769,
401
+ "eval_runtime": 1837.7585,
402
+ "eval_samples_per_second": 4.705,
403
+ "eval_steps_per_second": 0.471,
404
+ "eval_wer": 46.916703874584506,
405
+ "step": 1200
406
+ }
407
+ ],
408
+ "logging_steps": 25,
409
+ "max_steps": 10000,
410
+ "num_train_epochs": 39,
411
+ "save_steps": 100,
412
+ "total_flos": 1.93871671934976e+19,
413
+ "trial_name": null,
414
+ "trial_params": null
415
+ }
1/checkpoint-1200/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d286c349fc7a4c7419632d8b055a552e2835b4138f3ff268d56552451405f53b
3
+ size 4347
1/checkpoint-1300/config.json ADDED
@@ -0,0 +1,52 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "openai/whisper-small",
3
+ "activation_dropout": 0.0,
4
+ "activation_function": "gelu",
5
+ "apply_spec_augment": false,
6
+ "architectures": [
7
+ "WhisperForConditionalGeneration"
8
+ ],
9
+ "attention_dropout": 0.0,
10
+ "begin_suppress_tokens": [
11
+ 220,
12
+ 50257
13
+ ],
14
+ "bos_token_id": 50257,
15
+ "classifier_proj_size": 256,
16
+ "d_model": 768,
17
+ "decoder_attention_heads": 12,
18
+ "decoder_ffn_dim": 3072,
19
+ "decoder_layerdrop": 0.0,
20
+ "decoder_layers": 12,
21
+ "decoder_start_token_id": 50258,
22
+ "dropout": 0.0,
23
+ "encoder_attention_heads": 12,
24
+ "encoder_ffn_dim": 3072,
25
+ "encoder_layerdrop": 0.0,
26
+ "encoder_layers": 12,
27
+ "eos_token_id": 50257,
28
+ "forced_decoder_ids": null,
29
+ "init_std": 0.02,
30
+ "is_encoder_decoder": true,
31
+ "mask_feature_length": 10,
32
+ "mask_feature_min_masks": 0,
33
+ "mask_feature_prob": 0.0,
34
+ "mask_time_length": 10,
35
+ "mask_time_min_masks": 2,
36
+ "mask_time_prob": 0.05,
37
+ "max_length": 448,
38
+ "max_source_positions": 1500,
39
+ "max_target_positions": 448,
40
+ "median_filter_width": 7,
41
+ "model_type": "whisper",
42
+ "num_hidden_layers": 12,
43
+ "num_mel_bins": 80,
44
+ "pad_token_id": 50257,
45
+ "scale_embedding": false,
46
+ "suppress_tokens": [],
47
+ "torch_dtype": "float32",
48
+ "transformers_version": "4.35.0",
49
+ "use_cache": true,
50
+ "use_weighted_layer_sum": false,
51
+ "vocab_size": 51865
52
+ }
1/checkpoint-1300/generation_config.json ADDED
@@ -0,0 +1,263 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "alignment_heads": [
3
+ [
4
+ 5,
5
+ 3
6
+ ],
7
+ [
8
+ 5,
9
+ 9
10
+ ],
11
+ [
12
+ 8,
13
+ 0
14
+ ],
15
+ [
16
+ 8,
17
+ 4
18
+ ],
19
+ [
20
+ 8,
21
+ 7
22
+ ],
23
+ [
24
+ 8,
25
+ 8
26
+ ],
27
+ [
28
+ 9,
29
+ 0
30
+ ],
31
+ [
32
+ 9,
33
+ 7
34
+ ],
35
+ [
36
+ 9,
37
+ 9
38
+ ],
39
+ [
40
+ 10,
41
+ 5
42
+ ]
43
+ ],
44
+ "begin_suppress_tokens": [
45
+ 220,
46
+ 50257
47
+ ],
48
+ "bos_token_id": 50257,
49
+ "decoder_start_token_id": 50258,
50
+ "eos_token_id": 50257,
51
+ "forced_decoder_ids": [
52
+ [
53
+ 1,
54
+ null
55
+ ],
56
+ [
57
+ 2,
58
+ 50359
59
+ ]
60
+ ],
61
+ "is_multilingual": true,
62
+ "lang_to_id": {
63
+ "<|af|>": 50327,
64
+ "<|am|>": 50334,
65
+ "<|ar|>": 50272,
66
+ "<|as|>": 50350,
67
+ "<|az|>": 50304,
68
+ "<|ba|>": 50355,
69
+ "<|be|>": 50330,
70
+ "<|bg|>": 50292,
71
+ "<|bn|>": 50302,
72
+ "<|bo|>": 50347,
73
+ "<|br|>": 50309,
74
+ "<|bs|>": 50315,
75
+ "<|ca|>": 50270,
76
+ "<|cs|>": 50283,
77
+ "<|cy|>": 50297,
78
+ "<|da|>": 50285,
79
+ "<|de|>": 50261,
80
+ "<|el|>": 50281,
81
+ "<|en|>": 50259,
82
+ "<|es|>": 50262,
83
+ "<|et|>": 50307,
84
+ "<|eu|>": 50310,
85
+ "<|fa|>": 50300,
86
+ "<|fi|>": 50277,
87
+ "<|fo|>": 50338,
88
+ "<|fr|>": 50265,
89
+ "<|gl|>": 50319,
90
+ "<|gu|>": 50333,
91
+ "<|haw|>": 50352,
92
+ "<|ha|>": 50354,
93
+ "<|he|>": 50279,
94
+ "<|hi|>": 50276,
95
+ "<|hr|>": 50291,
96
+ "<|ht|>": 50339,
97
+ "<|hu|>": 50286,
98
+ "<|hy|>": 50312,
99
+ "<|id|>": 50275,
100
+ "<|is|>": 50311,
101
+ "<|it|>": 50274,
102
+ "<|ja|>": 50266,
103
+ "<|jw|>": 50356,
104
+ "<|ka|>": 50329,
105
+ "<|kk|>": 50316,
106
+ "<|km|>": 50323,
107
+ "<|kn|>": 50306,
108
+ "<|ko|>": 50264,
109
+ "<|la|>": 50294,
110
+ "<|lb|>": 50345,
111
+ "<|ln|>": 50353,
112
+ "<|lo|>": 50336,
113
+ "<|lt|>": 50293,
114
+ "<|lv|>": 50301,
115
+ "<|mg|>": 50349,
116
+ "<|mi|>": 50295,
117
+ "<|mk|>": 50308,
118
+ "<|ml|>": 50296,
119
+ "<|mn|>": 50314,
120
+ "<|mr|>": 50320,
121
+ "<|ms|>": 50282,
122
+ "<|mt|>": 50343,
123
+ "<|my|>": 50346,
124
+ "<|ne|>": 50313,
125
+ "<|nl|>": 50271,
126
+ "<|nn|>": 50342,
127
+ "<|no|>": 50288,
128
+ "<|oc|>": 50328,
129
+ "<|pa|>": 50321,
130
+ "<|pl|>": 50269,
131
+ "<|ps|>": 50340,
132
+ "<|pt|>": 50267,
133
+ "<|ro|>": 50284,
134
+ "<|ru|>": 50263,
135
+ "<|sa|>": 50344,
136
+ "<|sd|>": 50332,
137
+ "<|si|>": 50322,
138
+ "<|sk|>": 50298,
139
+ "<|sl|>": 50305,
140
+ "<|sn|>": 50324,
141
+ "<|so|>": 50326,
142
+ "<|sq|>": 50317,
143
+ "<|sr|>": 50303,
144
+ "<|su|>": 50357,
145
+ "<|sv|>": 50273,
146
+ "<|sw|>": 50318,
147
+ "<|ta|>": 50287,
148
+ "<|te|>": 50299,
149
+ "<|tg|>": 50331,
150
+ "<|th|>": 50289,
151
+ "<|tk|>": 50341,
152
+ "<|tl|>": 50348,
153
+ "<|tr|>": 50268,
154
+ "<|tt|>": 50351,
155
+ "<|uk|>": 50280,
156
+ "<|ur|>": 50290,
157
+ "<|uz|>": 50337,
158
+ "<|vi|>": 50278,
159
+ "<|yi|>": 50335,
160
+ "<|yo|>": 50325,
161
+ "<|zh|>": 50260
162
+ },
163
+ "max_initial_timestamp_index": 1,
164
+ "max_length": 448,
165
+ "no_timestamps_token_id": 50363,
166
+ "pad_token_id": 50257,
167
+ "return_timestamps": false,
168
+ "suppress_tokens": [
169
+ 1,
170
+ 2,
171
+ 7,
172
+ 8,
173
+ 9,
174
+ 10,
175
+ 14,
176
+ 25,
177
+ 26,
178
+ 27,
179
+ 28,
180
+ 29,
181
+ 31,
182
+ 58,
183
+ 59,
184
+ 60,
185
+ 61,
186
+ 62,
187
+ 63,
188
+ 90,
189
+ 91,
190
+ 92,
191
+ 93,
192
+ 359,
193
+ 503,
194
+ 522,
195
+ 542,
196
+ 873,
197
+ 893,
198
+ 902,
199
+ 918,
200
+ 922,
201
+ 931,
202
+ 1350,
203
+ 1853,
204
+ 1982,
205
+ 2460,
206
+ 2627,
207
+ 3246,
208
+ 3253,
209
+ 3268,
210
+ 3536,
211
+ 3846,
212
+ 3961,
213
+ 4183,
214
+ 4667,
215
+ 6585,
216
+ 6647,
217
+ 7273,
218
+ 9061,
219
+ 9383,
220
+ 10428,
221
+ 10929,
222
+ 11938,
223
+ 12033,
224
+ 12331,
225
+ 12562,
226
+ 13793,
227
+ 14157,
228
+ 14635,
229
+ 15265,
230
+ 15618,
231
+ 16553,
232
+ 16604,
233
+ 18362,
234
+ 18956,
235
+ 20075,
236
+ 21675,
237
+ 22520,
238
+ 26130,
239
+ 26161,
240
+ 26435,
241
+ 28279,
242
+ 29464,
243
+ 31650,
244
+ 32302,
245
+ 32470,
246
+ 36865,
247
+ 42863,
248
+ 47425,
249
+ 49870,
250
+ 50254,
251
+ 50258,
252
+ 50358,
253
+ 50359,
254
+ 50360,
255
+ 50361,
256
+ 50362
257
+ ],
258
+ "task_to_id": {
259
+ "transcribe": 50359,
260
+ "translate": 50358
261
+ },
262
+ "transformers_version": "4.35.0"
263
+ }
1/checkpoint-1300/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a397a0a42bc81ec3034ae7d021a1ba6fe949e9583f2738314fcdc76f4505a5c2
3
+ size 966995080
1/checkpoint-1300/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4f4ca09edcbca4ea196bef925b429b8a3d1d3ba3368eb3ef34905e0af83d5497
3
+ size 1925063607
1/checkpoint-1300/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
+ }