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
2ecb144
•
1 Parent(s): 814c436

Training in progress, step 600

Browse files
This view is limited to 50 files because it contains too many changes.   See raw diff
Files changed (50) hide show
  1. 2/checkpoint-100/config.json +52 -0
  2. 2/checkpoint-100/generation_config.json +263 -0
  3. 2/checkpoint-100/model.safetensors +3 -0
  4. 2/checkpoint-100/optimizer.pt +3 -0
  5. 2/checkpoint-100/preprocessor_config.json +14 -0
  6. 2/checkpoint-100/rng_state.pth +3 -0
  7. 2/checkpoint-100/scheduler.pt +3 -0
  8. 2/checkpoint-100/trainer_state.json +52 -0
  9. 2/checkpoint-100/training_args.bin +3 -0
  10. 2/checkpoint-10000/config.json +52 -0
  11. 2/checkpoint-10000/generation_config.json +263 -0
  12. 2/checkpoint-10000/model.safetensors +3 -0
  13. 2/checkpoint-10000/optimizer.pt +3 -0
  14. 2/checkpoint-10000/preprocessor_config.json +14 -0
  15. 2/checkpoint-10000/rng_state.pth +3 -0
  16. 2/checkpoint-10000/scheduler.pt +3 -0
  17. 2/checkpoint-10000/trainer_state.json +3319 -0
  18. 2/checkpoint-10000/training_args.bin +3 -0
  19. 2/checkpoint-200/config.json +52 -0
  20. 2/checkpoint-200/generation_config.json +263 -0
  21. 2/checkpoint-200/model.safetensors +3 -0
  22. 2/checkpoint-200/optimizer.pt +3 -0
  23. 2/checkpoint-200/preprocessor_config.json +14 -0
  24. 2/checkpoint-200/rng_state.pth +3 -0
  25. 2/checkpoint-200/scheduler.pt +3 -0
  26. 2/checkpoint-200/trainer_state.json +85 -0
  27. 2/checkpoint-200/training_args.bin +3 -0
  28. 2/checkpoint-300/config.json +52 -0
  29. 2/checkpoint-300/generation_config.json +263 -0
  30. 2/checkpoint-300/model.safetensors +3 -0
  31. 2/checkpoint-300/optimizer.pt +3 -0
  32. 2/checkpoint-300/preprocessor_config.json +14 -0
  33. 2/checkpoint-300/rng_state.pth +3 -0
  34. 2/checkpoint-300/scheduler.pt +3 -0
  35. 2/checkpoint-300/trainer_state.json +118 -0
  36. 2/checkpoint-300/training_args.bin +3 -0
  37. 2/checkpoint-400/config.json +52 -0
  38. 2/checkpoint-400/generation_config.json +263 -0
  39. 2/checkpoint-400/model.safetensors +3 -0
  40. 2/checkpoint-400/optimizer.pt +3 -0
  41. 2/checkpoint-400/preprocessor_config.json +14 -0
  42. 2/checkpoint-400/rng_state.pth +3 -0
  43. 2/checkpoint-400/scheduler.pt +3 -0
  44. 2/checkpoint-400/trainer_state.json +151 -0
  45. 2/checkpoint-400/training_args.bin +3 -0
  46. 2/checkpoint-9900/config.json +52 -0
  47. 2/checkpoint-9900/generation_config.json +263 -0
  48. 2/checkpoint-9900/model.safetensors +3 -0
  49. 2/checkpoint-9900/optimizer.pt +3 -0
  50. 2/checkpoint-9900/preprocessor_config.json +14 -0
2/checkpoint-100/config.json ADDED
@@ -0,0 +1,52 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "makhataei/Whisper-Small-Common-Voice",
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
+ }
2/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
+ }
2/checkpoint-100/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:205238ee740de2cd60895d324e096954a680e34479fc745d364e70127d96ec59
3
+ size 966995080
2/checkpoint-100/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0d6cf3d64de93f857a39fc8ba2cb150bb9b0c25671c537d7ac1aa6fe793c90de
3
+ size 1925063607
2/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
+ }
2/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
2/checkpoint-100/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f4386dc68e594fd4f4096cfca8d4e6b7bc06590992d114a94cd582d476f2e94f
3
+ size 627
2/checkpoint-100/trainer_state.json ADDED
@@ -0,0 +1,52 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": 49.12602735195383,
3
+ "best_model_checkpoint": "/media/makhataei/Backups/Whisper-Small-Common-Voice/checkpoint-100",
4
+ "epoch": 0.39254170755642787,
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": 5e-06,
14
+ "loss": 0.3008,
15
+ "step": 25
16
+ },
17
+ {
18
+ "epoch": 0.2,
19
+ "learning_rate": 1e-05,
20
+ "loss": 0.2368,
21
+ "step": 50
22
+ },
23
+ {
24
+ "epoch": 0.29,
25
+ "learning_rate": 9.974874371859297e-06,
26
+ "loss": 0.2001,
27
+ "step": 75
28
+ },
29
+ {
30
+ "epoch": 0.39,
31
+ "learning_rate": 9.949748743718594e-06,
32
+ "loss": 0.1801,
33
+ "step": 100
34
+ },
35
+ {
36
+ "epoch": 0.39,
37
+ "eval_loss": 0.4975946843624115,
38
+ "eval_runtime": 1546.7344,
39
+ "eval_samples_per_second": 5.59,
40
+ "eval_steps_per_second": 0.559,
41
+ "eval_wer": 49.12602735195383,
42
+ "step": 100
43
+ }
44
+ ],
45
+ "logging_steps": 25,
46
+ "max_steps": 10000,
47
+ "num_train_epochs": 40,
48
+ "save_steps": 100,
49
+ "total_flos": 1.616078241792e+18,
50
+ "trial_name": null,
51
+ "trial_params": null
52
+ }
2/checkpoint-100/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c69a6da4c323af31ce3f4997ac3ae843fb847a604926bb5641ee181988d9e27c
3
+ size 4347
2/checkpoint-10000/config.json ADDED
@@ -0,0 +1,52 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "makhataei/Whisper-Small-Common-Voice",
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
+ }
2/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
+ }
2/checkpoint-10000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6ecb99288f3bd45a1d9f276f5d52acce0b6eea2cc551612c04a1ac1d843aed8f
3
+ size 966995080
2/checkpoint-10000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2a140cf7f63cb787ffcaa98dc42aecdb2088c9ccf74f5f1c3786eb4eb88ed266
3
+ size 1925063607
2/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
+ }
2/checkpoint-10000/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6ce12248152ae56b00a9a9adcb35cdf6c5657fd39b42e42f2fab6f5d5c4ebfe7
3
+ size 14575
2/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
2/checkpoint-10000/trainer_state.json ADDED
@@ -0,0 +1,3319 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": 43.17606786724215,
3
+ "best_model_checkpoint": "/media/makhataei/Backups/Whisper-Small-Common-Voice/checkpoint-300",
4
+ "epoch": 39.25417075564279,
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": 5e-06,
14
+ "loss": 0.3008,
15
+ "step": 25
16
+ },
17
+ {
18
+ "epoch": 0.2,
19
+ "learning_rate": 1e-05,
20
+ "loss": 0.2368,
21
+ "step": 50
22
+ },
23
+ {
24
+ "epoch": 0.29,
25
+ "learning_rate": 9.974874371859297e-06,
26
+ "loss": 0.2001,
27
+ "step": 75
28
+ },
29
+ {
30
+ "epoch": 0.39,
31
+ "learning_rate": 9.949748743718594e-06,
32
+ "loss": 0.1801,
33
+ "step": 100
34
+ },
35
+ {
36
+ "epoch": 0.39,
37
+ "eval_loss": 0.4975946843624115,
38
+ "eval_runtime": 1546.7344,
39
+ "eval_samples_per_second": 5.59,
40
+ "eval_steps_per_second": 0.559,
41
+ "eval_wer": 49.12602735195383,
42
+ "step": 100
43
+ },
44
+ {
45
+ "epoch": 0.49,
46
+ "learning_rate": 9.92462311557789e-06,
47
+ "loss": 0.1713,
48
+ "step": 125
49
+ },
50
+ {
51
+ "epoch": 0.59,
52
+ "learning_rate": 9.899497487437186e-06,
53
+ "loss": 0.1657,
54
+ "step": 150
55
+ },
56
+ {
57
+ "epoch": 0.69,
58
+ "learning_rate": 9.874371859296484e-06,
59
+ "loss": 0.1574,
60
+ "step": 175
61
+ },
62
+ {
63
+ "epoch": 0.79,
64
+ "learning_rate": 9.84924623115578e-06,
65
+ "loss": 0.1597,
66
+ "step": 200
67
+ },
68
+ {
69
+ "epoch": 0.79,
70
+ "eval_loss": 0.46244511008262634,
71
+ "eval_runtime": 1507.9912,
72
+ "eval_samples_per_second": 5.733,
73
+ "eval_steps_per_second": 0.574,
74
+ "eval_wer": 46.749681665591766,
75
+ "step": 200
76
+ },
77
+ {
78
+ "epoch": 0.88,
79
+ "learning_rate": 9.824120603015075e-06,
80
+ "loss": 0.1509,
81
+ "step": 225
82
+ },
83
+ {
84
+ "epoch": 0.98,
85
+ "learning_rate": 9.798994974874372e-06,
86
+ "loss": 0.1434,
87
+ "step": 250
88
+ },
89
+ {
90
+ "epoch": 1.08,
91
+ "learning_rate": 9.773869346733669e-06,
92
+ "loss": 0.0919,
93
+ "step": 275
94
+ },
95
+ {
96
+ "epoch": 1.18,
97
+ "learning_rate": 9.748743718592965e-06,
98
+ "loss": 0.0776,
99
+ "step": 300
100
+ },
101
+ {
102
+ "epoch": 1.18,
103
+ "eval_loss": 0.4794461131095886,
104
+ "eval_runtime": 1499.4378,
105
+ "eval_samples_per_second": 5.766,
106
+ "eval_steps_per_second": 0.577,
107
+ "eval_wer": 43.17606786724215,
108
+ "step": 300
109
+ },
110
+ {
111
+ "epoch": 1.28,
112
+ "learning_rate": 9.723618090452262e-06,
113
+ "loss": 0.0772,
114
+ "step": 325
115
+ },
116
+ {
117
+ "epoch": 1.37,
118
+ "learning_rate": 9.698492462311559e-06,
119
+ "loss": 0.0799,
120
+ "step": 350
121
+ },
122
+ {
123
+ "epoch": 1.47,
124
+ "learning_rate": 9.673366834170855e-06,
125
+ "loss": 0.0872,
126
+ "step": 375
127
+ },
128
+ {
129
+ "epoch": 1.57,
130
+ "learning_rate": 9.648241206030152e-06,
131
+ "loss": 0.083,
132
+ "step": 400
133
+ },
134
+ {
135
+ "epoch": 1.57,
136
+ "eval_loss": 0.48234912753105164,
137
+ "eval_runtime": 1500.1085,
138
+ "eval_samples_per_second": 5.764,
139
+ "eval_steps_per_second": 0.577,
140
+ "eval_wer": 43.802814572274315,
141
+ "step": 400
142
+ },
143
+ {
144
+ "epoch": 1.67,
145
+ "learning_rate": 9.623115577889448e-06,
146
+ "loss": 0.0807,
147
+ "step": 425
148
+ },
149
+ {
150
+ "epoch": 1.77,
151
+ "learning_rate": 9.597989949748745e-06,
152
+ "loss": 0.0815,
153
+ "step": 450
154
+ },
155
+ {
156
+ "epoch": 1.86,
157
+ "learning_rate": 9.57286432160804e-06,
158
+ "loss": 0.0803,
159
+ "step": 475
160
+ },
161
+ {
162
+ "epoch": 1.96,
163
+ "learning_rate": 9.547738693467338e-06,
164
+ "loss": 0.0786,
165
+ "step": 500
166
+ },
167
+ {
168
+ "epoch": 1.96,
169
+ "eval_loss": 0.48833438754081726,
170
+ "eval_runtime": 1502.1848,
171
+ "eval_samples_per_second": 5.756,
172
+ "eval_steps_per_second": 0.576,
173
+ "eval_wer": 44.39152651684278,
174
+ "step": 500
175
+ },
176
+ {
177
+ "epoch": 2.06,
178
+ "learning_rate": 9.522613065326633e-06,
179
+ "loss": 0.0538,
180
+ "step": 525
181
+ },
182
+ {
183
+ "epoch": 2.16,
184
+ "learning_rate": 9.49748743718593e-06,
185
+ "loss": 0.0338,
186
+ "step": 550
187
+ },
188
+ {
189
+ "epoch": 2.26,
190
+ "learning_rate": 9.472361809045228e-06,
191
+ "loss": 0.0315,
192
+ "step": 575
193
+ },
194
+ {
195
+ "epoch": 2.36,
196
+ "learning_rate": 9.447236180904523e-06,
197
+ "loss": 0.0331,
198
+ "step": 600
199
+ },
200
+ {
201
+ "epoch": 2.36,
202
+ "eval_loss": 0.538537859916687,
203
+ "eval_runtime": 1606.4592,
204
+ "eval_samples_per_second": 5.382,
205
+ "eval_steps_per_second": 0.538,
206
+ "eval_wer": 46.243653982900895,
207
+ "step": 600
208
+ },
209
+ {
210
+ "epoch": 2.45,
211
+ "learning_rate": 9.42211055276382e-06,
212
+ "loss": 0.0347,
213
+ "step": 625
214
+ },
215
+ {
216
+ "epoch": 2.55,
217
+ "learning_rate": 9.396984924623116e-06,
218
+ "loss": 0.0319,
219
+ "step": 650
220
+ },
221
+ {
222
+ "epoch": 2.65,
223
+ "learning_rate": 9.371859296482413e-06,
224
+ "loss": 0.035,
225
+ "step": 675
226
+ },
227
+ {
228
+ "epoch": 2.75,
229
+ "learning_rate": 9.34673366834171e-06,
230
+ "loss": 0.0353,
231
+ "step": 700
232
+ },
233
+ {
234
+ "epoch": 2.75,
235
+ "eval_loss": 0.5605480670928955,
236
+ "eval_runtime": 1551.3357,
237
+ "eval_samples_per_second": 5.573,
238
+ "eval_steps_per_second": 0.558,
239
+ "eval_wer": 44.943857386185115,
240
+ "step": 700
241
+ },
242
+ {
243
+ "epoch": 2.85,
244
+ "learning_rate": 9.321608040201006e-06,
245
+ "loss": 0.0393,
246
+ "step": 725
247
+ },
248
+ {
249
+ "epoch": 2.94,
250
+ "learning_rate": 9.296482412060303e-06,
251
+ "loss": 0.0362,
252
+ "step": 750
253
+ },
254
+ {
255
+ "epoch": 3.04,
256
+ "learning_rate": 9.271356783919598e-06,
257
+ "loss": 0.0265,
258
+ "step": 775
259
+ },
260
+ {
261
+ "epoch": 3.14,
262
+ "learning_rate": 9.246231155778896e-06,
263
+ "loss": 0.0139,
264
+ "step": 800
265
+ },
266
+ {
267
+ "epoch": 3.14,
268
+ "eval_loss": 0.5941358804702759,
269
+ "eval_runtime": 1606.3004,
270
+ "eval_samples_per_second": 5.383,
271
+ "eval_steps_per_second": 0.539,
272
+ "eval_wer": 45.281209174645696,
273
+ "step": 800
274
+ },
275
+ {
276
+ "epoch": 3.24,
277
+ "learning_rate": 9.221105527638193e-06,
278
+ "loss": 0.0161,
279
+ "step": 825
280
+ },
281
+ {
282
+ "epoch": 3.34,
283
+ "learning_rate": 9.195979899497488e-06,
284
+ "loss": 0.0148,
285
+ "step": 850
286
+ },
287
+ {
288
+ "epoch": 3.43,
289
+ "learning_rate": 9.170854271356784e-06,
290
+ "loss": 0.0169,
291
+ "step": 875
292
+ },
293
+ {
294
+ "epoch": 3.53,
295
+ "learning_rate": 9.14572864321608e-06,
296
+ "loss": 0.0152,
297
+ "step": 900
298
+ },
299
+ {
300
+ "epoch": 3.53,
301
+ "eval_loss": 0.5978226065635681,
302
+ "eval_runtime": 1599.7307,
303
+ "eval_samples_per_second": 5.405,
304
+ "eval_steps_per_second": 0.541,
305
+ "eval_wer": 49.293049560946564,
306
+ "step": 900
307
+ },
308
+ {
309
+ "epoch": 3.63,
310
+ "learning_rate": 9.120603015075377e-06,
311
+ "loss": 0.0148,
312
+ "step": 925
313
+ },
314
+ {
315
+ "epoch": 3.73,
316
+ "learning_rate": 9.095477386934674e-06,
317
+ "loss": 0.0154,
318
+ "step": 950
319
+ },
320
+ {
321
+ "epoch": 3.83,
322
+ "learning_rate": 9.07035175879397e-06,
323
+ "loss": 0.0178,
324
+ "step": 975
325
+ },
326
+ {
327
+ "epoch": 3.93,
328
+ "learning_rate": 9.045226130653267e-06,
329
+ "loss": 0.0155,
330
+ "step": 1000
331
+ },
332
+ {
333
+ "epoch": 3.93,
334
+ "eval_loss": 0.6114091873168945,
335
+ "eval_runtime": 1660.7592,
336
+ "eval_samples_per_second": 5.206,
337
+ "eval_steps_per_second": 0.521,
338
+ "eval_wer": 49.97767524929305,
339
+ "step": 1000
340
+ },
341
+ {
342
+ "epoch": 4.02,
343
+ "learning_rate": 9.020100502512564e-06,
344
+ "loss": 0.0135,
345
+ "step": 1025
346
+ },
347
+ {
348
+ "epoch": 4.12,
349
+ "learning_rate": 8.99497487437186e-06,
350
+ "loss": 0.0069,
351
+ "step": 1050
352
+ },
353
+ {
354
+ "epoch": 4.22,
355
+ "learning_rate": 8.969849246231157e-06,
356
+ "loss": 0.006,
357
+ "step": 1075
358
+ },
359
+ {
360
+ "epoch": 4.32,
361
+ "learning_rate": 8.944723618090452e-06,
362
+ "loss": 0.0063,
363
+ "step": 1100
364
+ },
365
+ {
366
+ "epoch": 4.32,
367
+ "eval_loss": 0.6466676592826843,
368
+ "eval_runtime": 1663.0802,
369
+ "eval_samples_per_second": 5.199,
370
+ "eval_steps_per_second": 0.52,
371
+ "eval_wer": 50.00413421309388,
372
+ "step": 1100
373
+ },
374
+ {
375
+ "epoch": 4.42,
376
+ "learning_rate": 8.91959798994975e-06,
377
+ "loss": 0.0077,
378
+ "step": 1125
379
+ },
380
+ {
381
+ "epoch": 4.51,
382
+ "learning_rate": 8.894472361809045e-06,
383
+ "loss": 0.0063,
384
+ "step": 1150
385
+ },
386
+ {
387
+ "epoch": 4.61,
388
+ "learning_rate": 8.869346733668342e-06,
389
+ "loss": 0.0087,
390
+ "step": 1175
391
+ },
392
+ {
393
+ "epoch": 4.71,
394
+ "learning_rate": 8.84422110552764e-06,
395
+ "loss": 0.0079,
396
+ "step": 1200
397
+ },
398
+ {
399
+ "epoch": 4.71,
400
+ "eval_loss": 0.6383361220359802,
401
+ "eval_runtime": 1613.3597,
402
+ "eval_samples_per_second": 5.359,
403
+ "eval_steps_per_second": 0.536,
404
+ "eval_wer": 48.08751302277125,
405
+ "step": 1200
406
+ },
407
+ {
408
+ "epoch": 4.81,
409
+ "learning_rate": 8.819095477386935e-06,
410
+ "loss": 0.0081,
411
+ "step": 1225
412
+ },
413
+ {
414
+ "epoch": 4.91,
415
+ "learning_rate": 8.793969849246232e-06,
416
+ "loss": 0.0085,
417
+ "step": 1250
418
+ },
419
+ {
420
+ "epoch": 5.0,
421
+ "learning_rate": 8.768844221105528e-06,
422
+ "loss": 0.0094,
423
+ "step": 1275
424
+ },
425
+ {
426
+ "epoch": 5.1,
427
+ "learning_rate": 8.743718592964825e-06,
428
+ "loss": 0.0046,
429
+ "step": 1300
430
+ },
431
+ {
432
+ "epoch": 5.1,
433
+ "eval_loss": 0.6500093340873718,
434
+ "eval_runtime": 1575.0065,
435
+ "eval_samples_per_second": 5.49,
436
+ "eval_steps_per_second": 0.549,
437
+ "eval_wer": 45.49949562600254,
438
+ "step": 1300
439
+ },
440
+ {
441
+ "epoch": 5.2,
442
+ "learning_rate": 8.71859296482412e-06,
443
+ "loss": 0.0046,
444
+ "step": 1325
445
+ },
446
+ {
447
+ "epoch": 5.3,
448
+ "learning_rate": 8.693467336683418e-06,
449
+ "loss": 0.0058,
450
+ "step": 1350
451
+ },
452
+ {
453
+ "epoch": 5.4,
454
+ "learning_rate": 8.668341708542715e-06,
455
+ "loss": 0.0055,
456
+ "step": 1375
457
+ },
458
+ {
459
+ "epoch": 5.5,
460
+ "learning_rate": 8.64321608040201e-06,
461
+ "loss": 0.0042,
462
+ "step": 1400
463
+ },
464
+ {
465
+ "epoch": 5.5,
466
+ "eval_loss": 0.647581934928894,
467
+ "eval_runtime": 1605.9286,
468
+ "eval_samples_per_second": 5.384,
469
+ "eval_steps_per_second": 0.539,
470
+ "eval_wer": 47.44919052107622,
471
+ "step": 1400
472
+ },
473
+ {
474
+ "epoch": 5.59,
475
+ "learning_rate": 8.618090452261308e-06,
476
+ "loss": 0.0047,
477
+ "step": 1425
478
+ },
479
+ {
480
+ "epoch": 5.69,
481
+ "learning_rate": 8.592964824120603e-06,
482
+ "loss": 0.0049,
483
+ "step": 1450
484
+ },
485
+ {
486
+ "epoch": 5.79,
487
+ "learning_rate": 8.5678391959799e-06,
488
+ "loss": 0.0048,
489
+ "step": 1475
490
+ },
491
+ {
492
+ "epoch": 5.89,
493
+ "learning_rate": 8.542713567839196e-06,
494
+ "loss": 0.0052,
495
+ "step": 1500
496
+ },
497
+ {
498
+ "epoch": 5.89,
499
+ "eval_loss": 0.668513298034668,
500
+ "eval_runtime": 1668.8488,
501
+ "eval_samples_per_second": 5.181,
502
+ "eval_steps_per_second": 0.518,
503
+ "eval_wer": 52.18699872666237,
504
+ "step": 1500
505
+ },
506
+ {
507
+ "epoch": 5.99,
508
+ "learning_rate": 8.517587939698493e-06,
509
+ "loss": 0.006,
510
+ "step": 1525
511
+ },
512
+ {
513
+ "epoch": 6.08,
514
+ "learning_rate": 8.49246231155779e-06,
515
+ "loss": 0.0052,
516
+ "step": 1550
517
+ },
518
+ {
519
+ "epoch": 6.18,
520
+ "learning_rate": 8.467336683417086e-06,
521
+ "loss": 0.0038,
522
+ "step": 1575
523
+ },
524
+ {
525
+ "epoch": 6.28,
526
+ "learning_rate": 8.442211055276383e-06,
527
+ "loss": 0.0023,
528
+ "step": 1600
529
+ },
530
+ {
531
+ "epoch": 6.28,
532
+ "eval_loss": 0.6793932318687439,
533
+ "eval_runtime": 1564.2392,
534
+ "eval_samples_per_second": 5.527,
535
+ "eval_steps_per_second": 0.553,
536
+ "eval_wer": 44.250963271650875,
537
+ "step": 1600
538
+ },
539
+ {
540
+ "epoch": 6.38,
541
+ "learning_rate": 8.41708542713568e-06,
542
+ "loss": 0.0025,
543
+ "step": 1625
544
+ },
545
+ {
546
+ "epoch": 6.48,
547
+ "learning_rate": 8.391959798994976e-06,
548
+ "loss": 0.0027,
549
+ "step": 1650
550
+ },
551
+ {
552
+ "epoch": 6.58,
553
+ "learning_rate": 8.366834170854273e-06,
554
+ "loss": 0.0034,
555
+ "step": 1675
556
+ },
557
+ {
558
+ "epoch": 6.67,
559
+ "learning_rate": 8.341708542713568e-06,
560
+ "loss": 0.0032,
561
+ "step": 1700
562
+ },
563
+ {
564
+ "epoch": 6.67,
565
+ "eval_loss": 0.6724051237106323,
566
+ "eval_runtime": 1596.3328,
567
+ "eval_samples_per_second": 5.416,
568
+ "eval_steps_per_second": 0.542,
569
+ "eval_wer": 45.71612839212184,
570
+ "step": 1700
571
+ },
572
+ {
573
+ "epoch": 6.77,
574
+ "learning_rate": 8.316582914572864e-06,
575
+ "loss": 0.0036,
576
+ "step": 1725
577
+ },
578
+ {
579
+ "epoch": 6.87,
580
+ "learning_rate": 8.291457286432163e-06,
581
+ "loss": 0.0029,
582
+ "step": 1750
583
+ },
584
+ {
585
+ "epoch": 6.97,
586
+ "learning_rate": 8.266331658291457e-06,
587
+ "loss": 0.0026,
588
+ "step": 1775
589
+ },
590
+ {
591
+ "epoch": 7.07,
592
+ "learning_rate": 8.241206030150754e-06,
593
+ "loss": 0.0021,
594
+ "step": 1800
595
+ },
596
+ {
597
+ "epoch": 7.07,
598
+ "eval_loss": 0.6820480227470398,
599
+ "eval_runtime": 1590.9785,
600
+ "eval_samples_per_second": 5.434,
601
+ "eval_steps_per_second": 0.544,
602
+ "eval_wer": 47.60132956293099,
603
+ "step": 1800
604
+ },
605
+ {
606
+ "epoch": 7.16,
607
+ "learning_rate": 8.21608040201005e-06,
608
+ "loss": 0.0025,
609
+ "step": 1825
610
+ },
611
+ {
612
+ "epoch": 7.26,
613
+ "learning_rate": 8.190954773869347e-06,
614
+ "loss": 0.0017,
615
+ "step": 1850
616
+ },
617
+ {
618
+ "epoch": 7.36,
619
+ "learning_rate": 8.165829145728644e-06,
620
+ "loss": 0.0015,
621
+ "step": 1875
622
+ },
623
+ {
624
+ "epoch": 7.46,
625
+ "learning_rate": 8.14070351758794e-06,
626
+ "loss": 0.0015,
627
+ "step": 1900
628
+ },
629
+ {
630
+ "epoch": 7.46,
631
+ "eval_loss": 0.6925253868103027,
632
+ "eval_runtime": 1624.9343,
633
+ "eval_samples_per_second": 5.321,
634
+ "eval_steps_per_second": 0.532,
635
+ "eval_wer": 46.67195845942683,
636
+ "step": 1900
637
+ },
638
+ {
639
+ "epoch": 7.56,
640
+ "learning_rate": 8.115577889447237e-06,
641
+ "loss": 0.0023,
642
+ "step": 1925
643
+ },
644
+ {
645
+ "epoch": 7.65,
646
+ "learning_rate": 8.090452261306532e-06,
647
+ "loss": 0.0014,
648
+ "step": 1950
649
+ },
650
+ {
651
+ "epoch": 7.75,
652
+ "learning_rate": 8.06532663316583e-06,
653
+ "loss": 0.002,
654
+ "step": 1975
655
+ },
656
+ {
657
+ "epoch": 7.85,
658
+ "learning_rate": 8.040201005025127e-06,
659
+ "loss": 0.0024,
660
+ "step": 2000
661
+ },
662
+ {
663
+ "epoch": 7.85,
664
+ "eval_loss": 0.710422158241272,
665
+ "eval_runtime": 1663.5693,
666
+ "eval_samples_per_second": 5.197,
667
+ "eval_steps_per_second": 0.52,
668
+ "eval_wer": 50.29022175919036,
669
+ "step": 2000
670
+ },
671
+ {
672
+ "epoch": 7.95,
673
+ "learning_rate": 8.015075376884422e-06,
674
+ "loss": 0.0027,
675
+ "step": 2025
676
+ },
677
+ {
678
+ "epoch": 8.05,
679
+ "learning_rate": 7.989949748743719e-06,
680
+ "loss": 0.0026,
681
+ "step": 2050
682
+ },
683
+ {
684
+ "epoch": 8.15,
685
+ "learning_rate": 7.965829145728644e-06,
686
+ "loss": 0.0029,
687
+ "step": 2075
688
+ },
689
+ {
690
+ "epoch": 8.24,
691
+ "learning_rate": 7.940703517587939e-06,
692
+ "loss": 0.0029,
693
+ "step": 2100
694
+ },
695
+ {
696
+ "epoch": 8.24,
697
+ "eval_loss": 0.6837413311004639,
698
+ "eval_runtime": 1582.0652,
699
+ "eval_samples_per_second": 5.465,
700
+ "eval_steps_per_second": 0.547,
701
+ "eval_wer": 46.41729093284384,
702
+ "step": 2100
703
+ },
704
+ {
705
+ "epoch": 8.34,
706
+ "learning_rate": 7.915577889447237e-06,
707
+ "loss": 0.0017,
708
+ "step": 2125
709
+ },
710
+ {
711
+ "epoch": 8.44,
712
+ "learning_rate": 7.890452261306534e-06,
713
+ "loss": 0.0015,
714
+ "step": 2150
715
+ },
716
+ {
717
+ "epoch": 8.54,
718
+ "learning_rate": 7.865326633165829e-06,
719
+ "loss": 0.0017,
720
+ "step": 2175
721
+ },
722
+ {
723
+ "epoch": 8.64,
724
+ "learning_rate": 7.840201005025127e-06,
725
+ "loss": 0.0016,
726
+ "step": 2200
727
+ },
728
+ {
729
+ "epoch": 8.64,
730
+ "eval_loss": 0.7190621495246887,
731
+ "eval_runtime": 1555.4401,
732
+ "eval_samples_per_second": 5.559,
733
+ "eval_steps_per_second": 0.556,
734
+ "eval_wer": 46.008830679168526,
735
+ "step": 2200
736
+ },
737
+ {
738
+ "epoch": 8.73,
739
+ "learning_rate": 7.815075376884422e-06,
740
+ "loss": 0.0013,
741
+ "step": 2225
742
+ },
743
+ {
744
+ "epoch": 8.83,
745
+ "learning_rate": 7.789949748743719e-06,
746
+ "loss": 0.0017,
747
+ "step": 2250
748
+ },
749
+ {
750
+ "epoch": 8.93,
751
+ "learning_rate": 7.764824120603015e-06,
752
+ "loss": 0.0018,
753
+ "step": 2275
754
+ },
755
+ {
756
+ "epoch": 9.03,
757
+ "learning_rate": 7.739698492462312e-06,
758
+ "loss": 0.0017,
759
+ "step": 2300
760
+ },
761
+ {
762
+ "epoch": 9.03,
763
+ "eval_loss": 0.7105358242988586,
764
+ "eval_runtime": 1601.1616,
765
+ "eval_samples_per_second": 5.4,
766
+ "eval_steps_per_second": 0.54,
767
+ "eval_wer": 47.59636850721834,
768
+ "step": 2300
769
+ },
770
+ {
771
+ "epoch": 9.13,
772
+ "learning_rate": 7.714572864321609e-06,
773
+ "loss": 0.0016,
774
+ "step": 2325
775
+ },
776
+ {
777
+ "epoch": 9.22,
778
+ "learning_rate": 7.689447236180905e-06,
779
+ "loss": 0.0015,
780
+ "step": 2350
781
+ },
782
+ {
783
+ "epoch": 9.32,
784
+ "learning_rate": 7.664321608040202e-06,
785
+ "loss": 0.0013,
786
+ "step": 2375
787
+ },
788
+ {
789
+ "epoch": 9.42,
790
+ "learning_rate": 7.639195979899499e-06,
791
+ "loss": 0.0014,
792
+ "step": 2400
793
+ },
794
+ {
795
+ "epoch": 9.42,
796
+ "eval_loss": 0.7293367981910706,
797
+ "eval_runtime": 1636.4789,
798
+ "eval_samples_per_second": 5.283,
799
+ "eval_steps_per_second": 0.529,
800
+ "eval_wer": 44.76029832481686,
801
+ "step": 2400
802
+ },
803
+ {
804
+ "epoch": 9.52,
805
+ "learning_rate": 7.614070351758794e-06,
806
+ "loss": 0.0015,
807
+ "step": 2425
808
+ },
809
+ {
810
+ "epoch": 9.62,
811
+ "learning_rate": 7.588944723618091e-06,
812
+ "loss": 0.0014,
813
+ "step": 2450
814
+ },
815
+ {
816
+ "epoch": 9.72,
817
+ "learning_rate": 7.563819095477387e-06,
818
+ "loss": 0.0012,
819
+ "step": 2475
820
+ },
821
+ {
822
+ "epoch": 9.81,
823
+ "learning_rate": 7.538693467336684e-06,
824
+ "loss": 0.0018,
825
+ "step": 2500
826
+ },
827
+ {
828
+ "epoch": 9.81,
829
+ "eval_loss": 0.7364875078201294,
830
+ "eval_runtime": 1682.5135,
831
+ "eval_samples_per_second": 5.139,
832
+ "eval_steps_per_second": 0.514,
833
+ "eval_wer": 49.896644672653004,
834
+ "step": 2500
835
+ },
836
+ {
837
+ "epoch": 9.91,
838
+ "learning_rate": 7.513567839195981e-06,
839
+ "loss": 0.0015,
840
+ "step": 2525
841
+ },
842
+ {
843
+ "epoch": 10.01,
844
+ "learning_rate": 7.488442211055277e-06,
845
+ "loss": 0.0016,
846
+ "step": 2550
847
+ },
848
+ {
849
+ "epoch": 10.11,
850
+ "learning_rate": 7.463316582914574e-06,
851
+ "loss": 0.0008,
852
+ "step": 2575
853
+ },
854
+ {
855
+ "epoch": 10.21,
856
+ "learning_rate": 7.43819095477387e-06,
857
+ "loss": 0.0008,
858
+ "step": 2600
859
+ },
860
+ {
861
+ "epoch": 10.21,
862
+ "eval_loss": 0.7378189563751221,
863
+ "eval_runtime": 1631.5066,
864
+ "eval_samples_per_second": 5.299,
865
+ "eval_steps_per_second": 0.53,
866
+ "eval_wer": 47.473995799639496,
867
+ "step": 2600
868
+ },
869
+ {
870
+ "epoch": 10.3,
871
+ "learning_rate": 7.4130653266331665e-06,
872
+ "loss": 0.0012,
873
+ "step": 2625
874
+ },
875
+ {
876
+ "epoch": 10.4,
877
+ "learning_rate": 7.387939698492463e-06,
878
+ "loss": 0.001,
879
+ "step": 2650
880
+ },
881
+ {
882
+ "epoch": 10.5,
883
+ "learning_rate": 7.362814070351759e-06,
884
+ "loss": 0.0024,
885
+ "step": 2675
886
+ },
887
+ {
888
+ "epoch": 10.6,
889
+ "learning_rate": 7.337688442211056e-06,
890
+ "loss": 0.0016,
891
+ "step": 2700
892
+ },
893
+ {
894
+ "epoch": 10.6,
895
+ "eval_loss": 0.7303053140640259,
896
+ "eval_runtime": 1617.5646,
897
+ "eval_samples_per_second": 5.345,
898
+ "eval_steps_per_second": 0.535,
899
+ "eval_wer": 45.96914223346728,
900
+ "step": 2700
901
+ },
902
+ {
903
+ "epoch": 10.7,
904
+ "learning_rate": 7.312562814070352e-06,
905
+ "loss": 0.0015,
906
+ "step": 2725
907
+ },
908
+ {
909
+ "epoch": 10.79,
910
+ "learning_rate": 7.287437185929649e-06,
911
+ "loss": 0.0013,
912
+ "step": 2750
913
+ },
914
+ {
915
+ "epoch": 10.89,
916
+ "learning_rate": 7.262311557788946e-06,
917
+ "loss": 0.0021,
918
+ "step": 2775
919
+ },
920
+ {
921
+ "epoch": 10.99,
922
+ "learning_rate": 7.237185929648242e-06,
923
+ "loss": 0.0011,
924
+ "step": 2800
925
+ },
926
+ {
927
+ "epoch": 10.99,
928
+ "eval_loss": 0.7329912781715393,
929
+ "eval_runtime": 1626.8373,
930
+ "eval_samples_per_second": 5.315,
931
+ "eval_steps_per_second": 0.532,
932
+ "eval_wer": 47.72535595574738,
933
+ "step": 2800
934
+ },
935
+ {
936
+ "epoch": 11.09,
937
+ "learning_rate": 7.212060301507539e-06,
938
+ "loss": 0.0017,
939
+ "step": 2825
940
+ },
941
+ {
942
+ "epoch": 11.19,
943
+ "learning_rate": 7.186934673366834e-06,
944
+ "loss": 0.0013,
945
+ "step": 2850
946
+ },
947
+ {
948
+ "epoch": 11.29,
949
+ "learning_rate": 7.161809045226131e-06,
950
+ "loss": 0.0013,
951
+ "step": 2875
952
+ },
953
+ {
954
+ "epoch": 11.38,
955
+ "learning_rate": 7.1366834170854284e-06,
956
+ "loss": 0.0014,
957
+ "step": 2900
958
+ },
959
+ {
960
+ "epoch": 11.38,
961
+ "eval_loss": 0.7448384761810303,
962
+ "eval_runtime": 1554.4308,
963
+ "eval_samples_per_second": 5.562,
964
+ "eval_steps_per_second": 0.556,
965
+ "eval_wer": 44.857865753832414,
966
+ "step": 2900
967
+ },
968
+ {
969
+ "epoch": 11.48,
970
+ "learning_rate": 7.111557788944724e-06,
971
+ "loss": 0.0014,
972
+ "step": 2925
973
+ },
974
+ {
975
+ "epoch": 11.58,
976
+ "learning_rate": 7.086432160804021e-06,
977
+ "loss": 0.0012,
978
+ "step": 2950
979
+ },
980
+ {
981
+ "epoch": 11.68,
982
+ "learning_rate": 7.061306532663317e-06,
983
+ "loss": 0.0012,
984
+ "step": 2975
985
+ },
986
+ {
987
+ "epoch": 11.78,
988
+ "learning_rate": 7.036180904522614e-06,
989
+ "loss": 0.0013,
990
+ "step": 3000
991
+ },
992
+ {
993
+ "epoch": 11.78,
994
+ "eval_loss": 0.7471309900283813,
995
+ "eval_runtime": 1597.219,
996
+ "eval_samples_per_second": 5.413,
997
+ "eval_steps_per_second": 0.542,
998
+ "eval_wer": 46.5115509913843,
999
+ "step": 3000
1000
+ },
1001
+ {
1002
+ "epoch": 11.87,
1003
+ "learning_rate": 7.011055276381911e-06,
1004
+ "loss": 0.0021,
1005
+ "step": 3025
1006
+ },
1007
+ {
1008
+ "epoch": 11.97,
1009
+ "learning_rate": 6.9859296482412065e-06,
1010
+ "loss": 0.0021,
1011
+ "step": 3050
1012
+ },
1013
+ {
1014
+ "epoch": 12.07,
1015
+ "learning_rate": 6.960804020100503e-06,
1016
+ "loss": 0.0013,
1017
+ "step": 3075
1018
+ },
1019
+ {
1020
+ "epoch": 12.17,
1021
+ "learning_rate": 6.935678391959799e-06,
1022
+ "loss": 0.0015,
1023
+ "step": 3100
1024
+ },
1025
+ {
1026
+ "epoch": 12.17,
1027
+ "eval_loss": 0.7513495683670044,
1028
+ "eval_runtime": 1592.1737,
1029
+ "eval_samples_per_second": 5.43,
1030
+ "eval_steps_per_second": 0.543,
1031
+ "eval_wer": 47.5699095434175,
1032
+ "step": 3100
1033
+ },
1034
+ {
1035
+ "epoch": 12.27,
1036
+ "learning_rate": 6.910552763819096e-06,
1037
+ "loss": 0.0008,
1038
+ "step": 3125
1039
+ },
1040
+ {
1041
+ "epoch": 12.37,
1042
+ "learning_rate": 6.885427135678393e-06,
1043
+ "loss": 0.001,
1044
+ "step": 3150
1045
+ },
1046
+ {
1047
+ "epoch": 12.46,
1048
+ "learning_rate": 6.860301507537689e-06,
1049
+ "loss": 0.0012,
1050
+ "step": 3175
1051
+ },
1052
+ {
1053
+ "epoch": 12.56,
1054
+ "learning_rate": 6.835175879396986e-06,
1055
+ "loss": 0.0014,
1056
+ "step": 3200
1057
+ },
1058
+ {
1059
+ "epoch": 12.56,
1060
+ "eval_loss": 0.7380189299583435,
1061
+ "eval_runtime": 1594.7949,
1062
+ "eval_samples_per_second": 5.421,
1063
+ "eval_steps_per_second": 0.542,
1064
+ "eval_wer": 46.400754080468325,
1065
+ "step": 3200
1066
+ },
1067
+ {
1068
+ "epoch": 12.66,
1069
+ "learning_rate": 6.810050251256282e-06,
1070
+ "loss": 0.0013,
1071
+ "step": 3225
1072
+ },
1073
+ {
1074
+ "epoch": 12.76,
1075
+ "learning_rate": 6.784924623115579e-06,
1076
+ "loss": 0.0014,
1077
+ "step": 3250
1078
+ },
1079
+ {
1080
+ "epoch": 12.86,
1081
+ "learning_rate": 6.759798994974875e-06,
1082
+ "loss": 0.0018,
1083
+ "step": 3275
1084
+ },
1085
+ {
1086
+ "epoch": 12.95,
1087
+ "learning_rate": 6.734673366834171e-06,
1088
+ "loss": 0.0015,
1089
+ "step": 3300
1090
+ },
1091
+ {
1092
+ "epoch": 12.95,
1093
+ "eval_loss": 0.7520471811294556,
1094
+ "eval_runtime": 1574.6945,
1095
+ "eval_samples_per_second": 5.491,
1096
+ "eval_steps_per_second": 0.549,
1097
+ "eval_wer": 45.965834862992175,
1098
+ "step": 3300
1099
+ },
1100
+ {
1101
+ "epoch": 13.05,
1102
+ "learning_rate": 6.7095477386934685e-06,
1103
+ "loss": 0.0026,
1104
+ "step": 3325
1105
+ },
1106
+ {
1107
+ "epoch": 13.15,
1108
+ "learning_rate": 6.684422110552764e-06,
1109
+ "loss": 0.0009,
1110
+ "step": 3350
1111
+ },
1112
+ {
1113
+ "epoch": 13.25,
1114
+ "learning_rate": 6.659296482412061e-06,
1115
+ "loss": 0.0009,
1116
+ "step": 3375
1117
+ },
1118
+ {
1119
+ "epoch": 13.35,
1120
+ "learning_rate": 6.6341708542713575e-06,
1121
+ "loss": 0.0009,
1122
+ "step": 3400
1123
+ },
1124
+ {
1125
+ "epoch": 13.35,
1126
+ "eval_loss": 0.7481613159179688,
1127
+ "eval_runtime": 1618.0061,
1128
+ "eval_samples_per_second": 5.344,
1129
+ "eval_steps_per_second": 0.535,
1130
+ "eval_wer": 49.2269021514445,
1131
+ "step": 3400
1132
+ },
1133
+ {
1134
+ "epoch": 13.44,
1135
+ "learning_rate": 6.609045226130654e-06,
1136
+ "loss": 0.0009,
1137
+ "step": 3425
1138
+ },
1139
+ {
1140
+ "epoch": 13.54,
1141
+ "learning_rate": 6.583919597989951e-06,
1142
+ "loss": 0.0019,
1143
+ "step": 3450
1144
+ },
1145
+ {
1146
+ "epoch": 13.64,
1147
+ "learning_rate": 6.5587939698492465e-06,
1148
+ "loss": 0.0017,
1149
+ "step": 3475
1150
+ },
1151
+ {
1152
+ "epoch": 13.74,
1153
+ "learning_rate": 6.533668341708543e-06,
1154
+ "loss": 0.0021,
1155
+ "step": 3500
1156
+ },
1157
+ {
1158
+ "epoch": 13.74,
1159
+ "eval_loss": 0.7618550658226013,
1160
+ "eval_runtime": 1576.9068,
1161
+ "eval_samples_per_second": 5.483,
1162
+ "eval_steps_per_second": 0.549,
1163
+ "eval_wer": 47.1234145292785,
1164
+ "step": 3500
1165
+ },
1166
+ {
1167
+ "epoch": 13.84,
1168
+ "learning_rate": 6.508542713567839e-06,
1169
+ "loss": 0.003,
1170
+ "step": 3525
1171
+ },
1172
+ {
1173
+ "epoch": 13.94,
1174
+ "learning_rate": 6.483417085427136e-06,
1175
+ "loss": 0.0017,
1176
+ "step": 3550
1177
+ },
1178
+ {
1179
+ "epoch": 14.03,
1180
+ "learning_rate": 6.458291457286433e-06,
1181
+ "loss": 0.0016,
1182
+ "step": 3575
1183
+ },
1184
+ {
1185
+ "epoch": 14.13,
1186
+ "learning_rate": 6.433165829145729e-06,
1187
+ "loss": 0.0013,
1188
+ "step": 3600
1189
+ },
1190
+ {
1191
+ "epoch": 14.13,
1192
+ "eval_loss": 0.7452961802482605,
1193
+ "eval_runtime": 1635.9285,
1194
+ "eval_samples_per_second": 5.285,
1195
+ "eval_steps_per_second": 0.529,
1196
+ "eval_wer": 49.39557804567479,
1197
+ "step": 3600
1198
+ },
1199
+ {
1200
+ "epoch": 14.23,
1201
+ "learning_rate": 6.408040201005025e-06,
1202
+ "loss": 0.0011,
1203
+ "step": 3625
1204
+ },
1205
+ {
1206
+ "epoch": 14.33,
1207
+ "learning_rate": 6.382914572864322e-06,
1208
+ "loss": 0.0009,
1209
+ "step": 3650
1210
+ },
1211
+ {
1212
+ "epoch": 14.43,
1213
+ "learning_rate": 6.357788944723619e-06,
1214
+ "loss": 0.0017,
1215
+ "step": 3675
1216
+ },
1217
+ {
1218
+ "epoch": 14.52,
1219
+ "learning_rate": 6.332663316582915e-06,
1220
+ "loss": 0.001,
1221
+ "step": 3700
1222
+ },
1223
+ {
1224
+ "epoch": 14.52,
1225
+ "eval_loss": 0.7582494020462036,
1226
+ "eval_runtime": 1594.5778,
1227
+ "eval_samples_per_second": 5.422,
1228
+ "eval_steps_per_second": 0.542,
1229
+ "eval_wer": 47.67409171338328,
1230
+ "step": 3700
1231
+ },
1232
+ {
1233
+ "epoch": 14.62,
1234
+ "learning_rate": 6.307537688442211e-06,
1235
+ "loss": 0.0012,
1236
+ "step": 3725
1237
+ },
1238
+ {
1239
+ "epoch": 14.72,
1240
+ "learning_rate": 6.2824120603015085e-06,
1241
+ "loss": 0.0014,
1242
+ "step": 3750
1243
+ },
1244
+ {
1245
+ "epoch": 14.82,
1246
+ "learning_rate": 6.257286432160804e-06,
1247
+ "loss": 0.001,
1248
+ "step": 3775
1249
+ },
1250
+ {
1251
+ "epoch": 14.92,
1252
+ "learning_rate": 6.232160804020101e-06,
1253
+ "loss": 0.0009,
1254
+ "step": 3800
1255
+ },
1256
+ {
1257
+ "epoch": 14.92,
1258
+ "eval_loss": 0.7637260556221008,
1259
+ "eval_runtime": 1592.8989,
1260
+ "eval_samples_per_second": 5.428,
1261
+ "eval_steps_per_second": 0.543,
1262
+ "eval_wer": 46.971275487423725,
1263
+ "step": 3800
1264
+ },
1265
+ {
1266
+ "epoch": 15.01,
1267
+ "learning_rate": 6.2070351758793975e-06,
1268
+ "loss": 0.0008,
1269
+ "step": 3825
1270
+ },
1271
+ {
1272
+ "epoch": 15.11,
1273
+ "learning_rate": 6.181909547738693e-06,
1274
+ "loss": 0.0007,
1275
+ "step": 3850
1276
+ },
1277
+ {
1278
+ "epoch": 15.21,
1279
+ "learning_rate": 6.156783919597991e-06,
1280
+ "loss": 0.0003,
1281
+ "step": 3875
1282
+ },
1283
+ {
1284
+ "epoch": 15.31,
1285
+ "learning_rate": 6.1316582914572865e-06,
1286
+ "loss": 0.0014,
1287
+ "step": 3900
1288
+ },
1289
+ {
1290
+ "epoch": 15.31,
1291
+ "eval_loss": 0.7722346782684326,
1292
+ "eval_runtime": 1581.3202,
1293
+ "eval_samples_per_second": 5.468,
1294
+ "eval_steps_per_second": 0.547,
1295
+ "eval_wer": 47.27059251542062,
1296
+ "step": 3900
1297
+ },
1298
+ {
1299
+ "epoch": 15.41,
1300
+ "learning_rate": 6.106532663316583e-06,
1301
+ "loss": 0.0007,
1302
+ "step": 3925
1303
+ },
1304
+ {
1305
+ "epoch": 15.51,
1306
+ "learning_rate": 6.081407035175881e-06,
1307
+ "loss": 0.0015,
1308
+ "step": 3950
1309
+ },
1310
+ {
1311
+ "epoch": 15.6,
1312
+ "learning_rate": 6.056281407035176e-06,
1313
+ "loss": 0.0012,
1314
+ "step": 3975
1315
+ },
1316
+ {
1317
+ "epoch": 15.7,
1318
+ "learning_rate": 6.031155778894473e-06,
1319
+ "loss": 0.001,
1320
+ "step": 4000
1321
+ },
1322
+ {
1323
+ "epoch": 15.7,
1324
+ "eval_loss": 0.7691770195960999,
1325
+ "eval_runtime": 1608.3109,
1326
+ "eval_samples_per_second": 5.376,
1327
+ "eval_steps_per_second": 0.538,
1328
+ "eval_wer": 46.96631443171107,
1329
+ "step": 4000
1330
+ },
1331
+ {
1332
+ "epoch": 15.8,
1333
+ "learning_rate": 6.006030150753769e-06,
1334
+ "loss": 0.0006,
1335
+ "step": 4025
1336
+ },
1337
+ {
1338
+ "epoch": 15.9,
1339
+ "learning_rate": 5.980904522613065e-06,
1340
+ "loss": 0.0007,
1341
+ "step": 4050
1342
+ },
1343
+ {
1344
+ "epoch": 16.0,
1345
+ "learning_rate": 5.955778894472363e-06,
1346
+ "loss": 0.0008,
1347
+ "step": 4075
1348
+ },
1349
+ {
1350
+ "epoch": 16.09,
1351
+ "learning_rate": 5.930653266331659e-06,
1352
+ "loss": 0.0003,
1353
+ "step": 4100
1354
+ },
1355
+ {
1356
+ "epoch": 16.09,
1357
+ "eval_loss": 0.774411678314209,
1358
+ "eval_runtime": 1594.2731,
1359
+ "eval_samples_per_second": 5.423,
1360
+ "eval_steps_per_second": 0.543,
1361
+ "eval_wer": 47.173025086405055,
1362
+ "step": 4100
1363
+ },
1364
+ {
1365
+ "epoch": 16.19,
1366
+ "learning_rate": 5.905527638190955e-06,
1367
+ "loss": 0.0004,
1368
+ "step": 4125
1369
+ },
1370
+ {
1371
+ "epoch": 16.29,
1372
+ "learning_rate": 5.880402010050251e-06,
1373
+ "loss": 0.0003,
1374
+ "step": 4150
1375
+ },
1376
+ {
1377
+ "epoch": 16.39,
1378
+ "learning_rate": 5.8552763819095485e-06,
1379
+ "loss": 0.0015,
1380
+ "step": 4175
1381
+ },
1382
+ {
1383
+ "epoch": 16.49,
1384
+ "learning_rate": 5.830150753768845e-06,
1385
+ "loss": 0.0004,
1386
+ "step": 4200
1387
+ },
1388
+ {
1389
+ "epoch": 16.49,
1390
+ "eval_loss": 0.7842475175857544,
1391
+ "eval_runtime": 1587.8752,
1392
+ "eval_samples_per_second": 5.445,
1393
+ "eval_steps_per_second": 0.545,
1394
+ "eval_wer": 47.33508623968514,
1395
+ "step": 4200
1396
+ },
1397
+ {
1398
+ "epoch": 16.58,
1399
+ "learning_rate": 5.805025125628141e-06,
1400
+ "loss": 0.0005,
1401
+ "step": 4225
1402
+ },
1403
+ {
1404
+ "epoch": 16.68,
1405
+ "learning_rate": 5.780904522613066e-06,
1406
+ "loss": 0.0008,
1407
+ "step": 4250
1408
+ },
1409
+ {
1410
+ "epoch": 16.78,
1411
+ "learning_rate": 5.755778894472362e-06,
1412
+ "loss": 0.0003,
1413
+ "step": 4275
1414
+ },
1415
+ {
1416
+ "epoch": 16.88,
1417
+ "learning_rate": 5.730653266331658e-06,
1418
+ "loss": 0.0003,
1419
+ "step": 4300
1420
+ },
1421
+ {
1422
+ "epoch": 16.88,
1423
+ "eval_loss": 0.7783875465393066,
1424
+ "eval_runtime": 1573.9796,
1425
+ "eval_samples_per_second": 5.493,
1426
+ "eval_steps_per_second": 0.55,
1427
+ "eval_wer": 47.07711134262704,
1428
+ "step": 4300
1429
+ },
1430
+ {
1431
+ "epoch": 16.98,
1432
+ "learning_rate": 5.7055276381909555e-06,
1433
+ "loss": 0.0005,
1434
+ "step": 4325
1435
+ },
1436
+ {
1437
+ "epoch": 17.08,
1438
+ "learning_rate": 5.680402010050252e-06,
1439
+ "loss": 0.0008,
1440
+ "step": 4350
1441
+ },
1442
+ {
1443
+ "epoch": 17.17,
1444
+ "learning_rate": 5.655276381909548e-06,
1445
+ "loss": 0.0005,
1446
+ "step": 4375
1447
+ },
1448
+ {
1449
+ "epoch": 17.27,
1450
+ "learning_rate": 5.6301507537688445e-06,
1451
+ "loss": 0.0002,
1452
+ "step": 4400
1453
+ },
1454
+ {
1455
+ "epoch": 17.27,
1456
+ "eval_loss": 0.7879212498664856,
1457
+ "eval_runtime": 1564.3111,
1458
+ "eval_samples_per_second": 5.527,
1459
+ "eval_steps_per_second": 0.553,
1460
+ "eval_wer": 45.76408526401085,
1461
+ "step": 4400
1462
+ },
1463
+ {
1464
+ "epoch": 17.37,
1465
+ "learning_rate": 5.605025125628141e-06,
1466
+ "loss": 0.0004,
1467
+ "step": 4425
1468
+ },
1469
+ {
1470
+ "epoch": 17.47,
1471
+ "learning_rate": 5.579899497487438e-06,
1472
+ "loss": 0.0007,
1473
+ "step": 4450
1474
+ },
1475
+ {
1476
+ "epoch": 17.57,
1477
+ "learning_rate": 5.554773869346734e-06,
1478
+ "loss": 0.0005,
1479
+ "step": 4475
1480
+ },
1481
+ {
1482
+ "epoch": 17.66,
1483
+ "learning_rate": 5.52964824120603e-06,
1484
+ "loss": 0.0005,
1485
+ "step": 4500
1486
+ },
1487
+ {
1488
+ "epoch": 17.66,
1489
+ "eval_loss": 0.7964885830879211,
1490
+ "eval_runtime": 1592.8224,
1491
+ "eval_samples_per_second": 5.428,
1492
+ "eval_steps_per_second": 0.543,
1493
+ "eval_wer": 50.02397843594451,
1494
+ "step": 4500
1495
+ },
1496
+ {
1497
+ "epoch": 17.76,
1498
+ "learning_rate": 5.504522613065328e-06,
1499
+ "loss": 0.0005,
1500
+ "step": 4525
1501
+ },
1502
+ {
1503
+ "epoch": 17.86,
1504
+ "learning_rate": 5.479396984924623e-06,
1505
+ "loss": 0.0005,
1506
+ "step": 4550
1507
+ },
1508
+ {
1509
+ "epoch": 17.96,
1510
+ "learning_rate": 5.45427135678392e-06,
1511
+ "loss": 0.0009,
1512
+ "step": 4575
1513
+ },
1514
+ {
1515
+ "epoch": 18.06,
1516
+ "learning_rate": 5.429145728643217e-06,
1517
+ "loss": 0.0004,
1518
+ "step": 4600
1519
+ },
1520
+ {
1521
+ "epoch": 18.06,
1522
+ "eval_loss": 0.8001408576965332,
1523
+ "eval_runtime": 1577.1711,
1524
+ "eval_samples_per_second": 5.482,
1525
+ "eval_steps_per_second": 0.548,
1526
+ "eval_wer": 48.438094293132245,
1527
+ "step": 4600
1528
+ },
1529
+ {
1530
+ "epoch": 18.16,
1531
+ "learning_rate": 5.404020100502513e-06,
1532
+ "loss": 0.0007,
1533
+ "step": 4625
1534
+ },
1535
+ {
1536
+ "epoch": 18.25,
1537
+ "learning_rate": 5.37889447236181e-06,
1538
+ "loss": 0.0004,
1539
+ "step": 4650
1540
+ },
1541
+ {
1542
+ "epoch": 18.35,
1543
+ "learning_rate": 5.353768844221106e-06,
1544
+ "loss": 0.0002,
1545
+ "step": 4675
1546
+ },
1547
+ {
1548
+ "epoch": 18.45,
1549
+ "learning_rate": 5.328643216080402e-06,
1550
+ "loss": 0.0002,
1551
+ "step": 4700
1552
+ },
1553
+ {
1554
+ "epoch": 18.45,
1555
+ "eval_loss": 0.8016232252120972,
1556
+ "eval_runtime": 1606.7697,
1557
+ "eval_samples_per_second": 5.381,
1558
+ "eval_steps_per_second": 0.538,
1559
+ "eval_wer": 49.00365464437499,
1560
+ "step": 4700
1561
+ },
1562
+ {
1563
+ "epoch": 18.55,
1564
+ "learning_rate": 5.3035175879397e-06,
1565
+ "loss": 0.0002,
1566
+ "step": 4725
1567
+ },
1568
+ {
1569
+ "epoch": 18.65,
1570
+ "learning_rate": 5.2783919597989955e-06,
1571
+ "loss": 0.0004,
1572
+ "step": 4750
1573
+ },
1574
+ {
1575
+ "epoch": 18.74,
1576
+ "learning_rate": 5.253266331658292e-06,
1577
+ "loss": 0.0002,
1578
+ "step": 4775
1579
+ },
1580
+ {
1581
+ "epoch": 18.84,
1582
+ "learning_rate": 5.228140703517588e-06,
1583
+ "loss": 0.0002,
1584
+ "step": 4800
1585
+ },
1586
+ {
1587
+ "epoch": 18.84,
1588
+ "eval_loss": 0.8065798878669739,
1589
+ "eval_runtime": 1608.099,
1590
+ "eval_samples_per_second": 5.377,
1591
+ "eval_steps_per_second": 0.538,
1592
+ "eval_wer": 50.08681847497147,
1593
+ "step": 4800
1594
+ },
1595
+ {
1596
+ "epoch": 18.94,
1597
+ "learning_rate": 5.2030150753768845e-06,
1598
+ "loss": 0.0006,
1599
+ "step": 4825
1600
+ },
1601
+ {
1602
+ "epoch": 19.04,
1603
+ "learning_rate": 5.177889447236182e-06,
1604
+ "loss": 0.0002,
1605
+ "step": 4850
1606
+ },
1607
+ {
1608
+ "epoch": 19.14,
1609
+ "learning_rate": 5.152763819095478e-06,
1610
+ "loss": 0.0005,
1611
+ "step": 4875
1612
+ },
1613
+ {
1614
+ "epoch": 19.23,
1615
+ "learning_rate": 5.127638190954774e-06,
1616
+ "loss": 0.0009,
1617
+ "step": 4900
1618
+ },
1619
+ {
1620
+ "epoch": 19.23,
1621
+ "eval_loss": 0.8020856380462646,
1622
+ "eval_runtime": 1561.8925,
1623
+ "eval_samples_per_second": 5.536,
1624
+ "eval_steps_per_second": 0.554,
1625
+ "eval_wer": 47.227596699244266,
1626
+ "step": 4900
1627
+ },
1628
+ {
1629
+ "epoch": 19.33,
1630
+ "learning_rate": 5.10251256281407e-06,
1631
+ "loss": 0.0005,
1632
+ "step": 4925
1633
+ },
1634
+ {
1635
+ "epoch": 19.43,
1636
+ "learning_rate": 5.077386934673368e-06,
1637
+ "loss": 0.0004,
1638
+ "step": 4950
1639
+ },
1640
+ {
1641
+ "epoch": 19.53,
1642
+ "learning_rate": 5.052261306532664e-06,
1643
+ "loss": 0.0007,
1644
+ "step": 4975
1645
+ },
1646
+ {
1647
+ "epoch": 19.63,
1648
+ "learning_rate": 5.02713567839196e-06,
1649
+ "loss": 0.0005,
1650
+ "step": 5000
1651
+ },
1652
+ {
1653
+ "epoch": 19.63,
1654
+ "eval_loss": 0.8162133693695068,
1655
+ "eval_runtime": 1588.3658,
1656
+ "eval_samples_per_second": 5.443,
1657
+ "eval_steps_per_second": 0.545,
1658
+ "eval_wer": 47.34996940682311,
1659
+ "step": 5000
1660
+ },
1661
+ {
1662
+ "epoch": 19.73,
1663
+ "learning_rate": 5.002010050251257e-06,
1664
+ "loss": 0.0003,
1665
+ "step": 5025
1666
+ },
1667
+ {
1668
+ "epoch": 19.82,
1669
+ "learning_rate": 4.976884422110553e-06,
1670
+ "loss": 0.0004,
1671
+ "step": 5050
1672
+ },
1673
+ {
1674
+ "epoch": 19.92,
1675
+ "learning_rate": 4.95175879396985e-06,
1676
+ "loss": 0.0004,
1677
+ "step": 5075
1678
+ },
1679
+ {
1680
+ "epoch": 20.02,
1681
+ "learning_rate": 4.9266331658291465e-06,
1682
+ "loss": 0.0003,
1683
+ "step": 5100
1684
+ },
1685
+ {
1686
+ "epoch": 20.02,
1687
+ "eval_loss": 0.8091246485710144,
1688
+ "eval_runtime": 1607.5038,
1689
+ "eval_samples_per_second": 5.379,
1690
+ "eval_steps_per_second": 0.538,
1691
+ "eval_wer": 48.72252815399117,
1692
+ "step": 5100
1693
+ },
1694
+ {
1695
+ "epoch": 20.12,
1696
+ "learning_rate": 4.901507537688442e-06,
1697
+ "loss": 0.0003,
1698
+ "step": 5125
1699
+ },
1700
+ {
1701
+ "epoch": 20.22,
1702
+ "learning_rate": 4.876381909547739e-06,
1703
+ "loss": 0.0003,
1704
+ "step": 5150
1705
+ },
1706
+ {
1707
+ "epoch": 20.31,
1708
+ "learning_rate": 4.8512562814070355e-06,
1709
+ "loss": 0.0003,
1710
+ "step": 5175
1711
+ },
1712
+ {
1713
+ "epoch": 20.41,
1714
+ "learning_rate": 4.826130653266332e-06,
1715
+ "loss": 0.0003,
1716
+ "step": 5200
1717
+ },
1718
+ {
1719
+ "epoch": 20.41,
1720
+ "eval_loss": 0.8060252666473389,
1721
+ "eval_runtime": 1624.2869,
1722
+ "eval_samples_per_second": 5.323,
1723
+ "eval_steps_per_second": 0.533,
1724
+ "eval_wer": 51.50237303831589,
1725
+ "step": 5200
1726
+ },
1727
+ {
1728
+ "epoch": 20.51,
1729
+ "learning_rate": 4.801005025125629e-06,
1730
+ "loss": 0.0001,
1731
+ "step": 5225
1732
+ },
1733
+ {
1734
+ "epoch": 20.61,
1735
+ "learning_rate": 4.7758793969849245e-06,
1736
+ "loss": 0.0001,
1737
+ "step": 5250
1738
+ },
1739
+ {
1740
+ "epoch": 20.71,
1741
+ "learning_rate": 4.750753768844221e-06,
1742
+ "loss": 0.0003,
1743
+ "step": 5275
1744
+ },
1745
+ {
1746
+ "epoch": 20.8,
1747
+ "learning_rate": 4.725628140703518e-06,
1748
+ "loss": 0.0003,
1749
+ "step": 5300
1750
+ },
1751
+ {
1752
+ "epoch": 20.8,
1753
+ "eval_loss": 0.8219889402389526,
1754
+ "eval_runtime": 1625.3438,
1755
+ "eval_samples_per_second": 5.319,
1756
+ "eval_steps_per_second": 0.532,
1757
+ "eval_wer": 51.487489871177914,
1758
+ "step": 5300
1759
+ },
1760
+ {
1761
+ "epoch": 20.9,
1762
+ "learning_rate": 4.700502512562814e-06,
1763
+ "loss": 0.0005,
1764
+ "step": 5325
1765
+ },
1766
+ {
1767
+ "epoch": 21.0,
1768
+ "learning_rate": 4.675376884422111e-06,
1769
+ "loss": 0.0004,
1770
+ "step": 5350
1771
+ },
1772
+ {
1773
+ "epoch": 21.1,
1774
+ "learning_rate": 4.650251256281408e-06,
1775
+ "loss": 0.0013,
1776
+ "step": 5375
1777
+ },
1778
+ {
1779
+ "epoch": 21.2,
1780
+ "learning_rate": 4.625125628140703e-06,
1781
+ "loss": 0.0003,
1782
+ "step": 5400
1783
+ },
1784
+ {
1785
+ "epoch": 21.2,
1786
+ "eval_loss": 0.809751033782959,
1787
+ "eval_runtime": 1570.6528,
1788
+ "eval_samples_per_second": 5.505,
1789
+ "eval_steps_per_second": 0.551,
1790
+ "eval_wer": 45.861652693026414,
1791
+ "step": 5400
1792
+ },
1793
+ {
1794
+ "epoch": 21.3,
1795
+ "learning_rate": 4.600000000000001e-06,
1796
+ "loss": 0.0003,
1797
+ "step": 5425
1798
+ },
1799
+ {
1800
+ "epoch": 21.39,
1801
+ "learning_rate": 4.574874371859297e-06,
1802
+ "loss": 0.0003,
1803
+ "step": 5450
1804
+ },
1805
+ {
1806
+ "epoch": 21.49,
1807
+ "learning_rate": 4.549748743718593e-06,
1808
+ "loss": 0.0006,
1809
+ "step": 5475
1810
+ },
1811
+ {
1812
+ "epoch": 21.59,
1813
+ "learning_rate": 4.52462311557789e-06,
1814
+ "loss": 0.0003,
1815
+ "step": 5500
1816
+ },
1817
+ {
1818
+ "epoch": 21.59,
1819
+ "eval_loss": 0.8131942749023438,
1820
+ "eval_runtime": 1556.938,
1821
+ "eval_samples_per_second": 5.553,
1822
+ "eval_steps_per_second": 0.556,
1823
+ "eval_wer": 44.87109523573283,
1824
+ "step": 5500
1825
+ },
1826
+ {
1827
+ "epoch": 21.69,
1828
+ "learning_rate": 4.499497487437186e-06,
1829
+ "loss": 0.0002,
1830
+ "step": 5525
1831
+ },
1832
+ {
1833
+ "epoch": 21.79,
1834
+ "learning_rate": 4.474371859296483e-06,
1835
+ "loss": 0.0003,
1836
+ "step": 5550
1837
+ },
1838
+ {
1839
+ "epoch": 21.88,
1840
+ "learning_rate": 4.44924623115578e-06,
1841
+ "loss": 0.0006,
1842
+ "step": 5575
1843
+ },
1844
+ {
1845
+ "epoch": 21.98,
1846
+ "learning_rate": 4.4241206030150755e-06,
1847
+ "loss": 0.0009,
1848
+ "step": 5600
1849
+ },
1850
+ {
1851
+ "epoch": 21.98,
1852
+ "eval_loss": 0.8005769848823547,
1853
+ "eval_runtime": 1557.485,
1854
+ "eval_samples_per_second": 5.551,
1855
+ "eval_steps_per_second": 0.555,
1856
+ "eval_wer": 45.393659770799225,
1857
+ "step": 5600
1858
+ },
1859
+ {
1860
+ "epoch": 22.08,
1861
+ "learning_rate": 4.398994974874372e-06,
1862
+ "loss": 0.0009,
1863
+ "step": 5625
1864
+ },
1865
+ {
1866
+ "epoch": 22.18,
1867
+ "learning_rate": 4.373869346733669e-06,
1868
+ "loss": 0.0007,
1869
+ "step": 5650
1870
+ },
1871
+ {
1872
+ "epoch": 22.28,
1873
+ "learning_rate": 4.348743718592965e-06,
1874
+ "loss": 0.0011,
1875
+ "step": 5675
1876
+ },
1877
+ {
1878
+ "epoch": 22.37,
1879
+ "learning_rate": 4.323618090452262e-06,
1880
+ "loss": 0.0003,
1881
+ "step": 5700
1882
+ },
1883
+ {
1884
+ "epoch": 22.37,
1885
+ "eval_loss": 0.8007655739784241,
1886
+ "eval_runtime": 1588.3662,
1887
+ "eval_samples_per_second": 5.443,
1888
+ "eval_steps_per_second": 0.545,
1889
+ "eval_wer": 45.61856096310628,
1890
+ "step": 5700
1891
+ },
1892
+ {
1893
+ "epoch": 22.47,
1894
+ "learning_rate": 4.298492462311558e-06,
1895
+ "loss": 0.0004,
1896
+ "step": 5725
1897
+ },
1898
+ {
1899
+ "epoch": 22.57,
1900
+ "learning_rate": 4.273366834170854e-06,
1901
+ "loss": 0.0004,
1902
+ "step": 5750
1903
+ },
1904
+ {
1905
+ "epoch": 22.67,
1906
+ "learning_rate": 4.248241206030151e-06,
1907
+ "loss": 0.0004,
1908
+ "step": 5775
1909
+ },
1910
+ {
1911
+ "epoch": 22.77,
1912
+ "learning_rate": 4.223115577889448e-06,
1913
+ "loss": 0.0002,
1914
+ "step": 5800
1915
+ },
1916
+ {
1917
+ "epoch": 22.77,
1918
+ "eval_loss": 0.8081113696098328,
1919
+ "eval_runtime": 1604.8728,
1920
+ "eval_samples_per_second": 5.387,
1921
+ "eval_steps_per_second": 0.539,
1922
+ "eval_wer": 46.324684559540934,
1923
+ "step": 5800
1924
+ },
1925
+ {
1926
+ "epoch": 22.87,
1927
+ "learning_rate": 4.197989949748744e-06,
1928
+ "loss": 0.0005,
1929
+ "step": 5825
1930
+ },
1931
+ {
1932
+ "epoch": 22.96,
1933
+ "learning_rate": 4.172864321608041e-06,
1934
+ "loss": 0.0005,
1935
+ "step": 5850
1936
+ },
1937
+ {
1938
+ "epoch": 23.06,
1939
+ "learning_rate": 4.147738693467337e-06,
1940
+ "loss": 0.0008,
1941
+ "step": 5875
1942
+ },
1943
+ {
1944
+ "epoch": 23.16,
1945
+ "learning_rate": 4.122613065326633e-06,
1946
+ "loss": 0.0002,
1947
+ "step": 5900
1948
+ },
1949
+ {
1950
+ "epoch": 23.16,
1951
+ "eval_loss": 0.8082166314125061,
1952
+ "eval_runtime": 1581.3962,
1953
+ "eval_samples_per_second": 5.467,
1954
+ "eval_steps_per_second": 0.547,
1955
+ "eval_wer": 46.12789601627226,
1956
+ "step": 5900
1957
+ },
1958
+ {
1959
+ "epoch": 23.26,
1960
+ "learning_rate": 4.09748743718593e-06,
1961
+ "loss": 0.0002,
1962
+ "step": 5925
1963
+ },
1964
+ {
1965
+ "epoch": 23.36,
1966
+ "learning_rate": 4.0723618090452265e-06,
1967
+ "loss": 0.0003,
1968
+ "step": 5950
1969
+ },
1970
+ {
1971
+ "epoch": 23.45,
1972
+ "learning_rate": 4.047236180904523e-06,
1973
+ "loss": 0.0003,
1974
+ "step": 5975
1975
+ },
1976
+ {
1977
+ "epoch": 23.55,
1978
+ "learning_rate": 4.02211055276382e-06,
1979
+ "loss": 0.0002,
1980
+ "step": 6000
1981
+ },
1982
+ {
1983
+ "epoch": 23.55,
1984
+ "eval_loss": 0.8238235116004944,
1985
+ "eval_runtime": 1599.956,
1986
+ "eval_samples_per_second": 5.404,
1987
+ "eval_steps_per_second": 0.541,
1988
+ "eval_wer": 46.177506573398816,
1989
+ "step": 6000
1990
+ },
1991
+ {
1992
+ "epoch": 23.65,
1993
+ "learning_rate": 3.9969849246231155e-06,
1994
+ "loss": 0.0005,
1995
+ "step": 6025
1996
+ },
1997
+ {
1998
+ "epoch": 23.75,
1999
+ "learning_rate": 3.971859296482413e-06,
2000
+ "loss": 0.0007,
2001
+ "step": 6050
2002
+ },
2003
+ {
2004
+ "epoch": 23.85,
2005
+ "learning_rate": 3.946733668341709e-06,
2006
+ "loss": 0.0004,
2007
+ "step": 6075
2008
+ },
2009
+ {
2010
+ "epoch": 23.95,
2011
+ "learning_rate": 3.921608040201005e-06,
2012
+ "loss": 0.0005,
2013
+ "step": 6100
2014
+ },
2015
+ {
2016
+ "epoch": 23.95,
2017
+ "eval_loss": 0.8119234442710876,
2018
+ "eval_runtime": 1612.6898,
2019
+ "eval_samples_per_second": 5.361,
2020
+ "eval_steps_per_second": 0.536,
2021
+ "eval_wer": 49.972714193580394,
2022
+ "step": 6100
2023
+ },
2024
+ {
2025
+ "epoch": 24.04,
2026
+ "learning_rate": 3.896482412060302e-06,
2027
+ "loss": 0.0006,
2028
+ "step": 6125
2029
+ },
2030
+ {
2031
+ "epoch": 24.14,
2032
+ "learning_rate": 3.871356783919598e-06,
2033
+ "loss": 0.0003,
2034
+ "step": 6150
2035
+ },
2036
+ {
2037
+ "epoch": 24.24,
2038
+ "learning_rate": 3.846231155778895e-06,
2039
+ "loss": 0.0009,
2040
+ "step": 6175
2041
+ },
2042
+ {
2043
+ "epoch": 24.34,
2044
+ "learning_rate": 3.821105527638191e-06,
2045
+ "loss": 0.0002,
2046
+ "step": 6200
2047
+ },
2048
+ {
2049
+ "epoch": 24.34,
2050
+ "eval_loss": 0.8315232992172241,
2051
+ "eval_runtime": 1606.8991,
2052
+ "eval_samples_per_second": 5.381,
2053
+ "eval_steps_per_second": 0.538,
2054
+ "eval_wer": 49.086338906252585,
2055
+ "step": 6200
2056
+ },
2057
+ {
2058
+ "epoch": 24.44,
2059
+ "learning_rate": 3.7959798994974876e-06,
2060
+ "loss": 0.0003,
2061
+ "step": 6225
2062
+ },
2063
+ {
2064
+ "epoch": 24.53,
2065
+ "learning_rate": 3.7708542713567843e-06,
2066
+ "loss": 0.0003,
2067
+ "step": 6250
2068
+ },
2069
+ {
2070
+ "epoch": 24.63,
2071
+ "learning_rate": 3.7457286432160805e-06,
2072
+ "loss": 0.0004,
2073
+ "step": 6275
2074
+ },
2075
+ {
2076
+ "epoch": 24.73,
2077
+ "learning_rate": 3.720603015075377e-06,
2078
+ "loss": 0.0001,
2079
+ "step": 6300
2080
+ },
2081
+ {
2082
+ "epoch": 24.73,
2083
+ "eval_loss": 0.8223620057106018,
2084
+ "eval_runtime": 1589.1151,
2085
+ "eval_samples_per_second": 5.441,
2086
+ "eval_steps_per_second": 0.544,
2087
+ "eval_wer": 47.22428932876917,
2088
+ "step": 6300
2089
+ },
2090
+ {
2091
+ "epoch": 24.83,
2092
+ "learning_rate": 3.6954773869346737e-06,
2093
+ "loss": 0.0002,
2094
+ "step": 6325
2095
+ },
2096
+ {
2097
+ "epoch": 24.93,
2098
+ "learning_rate": 3.6703517587939703e-06,
2099
+ "loss": 0.0003,
2100
+ "step": 6350
2101
+ },
2102
+ {
2103
+ "epoch": 25.02,
2104
+ "learning_rate": 3.6452261306532665e-06,
2105
+ "loss": 0.0001,
2106
+ "step": 6375
2107
+ },
2108
+ {
2109
+ "epoch": 25.12,
2110
+ "learning_rate": 3.620100502512563e-06,
2111
+ "loss": 0.0001,
2112
+ "step": 6400
2113
+ },
2114
+ {
2115
+ "epoch": 25.12,
2116
+ "eval_loss": 0.8258624076843262,
2117
+ "eval_runtime": 1596.5065,
2118
+ "eval_samples_per_second": 5.416,
2119
+ "eval_steps_per_second": 0.542,
2120
+ "eval_wer": 47.1680640306924,
2121
+ "step": 6400
2122
+ },
2123
+ {
2124
+ "epoch": 25.22,
2125
+ "learning_rate": 3.5949748743718593e-06,
2126
+ "loss": 0.0002,
2127
+ "step": 6425
2128
+ },
2129
+ {
2130
+ "epoch": 25.32,
2131
+ "learning_rate": 3.5698492462311564e-06,
2132
+ "loss": 0.0001,
2133
+ "step": 6450
2134
+ },
2135
+ {
2136
+ "epoch": 25.42,
2137
+ "learning_rate": 3.5447236180904526e-06,
2138
+ "loss": 0.0004,
2139
+ "step": 6475
2140
+ },
2141
+ {
2142
+ "epoch": 25.52,
2143
+ "learning_rate": 3.519597989949749e-06,
2144
+ "loss": 0.0001,
2145
+ "step": 6500
2146
+ },
2147
+ {
2148
+ "epoch": 25.52,
2149
+ "eval_loss": 0.8219370245933533,
2150
+ "eval_runtime": 1629.406,
2151
+ "eval_samples_per_second": 5.306,
2152
+ "eval_steps_per_second": 0.531,
2153
+ "eval_wer": 48.5736964826115,
2154
+ "step": 6500
2155
+ },
2156
+ {
2157
+ "epoch": 25.61,
2158
+ "learning_rate": 3.4944723618090454e-06,
2159
+ "loss": 0.0002,
2160
+ "step": 6525
2161
+ },
2162
+ {
2163
+ "epoch": 25.71,
2164
+ "learning_rate": 3.4693467336683416e-06,
2165
+ "loss": 0.0001,
2166
+ "step": 6550
2167
+ },
2168
+ {
2169
+ "epoch": 25.81,
2170
+ "learning_rate": 3.4442211055276386e-06,
2171
+ "loss": 0.0001,
2172
+ "step": 6575
2173
+ },
2174
+ {
2175
+ "epoch": 25.91,
2176
+ "learning_rate": 3.419095477386935e-06,
2177
+ "loss": 0.0002,
2178
+ "step": 6600
2179
+ },
2180
+ {
2181
+ "epoch": 25.91,
2182
+ "eval_loss": 0.839992344379425,
2183
+ "eval_runtime": 1605.3111,
2184
+ "eval_samples_per_second": 5.386,
2185
+ "eval_steps_per_second": 0.539,
2186
+ "eval_wer": 48.90774090059698,
2187
+ "step": 6600
2188
+ },
2189
+ {
2190
+ "epoch": 26.01,
2191
+ "learning_rate": 3.3939698492462315e-06,
2192
+ "loss": 0.0009,
2193
+ "step": 6625
2194
+ },
2195
+ {
2196
+ "epoch": 26.1,
2197
+ "learning_rate": 3.3688442211055276e-06,
2198
+ "loss": 0.0004,
2199
+ "step": 6650
2200
+ },
2201
+ {
2202
+ "epoch": 26.2,
2203
+ "learning_rate": 3.3437185929648243e-06,
2204
+ "loss": 0.0002,
2205
+ "step": 6675
2206
+ },
2207
+ {
2208
+ "epoch": 26.3,
2209
+ "learning_rate": 3.318592964824121e-06,
2210
+ "loss": 0.0005,
2211
+ "step": 6700
2212
+ },
2213
+ {
2214
+ "epoch": 26.3,
2215
+ "eval_loss": 0.8318980932235718,
2216
+ "eval_runtime": 1614.1036,
2217
+ "eval_samples_per_second": 5.357,
2218
+ "eval_steps_per_second": 0.536,
2219
+ "eval_wer": 47.55668006151709,
2220
+ "step": 6700
2221
+ },
2222
+ {
2223
+ "epoch": 26.4,
2224
+ "learning_rate": 3.2934673366834175e-06,
2225
+ "loss": 0.0002,
2226
+ "step": 6725
2227
+ },
2228
+ {
2229
+ "epoch": 26.5,
2230
+ "learning_rate": 3.2683417085427137e-06,
2231
+ "loss": 0.0002,
2232
+ "step": 6750
2233
+ },
2234
+ {
2235
+ "epoch": 26.59,
2236
+ "learning_rate": 3.2432160804020103e-06,
2237
+ "loss": 0.0006,
2238
+ "step": 6775
2239
+ },
2240
+ {
2241
+ "epoch": 26.69,
2242
+ "learning_rate": 3.2180904522613065e-06,
2243
+ "loss": 0.0001,
2244
+ "step": 6800
2245
+ },
2246
+ {
2247
+ "epoch": 26.69,
2248
+ "eval_loss": 0.8393569588661194,
2249
+ "eval_runtime": 1658.5859,
2250
+ "eval_samples_per_second": 5.213,
2251
+ "eval_steps_per_second": 0.522,
2252
+ "eval_wer": 50.23565014635114,
2253
+ "step": 6800
2254
+ },
2255
+ {
2256
+ "epoch": 26.79,
2257
+ "learning_rate": 3.1929648241206036e-06,
2258
+ "loss": 0.0001,
2259
+ "step": 6825
2260
+ },
2261
+ {
2262
+ "epoch": 26.89,
2263
+ "learning_rate": 3.1688442211055283e-06,
2264
+ "loss": 0.0002,
2265
+ "step": 6850
2266
+ },
2267
+ {
2268
+ "epoch": 26.99,
2269
+ "learning_rate": 3.1437185929648245e-06,
2270
+ "loss": 0.0005,
2271
+ "step": 6875
2272
+ },
2273
+ {
2274
+ "epoch": 27.09,
2275
+ "learning_rate": 3.1185929648241207e-06,
2276
+ "loss": 0.0001,
2277
+ "step": 6900
2278
+ },
2279
+ {
2280
+ "epoch": 27.09,
2281
+ "eval_loss": 0.8479753136634827,
2282
+ "eval_runtime": 1618.6719,
2283
+ "eval_samples_per_second": 5.341,
2284
+ "eval_steps_per_second": 0.534,
2285
+ "eval_wer": 48.46289957169552,
2286
+ "step": 6900
2287
+ },
2288
+ {
2289
+ "epoch": 27.18,
2290
+ "learning_rate": 3.0934673366834173e-06,
2291
+ "loss": 0.0004,
2292
+ "step": 6925
2293
+ },
2294
+ {
2295
+ "epoch": 27.28,
2296
+ "learning_rate": 3.0683417085427135e-06,
2297
+ "loss": 0.0001,
2298
+ "step": 6950
2299
+ },
2300
+ {
2301
+ "epoch": 27.38,
2302
+ "learning_rate": 3.0432160804020105e-06,
2303
+ "loss": 0.0003,
2304
+ "step": 6975
2305
+ },
2306
+ {
2307
+ "epoch": 27.48,
2308
+ "learning_rate": 3.0180904522613067e-06,
2309
+ "loss": 0.0001,
2310
+ "step": 7000
2311
+ },
2312
+ {
2313
+ "epoch": 27.48,
2314
+ "eval_loss": 0.8498477339744568,
2315
+ "eval_runtime": 1594.4093,
2316
+ "eval_samples_per_second": 5.423,
2317
+ "eval_steps_per_second": 0.543,
2318
+ "eval_wer": 47.11514610309074,
2319
+ "step": 7000
2320
+ },
2321
+ {
2322
+ "epoch": 27.58,
2323
+ "learning_rate": 2.9929648241206034e-06,
2324
+ "loss": 0.0001,
2325
+ "step": 7025
2326
+ },
2327
+ {
2328
+ "epoch": 27.67,
2329
+ "learning_rate": 2.9678391959798996e-06,
2330
+ "loss": 0.0001,
2331
+ "step": 7050
2332
+ },
2333
+ {
2334
+ "epoch": 27.77,
2335
+ "learning_rate": 2.942713567839196e-06,
2336
+ "loss": 0.0003,
2337
+ "step": 7075
2338
+ },
2339
+ {
2340
+ "epoch": 27.87,
2341
+ "learning_rate": 2.917587939698493e-06,
2342
+ "loss": 0.0002,
2343
+ "step": 7100
2344
+ },
2345
+ {
2346
+ "epoch": 27.87,
2347
+ "eval_loss": 0.8341818451881409,
2348
+ "eval_runtime": 1613.2296,
2349
+ "eval_samples_per_second": 5.359,
2350
+ "eval_steps_per_second": 0.536,
2351
+ "eval_wer": 48.9242777529725,
2352
+ "step": 7100
2353
+ },
2354
+ {
2355
+ "epoch": 27.97,
2356
+ "learning_rate": 2.8924623115577894e-06,
2357
+ "loss": 0.0002,
2358
+ "step": 7125
2359
+ },
2360
+ {
2361
+ "epoch": 28.07,
2362
+ "learning_rate": 2.8673366834170856e-06,
2363
+ "loss": 0.0004,
2364
+ "step": 7150
2365
+ },
2366
+ {
2367
+ "epoch": 28.16,
2368
+ "learning_rate": 2.8422110552763822e-06,
2369
+ "loss": 0.0005,
2370
+ "step": 7175
2371
+ },
2372
+ {
2373
+ "epoch": 28.26,
2374
+ "learning_rate": 2.8170854271356784e-06,
2375
+ "loss": 0.0003,
2376
+ "step": 7200
2377
+ },
2378
+ {
2379
+ "epoch": 28.26,
2380
+ "eval_loss": 0.8183611035346985,
2381
+ "eval_runtime": 1606.5293,
2382
+ "eval_samples_per_second": 5.382,
2383
+ "eval_steps_per_second": 0.538,
2384
+ "eval_wer": 47.373121000148835,
2385
+ "step": 7200
2386
+ },
2387
+ {
2388
+ "epoch": 28.36,
2389
+ "learning_rate": 2.7919597989949755e-06,
2390
+ "loss": 0.0002,
2391
+ "step": 7225
2392
+ },
2393
+ {
2394
+ "epoch": 28.46,
2395
+ "learning_rate": 2.7668341708542717e-06,
2396
+ "loss": 0.0002,
2397
+ "step": 7250
2398
+ },
2399
+ {
2400
+ "epoch": 28.56,
2401
+ "learning_rate": 2.7417085427135683e-06,
2402
+ "loss": 0.0002,
2403
+ "step": 7275
2404
+ },
2405
+ {
2406
+ "epoch": 28.66,
2407
+ "learning_rate": 2.7165829145728645e-06,
2408
+ "loss": 0.0001,
2409
+ "step": 7300
2410
+ },
2411
+ {
2412
+ "epoch": 28.66,
2413
+ "eval_loss": 0.8278088569641113,
2414
+ "eval_runtime": 1611.8451,
2415
+ "eval_samples_per_second": 5.364,
2416
+ "eval_steps_per_second": 0.537,
2417
+ "eval_wer": 47.92875923996626,
2418
+ "step": 7300
2419
+ },
2420
+ {
2421
+ "epoch": 28.75,
2422
+ "learning_rate": 2.6914572864321607e-06,
2423
+ "loss": 0.0001,
2424
+ "step": 7325
2425
+ },
2426
+ {
2427
+ "epoch": 28.85,
2428
+ "learning_rate": 2.6663316582914573e-06,
2429
+ "loss": 0.0001,
2430
+ "step": 7350
2431
+ },
2432
+ {
2433
+ "epoch": 28.95,
2434
+ "learning_rate": 2.6412060301507544e-06,
2435
+ "loss": 0.0002,
2436
+ "step": 7375
2437
+ },
2438
+ {
2439
+ "epoch": 29.05,
2440
+ "learning_rate": 2.6160804020100506e-06,
2441
+ "loss": 0.0002,
2442
+ "step": 7400
2443
+ },
2444
+ {
2445
+ "epoch": 29.05,
2446
+ "eval_loss": 0.8438936471939087,
2447
+ "eval_runtime": 1612.1092,
2448
+ "eval_samples_per_second": 5.363,
2449
+ "eval_steps_per_second": 0.537,
2450
+ "eval_wer": 47.86095814522664,
2451
+ "step": 7400
2452
+ },
2453
+ {
2454
+ "epoch": 29.15,
2455
+ "learning_rate": 2.5909547738693468e-06,
2456
+ "loss": 0.0003,
2457
+ "step": 7425
2458
+ },
2459
+ {
2460
+ "epoch": 29.24,
2461
+ "learning_rate": 2.5658291457286434e-06,
2462
+ "loss": 0.0001,
2463
+ "step": 7450
2464
+ },
2465
+ {
2466
+ "epoch": 29.34,
2467
+ "learning_rate": 2.5407035175879396e-06,
2468
+ "loss": 0.0001,
2469
+ "step": 7475
2470
+ },
2471
+ {
2472
+ "epoch": 29.44,
2473
+ "learning_rate": 2.5155778894472366e-06,
2474
+ "loss": 0.0001,
2475
+ "step": 7500
2476
+ },
2477
+ {
2478
+ "epoch": 29.44,
2479
+ "eval_loss": 0.8461014628410339,
2480
+ "eval_runtime": 1631.5395,
2481
+ "eval_samples_per_second": 5.299,
2482
+ "eval_steps_per_second": 0.53,
2483
+ "eval_wer": 49.94625522977956,
2484
+ "step": 7500
2485
+ },
2486
+ {
2487
+ "epoch": 29.54,
2488
+ "learning_rate": 2.490452261306533e-06,
2489
+ "loss": 0.0001,
2490
+ "step": 7525
2491
+ },
2492
+ {
2493
+ "epoch": 29.64,
2494
+ "learning_rate": 2.4653266331658294e-06,
2495
+ "loss": 0.0002,
2496
+ "step": 7550
2497
+ },
2498
+ {
2499
+ "epoch": 29.74,
2500
+ "learning_rate": 2.4402010050251256e-06,
2501
+ "loss": 0.0001,
2502
+ "step": 7575
2503
+ },
2504
+ {
2505
+ "epoch": 29.83,
2506
+ "learning_rate": 2.4150753768844223e-06,
2507
+ "loss": 0.0001,
2508
+ "step": 7600
2509
+ },
2510
+ {
2511
+ "epoch": 29.83,
2512
+ "eval_loss": 0.8448674082756042,
2513
+ "eval_runtime": 1642.4029,
2514
+ "eval_samples_per_second": 5.264,
2515
+ "eval_steps_per_second": 0.527,
2516
+ "eval_wer": 48.48605116502125,
2517
+ "step": 7600
2518
+ },
2519
+ {
2520
+ "epoch": 29.93,
2521
+ "learning_rate": 2.389949748743719e-06,
2522
+ "loss": 0.0002,
2523
+ "step": 7625
2524
+ },
2525
+ {
2526
+ "epoch": 30.03,
2527
+ "learning_rate": 2.3648241206030155e-06,
2528
+ "loss": 0.0001,
2529
+ "step": 7650
2530
+ },
2531
+ {
2532
+ "epoch": 30.13,
2533
+ "learning_rate": 2.3396984924623117e-06,
2534
+ "loss": 0.0001,
2535
+ "step": 7675
2536
+ },
2537
+ {
2538
+ "epoch": 30.23,
2539
+ "learning_rate": 2.314572864321608e-06,
2540
+ "loss": 0.0001,
2541
+ "step": 7700
2542
+ },
2543
+ {
2544
+ "epoch": 30.23,
2545
+ "eval_loss": 0.851162850856781,
2546
+ "eval_runtime": 1646.9818,
2547
+ "eval_samples_per_second": 5.25,
2548
+ "eval_steps_per_second": 0.525,
2549
+ "eval_wer": 49.000347273899884,
2550
+ "step": 7700
2551
+ },
2552
+ {
2553
+ "epoch": 30.32,
2554
+ "learning_rate": 2.289447236180905e-06,
2555
+ "loss": 0.0001,
2556
+ "step": 7725
2557
+ },
2558
+ {
2559
+ "epoch": 30.42,
2560
+ "learning_rate": 2.264321608040201e-06,
2561
+ "loss": 0.0001,
2562
+ "step": 7750
2563
+ },
2564
+ {
2565
+ "epoch": 30.52,
2566
+ "learning_rate": 2.2391959798994978e-06,
2567
+ "loss": 0.0001,
2568
+ "step": 7775
2569
+ },
2570
+ {
2571
+ "epoch": 30.62,
2572
+ "learning_rate": 2.214070351758794e-06,
2573
+ "loss": 0.0001,
2574
+ "step": 7800
2575
+ },
2576
+ {
2577
+ "epoch": 30.62,
2578
+ "eval_loss": 0.8554991483688354,
2579
+ "eval_runtime": 1638.7791,
2580
+ "eval_samples_per_second": 5.276,
2581
+ "eval_steps_per_second": 0.528,
2582
+ "eval_wer": 48.27768682508971,
2583
+ "step": 7800
2584
+ },
2585
+ {
2586
+ "epoch": 30.72,
2587
+ "learning_rate": 2.1889447236180906e-06,
2588
+ "loss": 0.0001,
2589
+ "step": 7825
2590
+ },
2591
+ {
2592
+ "epoch": 30.81,
2593
+ "learning_rate": 2.163819095477387e-06,
2594
+ "loss": 0.0001,
2595
+ "step": 7850
2596
+ },
2597
+ {
2598
+ "epoch": 30.91,
2599
+ "learning_rate": 2.1386934673366834e-06,
2600
+ "loss": 0.0001,
2601
+ "step": 7875
2602
+ },
2603
+ {
2604
+ "epoch": 31.01,
2605
+ "learning_rate": 2.11356783919598e-06,
2606
+ "loss": 0.0001,
2607
+ "step": 7900
2608
+ },
2609
+ {
2610
+ "epoch": 31.01,
2611
+ "eval_loss": 0.854286253452301,
2612
+ "eval_runtime": 1639.4413,
2613
+ "eval_samples_per_second": 5.274,
2614
+ "eval_steps_per_second": 0.528,
2615
+ "eval_wer": 48.67953233781482,
2616
+ "step": 7900
2617
+ },
2618
+ {
2619
+ "epoch": 31.11,
2620
+ "learning_rate": 2.0884422110552766e-06,
2621
+ "loss": 0.0001,
2622
+ "step": 7925
2623
+ },
2624
+ {
2625
+ "epoch": 31.21,
2626
+ "learning_rate": 2.063316582914573e-06,
2627
+ "loss": 0.0001,
2628
+ "step": 7950
2629
+ },
2630
+ {
2631
+ "epoch": 31.31,
2632
+ "learning_rate": 2.0381909547738695e-06,
2633
+ "loss": 0.0001,
2634
+ "step": 7975
2635
+ },
2636
+ {
2637
+ "epoch": 31.4,
2638
+ "learning_rate": 2.013065326633166e-06,
2639
+ "loss": 0.0001,
2640
+ "step": 8000
2641
+ },
2642
+ {
2643
+ "epoch": 31.4,
2644
+ "eval_loss": 0.8566484451293945,
2645
+ "eval_runtime": 1631.2411,
2646
+ "eval_samples_per_second": 5.3,
2647
+ "eval_steps_per_second": 0.53,
2648
+ "eval_wer": 48.76552397016752,
2649
+ "step": 8000
2650
+ },
2651
+ {
2652
+ "epoch": 31.5,
2653
+ "learning_rate": 1.9879396984924627e-06,
2654
+ "loss": 0.0001,
2655
+ "step": 8025
2656
+ },
2657
+ {
2658
+ "epoch": 31.6,
2659
+ "learning_rate": 1.962814070351759e-06,
2660
+ "loss": 0.0001,
2661
+ "step": 8050
2662
+ },
2663
+ {
2664
+ "epoch": 31.7,
2665
+ "learning_rate": 1.9376884422110555e-06,
2666
+ "loss": 0.0001,
2667
+ "step": 8075
2668
+ },
2669
+ {
2670
+ "epoch": 31.8,
2671
+ "learning_rate": 1.912562814070352e-06,
2672
+ "loss": 0.0001,
2673
+ "step": 8100
2674
+ },
2675
+ {
2676
+ "epoch": 31.8,
2677
+ "eval_loss": 0.8605026602745056,
2678
+ "eval_runtime": 1651.0794,
2679
+ "eval_samples_per_second": 5.237,
2680
+ "eval_steps_per_second": 0.524,
2681
+ "eval_wer": 48.67787865257726,
2682
+ "step": 8100
2683
+ },
2684
+ {
2685
+ "epoch": 31.89,
2686
+ "learning_rate": 1.8874371859296483e-06,
2687
+ "loss": 0.0,
2688
+ "step": 8125
2689
+ },
2690
+ {
2691
+ "epoch": 31.99,
2692
+ "learning_rate": 1.862311557788945e-06,
2693
+ "loss": 0.0,
2694
+ "step": 8150
2695
+ },
2696
+ {
2697
+ "epoch": 32.09,
2698
+ "learning_rate": 1.8371859296482414e-06,
2699
+ "loss": 0.0,
2700
+ "step": 8175
2701
+ },
2702
+ {
2703
+ "epoch": 32.19,
2704
+ "learning_rate": 1.8120603015075378e-06,
2705
+ "loss": 0.0,
2706
+ "step": 8200
2707
+ },
2708
+ {
2709
+ "epoch": 32.19,
2710
+ "eval_loss": 0.8633646965026855,
2711
+ "eval_runtime": 1649.2073,
2712
+ "eval_samples_per_second": 5.243,
2713
+ "eval_steps_per_second": 0.524,
2714
+ "eval_wer": 49.36911908187396,
2715
+ "step": 8200
2716
+ },
2717
+ {
2718
+ "epoch": 32.29,
2719
+ "learning_rate": 1.7869346733668344e-06,
2720
+ "loss": 0.0,
2721
+ "step": 8225
2722
+ },
2723
+ {
2724
+ "epoch": 32.38,
2725
+ "learning_rate": 1.7618090452261308e-06,
2726
+ "loss": 0.0,
2727
+ "step": 8250
2728
+ },
2729
+ {
2730
+ "epoch": 32.48,
2731
+ "learning_rate": 1.7366834170854272e-06,
2732
+ "loss": 0.0,
2733
+ "step": 8275
2734
+ },
2735
+ {
2736
+ "epoch": 32.58,
2737
+ "learning_rate": 1.7115577889447238e-06,
2738
+ "loss": 0.0,
2739
+ "step": 8300
2740
+ },
2741
+ {
2742
+ "epoch": 32.58,
2743
+ "eval_loss": 0.8662922978401184,
2744
+ "eval_runtime": 1644.0444,
2745
+ "eval_samples_per_second": 5.259,
2746
+ "eval_steps_per_second": 0.526,
2747
+ "eval_wer": 50.04382265879512,
2748
+ "step": 8300
2749
+ },
2750
+ {
2751
+ "epoch": 32.68,
2752
+ "learning_rate": 1.6864321608040202e-06,
2753
+ "loss": 0.0,
2754
+ "step": 8325
2755
+ },
2756
+ {
2757
+ "epoch": 32.78,
2758
+ "learning_rate": 1.6613065326633169e-06,
2759
+ "loss": 0.0,
2760
+ "step": 8350
2761
+ },
2762
+ {
2763
+ "epoch": 32.88,
2764
+ "learning_rate": 1.6361809045226133e-06,
2765
+ "loss": 0.0,
2766
+ "step": 8375
2767
+ },
2768
+ {
2769
+ "epoch": 32.97,
2770
+ "learning_rate": 1.6110552763819095e-06,
2771
+ "loss": 0.0,
2772
+ "step": 8400
2773
+ },
2774
+ {
2775
+ "epoch": 32.97,
2776
+ "eval_loss": 0.8684667944908142,
2777
+ "eval_runtime": 1656.8342,
2778
+ "eval_samples_per_second": 5.218,
2779
+ "eval_steps_per_second": 0.522,
2780
+ "eval_wer": 49.727968778422714,
2781
+ "step": 8400
2782
+ },
2783
+ {
2784
+ "epoch": 33.07,
2785
+ "learning_rate": 1.5859296482412063e-06,
2786
+ "loss": 0.0,
2787
+ "step": 8425
2788
+ },
2789
+ {
2790
+ "epoch": 33.17,
2791
+ "learning_rate": 1.5608040201005025e-06,
2792
+ "loss": 0.0,
2793
+ "step": 8450
2794
+ },
2795
+ {
2796
+ "epoch": 33.27,
2797
+ "learning_rate": 1.5356783919597993e-06,
2798
+ "loss": 0.0,
2799
+ "step": 8475
2800
+ },
2801
+ {
2802
+ "epoch": 33.37,
2803
+ "learning_rate": 1.5105527638190955e-06,
2804
+ "loss": 0.0,
2805
+ "step": 8500
2806
+ },
2807
+ {
2808
+ "epoch": 33.37,
2809
+ "eval_loss": 0.8704341053962708,
2810
+ "eval_runtime": 1651.7464,
2811
+ "eval_samples_per_second": 5.234,
2812
+ "eval_steps_per_second": 0.524,
2813
+ "eval_wer": 49.164062112417525,
2814
+ "step": 8500
2815
+ },
2816
+ {
2817
+ "epoch": 33.46,
2818
+ "learning_rate": 1.485427135678392e-06,
2819
+ "loss": 0.0,
2820
+ "step": 8525
2821
+ },
2822
+ {
2823
+ "epoch": 33.56,
2824
+ "learning_rate": 1.4603015075376886e-06,
2825
+ "loss": 0.0,
2826
+ "step": 8550
2827
+ },
2828
+ {
2829
+ "epoch": 33.66,
2830
+ "learning_rate": 1.435175879396985e-06,
2831
+ "loss": 0.0,
2832
+ "step": 8575
2833
+ },
2834
+ {
2835
+ "epoch": 33.76,
2836
+ "learning_rate": 1.4100502512562816e-06,
2837
+ "loss": 0.0,
2838
+ "step": 8600
2839
+ },
2840
+ {
2841
+ "epoch": 33.76,
2842
+ "eval_loss": 0.8724026083946228,
2843
+ "eval_runtime": 1643.6363,
2844
+ "eval_samples_per_second": 5.26,
2845
+ "eval_steps_per_second": 0.526,
2846
+ "eval_wer": 48.841593491094905,
2847
+ "step": 8600
2848
+ },
2849
+ {
2850
+ "epoch": 33.86,
2851
+ "learning_rate": 1.384924623115578e-06,
2852
+ "loss": 0.0,
2853
+ "step": 8625
2854
+ },
2855
+ {
2856
+ "epoch": 33.95,
2857
+ "learning_rate": 1.3597989949748744e-06,
2858
+ "loss": 0.0,
2859
+ "step": 8650
2860
+ },
2861
+ {
2862
+ "epoch": 34.05,
2863
+ "learning_rate": 1.334673366834171e-06,
2864
+ "loss": 0.0,
2865
+ "step": 8675
2866
+ },
2867
+ {
2868
+ "epoch": 34.15,
2869
+ "learning_rate": 1.3095477386934674e-06,
2870
+ "loss": 0.0,
2871
+ "step": 8700
2872
+ },
2873
+ {
2874
+ "epoch": 34.15,
2875
+ "eval_loss": 0.8736163377761841,
2876
+ "eval_runtime": 1654.9145,
2877
+ "eval_samples_per_second": 5.224,
2878
+ "eval_steps_per_second": 0.523,
2879
+ "eval_wer": 49.22855583668205,
2880
+ "step": 8700
2881
+ },
2882
+ {
2883
+ "epoch": 34.25,
2884
+ "learning_rate": 1.284422110552764e-06,
2885
+ "loss": 0.0,
2886
+ "step": 8725
2887
+ },
2888
+ {
2889
+ "epoch": 34.35,
2890
+ "learning_rate": 1.2592964824120605e-06,
2891
+ "loss": 0.0,
2892
+ "step": 8750
2893
+ },
2894
+ {
2895
+ "epoch": 34.45,
2896
+ "learning_rate": 1.2341708542713569e-06,
2897
+ "loss": 0.0,
2898
+ "step": 8775
2899
+ },
2900
+ {
2901
+ "epoch": 34.54,
2902
+ "learning_rate": 1.2090452261306533e-06,
2903
+ "loss": 0.0,
2904
+ "step": 8800
2905
+ },
2906
+ {
2907
+ "epoch": 34.54,
2908
+ "eval_loss": 0.8755289316177368,
2909
+ "eval_runtime": 1642.3091,
2910
+ "eval_samples_per_second": 5.265,
2911
+ "eval_steps_per_second": 0.527,
2912
+ "eval_wer": 48.61338492831275,
2913
+ "step": 8800
2914
+ },
2915
+ {
2916
+ "epoch": 34.64,
2917
+ "learning_rate": 1.18391959798995e-06,
2918
+ "loss": 0.0,
2919
+ "step": 8825
2920
+ },
2921
+ {
2922
+ "epoch": 34.74,
2923
+ "learning_rate": 1.1587939698492463e-06,
2924
+ "loss": 0.0,
2925
+ "step": 8850
2926
+ },
2927
+ {
2928
+ "epoch": 34.84,
2929
+ "learning_rate": 1.133668341708543e-06,
2930
+ "loss": 0.0,
2931
+ "step": 8875
2932
+ },
2933
+ {
2934
+ "epoch": 34.94,
2935
+ "learning_rate": 1.1085427135678393e-06,
2936
+ "loss": 0.0,
2937
+ "step": 8900
2938
+ },
2939
+ {
2940
+ "epoch": 34.94,
2941
+ "eval_loss": 0.8766536116600037,
2942
+ "eval_runtime": 1656.0111,
2943
+ "eval_samples_per_second": 5.221,
2944
+ "eval_steps_per_second": 0.522,
2945
+ "eval_wer": 48.92593143821005,
2946
+ "step": 8900
2947
+ },
2948
+ {
2949
+ "epoch": 35.03,
2950
+ "learning_rate": 1.0834170854271357e-06,
2951
+ "loss": 0.0,
2952
+ "step": 8925
2953
+ },
2954
+ {
2955
+ "epoch": 35.13,
2956
+ "learning_rate": 1.0582914572864322e-06,
2957
+ "loss": 0.0,
2958
+ "step": 8950
2959
+ },
2960
+ {
2961
+ "epoch": 35.23,
2962
+ "learning_rate": 1.0331658291457288e-06,
2963
+ "loss": 0.0,
2964
+ "step": 8975
2965
+ },
2966
+ {
2967
+ "epoch": 35.33,
2968
+ "learning_rate": 1.0080402010050252e-06,
2969
+ "loss": 0.0,
2970
+ "step": 9000
2971
+ },
2972
+ {
2973
+ "epoch": 35.33,
2974
+ "eval_loss": 0.8778066039085388,
2975
+ "eval_runtime": 1650.953,
2976
+ "eval_samples_per_second": 5.237,
2977
+ "eval_steps_per_second": 0.524,
2978
+ "eval_wer": 48.98050305104926,
2979
+ "step": 9000
2980
+ },
2981
+ {
2982
+ "epoch": 35.43,
2983
+ "learning_rate": 9.829145728643216e-07,
2984
+ "loss": 0.0,
2985
+ "step": 9025
2986
+ },
2987
+ {
2988
+ "epoch": 35.53,
2989
+ "learning_rate": 9.577889447236182e-07,
2990
+ "loss": 0.0,
2991
+ "step": 9050
2992
+ },
2993
+ {
2994
+ "epoch": 35.62,
2995
+ "learning_rate": 9.326633165829146e-07,
2996
+ "loss": 0.0,
2997
+ "step": 9075
2998
+ },
2999
+ {
3000
+ "epoch": 35.72,
3001
+ "learning_rate": 9.075376884422111e-07,
3002
+ "loss": 0.0,
3003
+ "step": 9100
3004
+ },
3005
+ {
3006
+ "epoch": 35.72,
3007
+ "eval_loss": 0.8791189193725586,
3008
+ "eval_runtime": 1640.1356,
3009
+ "eval_samples_per_second": 5.272,
3010
+ "eval_steps_per_second": 0.527,
3011
+ "eval_wer": 49.321162209984955,
3012
+ "step": 9100
3013
+ },
3014
+ {
3015
+ "epoch": 35.82,
3016
+ "learning_rate": 8.824120603015077e-07,
3017
+ "loss": 0.0,
3018
+ "step": 9125
3019
+ },
3020
+ {
3021
+ "epoch": 35.92,
3022
+ "learning_rate": 8.572864321608041e-07,
3023
+ "loss": 0.0,
3024
+ "step": 9150
3025
+ },
3026
+ {
3027
+ "epoch": 36.02,
3028
+ "learning_rate": 8.321608040201006e-07,
3029
+ "loss": 0.0,
3030
+ "step": 9175
3031
+ },
3032
+ {
3033
+ "epoch": 36.11,
3034
+ "learning_rate": 8.070351758793971e-07,
3035
+ "loss": 0.0,
3036
+ "step": 9200
3037
+ },
3038
+ {
3039
+ "epoch": 36.11,
3040
+ "eval_loss": 0.8800622820854187,
3041
+ "eval_runtime": 1655.7118,
3042
+ "eval_samples_per_second": 5.222,
3043
+ "eval_steps_per_second": 0.522,
3044
+ "eval_wer": 49.37242645234906,
3045
+ "step": 9200
3046
+ },
3047
+ {
3048
+ "epoch": 36.21,
3049
+ "learning_rate": 7.819095477386936e-07,
3050
+ "loss": 0.0,
3051
+ "step": 9225
3052
+ },
3053
+ {
3054
+ "epoch": 36.31,
3055
+ "learning_rate": 7.567839195979901e-07,
3056
+ "loss": 0.0,
3057
+ "step": 9250
3058
+ },
3059
+ {
3060
+ "epoch": 36.41,
3061
+ "learning_rate": 7.316582914572864e-07,
3062
+ "loss": 0.0,
3063
+ "step": 9275
3064
+ },
3065
+ {
3066
+ "epoch": 36.51,
3067
+ "learning_rate": 7.065326633165829e-07,
3068
+ "loss": 0.0,
3069
+ "step": 9300
3070
+ },
3071
+ {
3072
+ "epoch": 36.51,
3073
+ "eval_loss": 0.881312906742096,
3074
+ "eval_runtime": 1661.7481,
3075
+ "eval_samples_per_second": 5.203,
3076
+ "eval_steps_per_second": 0.521,
3077
+ "eval_wer": 49.43361280613848,
3078
+ "step": 9300
3079
+ },
3080
+ {
3081
+ "epoch": 36.6,
3082
+ "learning_rate": 6.814070351758795e-07,
3083
+ "loss": 0.0,
3084
+ "step": 9325
3085
+ },
3086
+ {
3087
+ "epoch": 36.7,
3088
+ "learning_rate": 6.56281407035176e-07,
3089
+ "loss": 0.0,
3090
+ "step": 9350
3091
+ },
3092
+ {
3093
+ "epoch": 36.8,
3094
+ "learning_rate": 6.311557788944724e-07,
3095
+ "loss": 0.0,
3096
+ "step": 9375
3097
+ },
3098
+ {
3099
+ "epoch": 36.9,
3100
+ "learning_rate": 6.060301507537689e-07,
3101
+ "loss": 0.0,
3102
+ "step": 9400
3103
+ },
3104
+ {
3105
+ "epoch": 36.9,
3106
+ "eval_loss": 0.8819226622581482,
3107
+ "eval_runtime": 1659.1244,
3108
+ "eval_samples_per_second": 5.211,
3109
+ "eval_steps_per_second": 0.521,
3110
+ "eval_wer": 49.10452944386565,
3111
+ "step": 9400
3112
+ },
3113
+ {
3114
+ "epoch": 37.0,
3115
+ "learning_rate": 5.809045226130654e-07,
3116
+ "loss": 0.0,
3117
+ "step": 9425
3118
+ },
3119
+ {
3120
+ "epoch": 37.1,
3121
+ "learning_rate": 5.557788944723619e-07,
3122
+ "loss": 0.0,
3123
+ "step": 9450
3124
+ },
3125
+ {
3126
+ "epoch": 37.19,
3127
+ "learning_rate": 5.306532663316583e-07,
3128
+ "loss": 0.0,
3129
+ "step": 9475
3130
+ },
3131
+ {
3132
+ "epoch": 37.29,
3133
+ "learning_rate": 5.055276381909549e-07,
3134
+ "loss": 0.0,
3135
+ "step": 9500
3136
+ },
3137
+ {
3138
+ "epoch": 37.29,
3139
+ "eval_loss": 0.8826147317886353,
3140
+ "eval_runtime": 1662.0225,
3141
+ "eval_samples_per_second": 5.202,
3142
+ "eval_steps_per_second": 0.52,
3143
+ "eval_wer": 49.26328322667064,
3144
+ "step": 9500
3145
+ },
3146
+ {
3147
+ "epoch": 37.39,
3148
+ "learning_rate": 4.804020100502513e-07,
3149
+ "loss": 0.0,
3150
+ "step": 9525
3151
+ },
3152
+ {
3153
+ "epoch": 37.49,
3154
+ "learning_rate": 4.552763819095478e-07,
3155
+ "loss": 0.0,
3156
+ "step": 9550
3157
+ },
3158
+ {
3159
+ "epoch": 37.59,
3160
+ "learning_rate": 4.3015075376884424e-07,
3161
+ "loss": 0.0,
3162
+ "step": 9575
3163
+ },
3164
+ {
3165
+ "epoch": 37.68,
3166
+ "learning_rate": 4.0502512562814075e-07,
3167
+ "loss": 0.0,
3168
+ "step": 9600
3169
+ },
3170
+ {
3171
+ "epoch": 37.68,
3172
+ "eval_loss": 0.8831948041915894,
3173
+ "eval_runtime": 1662.4658,
3174
+ "eval_samples_per_second": 5.201,
3175
+ "eval_steps_per_second": 0.52,
3176
+ "eval_wer": 49.423690694713166,
3177
+ "step": 9600
3178
+ },
3179
+ {
3180
+ "epoch": 37.78,
3181
+ "learning_rate": 3.7989949748743727e-07,
3182
+ "loss": 0.0,
3183
+ "step": 9625
3184
+ },
3185
+ {
3186
+ "epoch": 37.88,
3187
+ "learning_rate": 3.547738693467337e-07,
3188
+ "loss": 0.0,
3189
+ "step": 9650
3190
+ },
3191
+ {
3192
+ "epoch": 37.98,
3193
+ "learning_rate": 3.296482412060302e-07,
3194
+ "loss": 0.0,
3195
+ "step": 9675
3196
+ },
3197
+ {
3198
+ "epoch": 38.08,
3199
+ "learning_rate": 3.0452261306532665e-07,
3200
+ "loss": 0.0,
3201
+ "step": 9700
3202
+ },
3203
+ {
3204
+ "epoch": 38.08,
3205
+ "eval_loss": 0.8836556673049927,
3206
+ "eval_runtime": 1645.335,
3207
+ "eval_samples_per_second": 5.255,
3208
+ "eval_steps_per_second": 0.526,
3209
+ "eval_wer": 48.63157546592581,
3210
+ "step": 9700
3211
+ },
3212
+ {
3213
+ "epoch": 38.17,
3214
+ "learning_rate": 2.7939698492462317e-07,
3215
+ "loss": 0.0,
3216
+ "step": 9725
3217
+ },
3218
+ {
3219
+ "epoch": 38.27,
3220
+ "learning_rate": 2.5427135678391963e-07,
3221
+ "loss": 0.0,
3222
+ "step": 9750
3223
+ },
3224
+ {
3225
+ "epoch": 38.37,
3226
+ "learning_rate": 2.291457286432161e-07,
3227
+ "loss": 0.0,
3228
+ "step": 9775
3229
+ },
3230
+ {
3231
+ "epoch": 38.47,
3232
+ "learning_rate": 2.0402010050251258e-07,
3233
+ "loss": 0.0,
3234
+ "step": 9800
3235
+ },
3236
+ {
3237
+ "epoch": 38.47,
3238
+ "eval_loss": 0.8840765953063965,
3239
+ "eval_runtime": 1659.0687,
3240
+ "eval_samples_per_second": 5.211,
3241
+ "eval_steps_per_second": 0.521,
3242
+ "eval_wer": 48.64645863306378,
3243
+ "step": 9800
3244
+ },
3245
+ {
3246
+ "epoch": 38.57,
3247
+ "learning_rate": 1.7889447236180904e-07,
3248
+ "loss": 0.0,
3249
+ "step": 9825
3250
+ },
3251
+ {
3252
+ "epoch": 38.67,
3253
+ "learning_rate": 1.5376884422110556e-07,
3254
+ "loss": 0.0,
3255
+ "step": 9850
3256
+ },
3257
+ {
3258
+ "epoch": 38.76,
3259
+ "learning_rate": 1.2864321608040202e-07,
3260
+ "loss": 0.0,
3261
+ "step": 9875
3262
+ },
3263
+ {
3264
+ "epoch": 38.86,
3265
+ "learning_rate": 1.035175879396985e-07,
3266
+ "loss": 0.0,
3267
+ "step": 9900
3268
+ },
3269
+ {
3270
+ "epoch": 38.86,
3271
+ "eval_loss": 0.8842225670814514,
3272
+ "eval_runtime": 1661.5936,
3273
+ "eval_samples_per_second": 5.203,
3274
+ "eval_steps_per_second": 0.521,
3275
+ "eval_wer": 48.93419986439781,
3276
+ "step": 9900
3277
+ },
3278
+ {
3279
+ "epoch": 38.96,
3280
+ "learning_rate": 7.839195979899497e-08,
3281
+ "loss": 0.0,
3282
+ "step": 9925
3283
+ },
3284
+ {
3285
+ "epoch": 39.06,
3286
+ "learning_rate": 5.326633165829146e-08,
3287
+ "loss": 0.0,
3288
+ "step": 9950
3289
+ },
3290
+ {
3291
+ "epoch": 39.16,
3292
+ "learning_rate": 2.8140703517587944e-08,
3293
+ "loss": 0.0,
3294
+ "step": 9975
3295
+ },
3296
+ {
3297
+ "epoch": 39.25,
3298
+ "learning_rate": 3.0150753768844226e-09,
3299
+ "loss": 0.0,
3300
+ "step": 10000
3301
+ },
3302
+ {
3303
+ "epoch": 39.25,
3304
+ "eval_loss": 0.8843041062355042,
3305
+ "eval_runtime": 1669.3748,
3306
+ "eval_samples_per_second": 5.179,
3307
+ "eval_steps_per_second": 0.518,
3308
+ "eval_wer": 48.64480494782623,
3309
+ "step": 10000
3310
+ }
3311
+ ],
3312
+ "logging_steps": 25,
3313
+ "max_steps": 10000,
3314
+ "num_train_epochs": 40,
3315
+ "save_steps": 100,
3316
+ "total_flos": 1.6150653070368768e+20,
3317
+ "trial_name": null,
3318
+ "trial_params": null
3319
+ }
2/checkpoint-10000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c69a6da4c323af31ce3f4997ac3ae843fb847a604926bb5641ee181988d9e27c
3
+ size 4347
2/checkpoint-200/config.json ADDED
@@ -0,0 +1,52 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "makhataei/Whisper-Small-Common-Voice",
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
+ }
2/checkpoint-200/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
+ }
2/checkpoint-200/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5fe23069e04495d3beeb4738e901000dbd76323d127393373d23bb07a35be8ab
3
+ size 966995080
2/checkpoint-200/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e86a5afe13eec6f591da03463eba2a217e45c837fd5ccfd9e7291871768cd804
3
+ size 1925063607
2/checkpoint-200/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
+ }
2/checkpoint-200/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:67df5bf74243ac315c7474eeadbf810db4487a6df210f027b5e897b41730ee7f
3
+ size 14575
2/checkpoint-200/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4512321894017a036bb1dcbff32b137561c42b6747f564a155868db24fc8e0e5
3
+ size 627
2/checkpoint-200/trainer_state.json ADDED
@@ -0,0 +1,85 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": 46.749681665591766,
3
+ "best_model_checkpoint": "/media/makhataei/Backups/Whisper-Small-Common-Voice/checkpoint-200",
4
+ "epoch": 0.7850834151128557,
5
+ "eval_steps": 100,
6
+ "global_step": 200,
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": 5e-06,
14
+ "loss": 0.3008,
15
+ "step": 25
16
+ },
17
+ {
18
+ "epoch": 0.2,
19
+ "learning_rate": 1e-05,
20
+ "loss": 0.2368,
21
+ "step": 50
22
+ },
23
+ {
24
+ "epoch": 0.29,
25
+ "learning_rate": 9.974874371859297e-06,
26
+ "loss": 0.2001,
27
+ "step": 75
28
+ },
29
+ {
30
+ "epoch": 0.39,
31
+ "learning_rate": 9.949748743718594e-06,
32
+ "loss": 0.1801,
33
+ "step": 100
34
+ },
35
+ {
36
+ "epoch": 0.39,
37
+ "eval_loss": 0.4975946843624115,
38
+ "eval_runtime": 1546.7344,
39
+ "eval_samples_per_second": 5.59,
40
+ "eval_steps_per_second": 0.559,
41
+ "eval_wer": 49.12602735195383,
42
+ "step": 100
43
+ },
44
+ {
45
+ "epoch": 0.49,
46
+ "learning_rate": 9.92462311557789e-06,
47
+ "loss": 0.1713,
48
+ "step": 125
49
+ },
50
+ {
51
+ "epoch": 0.59,
52
+ "learning_rate": 9.899497487437186e-06,
53
+ "loss": 0.1657,
54
+ "step": 150
55
+ },
56
+ {
57
+ "epoch": 0.69,
58
+ "learning_rate": 9.874371859296484e-06,
59
+ "loss": 0.1574,
60
+ "step": 175
61
+ },
62
+ {
63
+ "epoch": 0.79,
64
+ "learning_rate": 9.84924623115578e-06,
65
+ "loss": 0.1597,
66
+ "step": 200
67
+ },
68
+ {
69
+ "epoch": 0.79,
70
+ "eval_loss": 0.46244511008262634,
71
+ "eval_runtime": 1507.9912,
72
+ "eval_samples_per_second": 5.733,
73
+ "eval_steps_per_second": 0.574,
74
+ "eval_wer": 46.749681665591766,
75
+ "step": 200
76
+ }
77
+ ],
78
+ "logging_steps": 25,
79
+ "max_steps": 10000,
80
+ "num_train_epochs": 40,
81
+ "save_steps": 100,
82
+ "total_flos": 3.232156483584e+18,
83
+ "trial_name": null,
84
+ "trial_params": null
85
+ }
2/checkpoint-200/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c69a6da4c323af31ce3f4997ac3ae843fb847a604926bb5641ee181988d9e27c
3
+ size 4347
2/checkpoint-300/config.json ADDED
@@ -0,0 +1,52 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "makhataei/Whisper-Small-Common-Voice",
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
+ }
2/checkpoint-300/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
+ }
2/checkpoint-300/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:77ce651e445777c9df154b004876c74a7f58463ef81ef61d0de064ee51d01c14
3
+ size 966995080
2/checkpoint-300/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6c7c2bdf41c203f0b6f2aefe71f2a1f4b8ae6d5811d63bd2481575e07ba1fabd
3
+ size 1925063607
2/checkpoint-300/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
+ }
2/checkpoint-300/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:51803a34fb0e22cdc9beeed2b4da86c2002d214413a9d153e0bc7b2e2ba14113
3
+ size 14575
2/checkpoint-300/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9389f201fdda18d7033678307705012510ad3b49e15684df8ebb8a3e262dce0c
3
+ size 627
2/checkpoint-300/trainer_state.json ADDED
@@ -0,0 +1,118 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": 43.17606786724215,
3
+ "best_model_checkpoint": "/media/makhataei/Backups/Whisper-Small-Common-Voice/checkpoint-300",
4
+ "epoch": 1.1776251226692835,
5
+ "eval_steps": 100,
6
+ "global_step": 300,
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": 5e-06,
14
+ "loss": 0.3008,
15
+ "step": 25
16
+ },
17
+ {
18
+ "epoch": 0.2,
19
+ "learning_rate": 1e-05,
20
+ "loss": 0.2368,
21
+ "step": 50
22
+ },
23
+ {
24
+ "epoch": 0.29,
25
+ "learning_rate": 9.974874371859297e-06,
26
+ "loss": 0.2001,
27
+ "step": 75
28
+ },
29
+ {
30
+ "epoch": 0.39,
31
+ "learning_rate": 9.949748743718594e-06,
32
+ "loss": 0.1801,
33
+ "step": 100
34
+ },
35
+ {
36
+ "epoch": 0.39,
37
+ "eval_loss": 0.4975946843624115,
38
+ "eval_runtime": 1546.7344,
39
+ "eval_samples_per_second": 5.59,
40
+ "eval_steps_per_second": 0.559,
41
+ "eval_wer": 49.12602735195383,
42
+ "step": 100
43
+ },
44
+ {
45
+ "epoch": 0.49,
46
+ "learning_rate": 9.92462311557789e-06,
47
+ "loss": 0.1713,
48
+ "step": 125
49
+ },
50
+ {
51
+ "epoch": 0.59,
52
+ "learning_rate": 9.899497487437186e-06,
53
+ "loss": 0.1657,
54
+ "step": 150
55
+ },
56
+ {
57
+ "epoch": 0.69,
58
+ "learning_rate": 9.874371859296484e-06,
59
+ "loss": 0.1574,
60
+ "step": 175
61
+ },
62
+ {
63
+ "epoch": 0.79,
64
+ "learning_rate": 9.84924623115578e-06,
65
+ "loss": 0.1597,
66
+ "step": 200
67
+ },
68
+ {
69
+ "epoch": 0.79,
70
+ "eval_loss": 0.46244511008262634,
71
+ "eval_runtime": 1507.9912,
72
+ "eval_samples_per_second": 5.733,
73
+ "eval_steps_per_second": 0.574,
74
+ "eval_wer": 46.749681665591766,
75
+ "step": 200
76
+ },
77
+ {
78
+ "epoch": 0.88,
79
+ "learning_rate": 9.824120603015075e-06,
80
+ "loss": 0.1509,
81
+ "step": 225
82
+ },
83
+ {
84
+ "epoch": 0.98,
85
+ "learning_rate": 9.798994974874372e-06,
86
+ "loss": 0.1434,
87
+ "step": 250
88
+ },
89
+ {
90
+ "epoch": 1.08,
91
+ "learning_rate": 9.773869346733669e-06,
92
+ "loss": 0.0919,
93
+ "step": 275
94
+ },
95
+ {
96
+ "epoch": 1.18,
97
+ "learning_rate": 9.748743718592965e-06,
98
+ "loss": 0.0776,
99
+ "step": 300
100
+ },
101
+ {
102
+ "epoch": 1.18,
103
+ "eval_loss": 0.4794461131095886,
104
+ "eval_runtime": 1499.4378,
105
+ "eval_samples_per_second": 5.766,
106
+ "eval_steps_per_second": 0.577,
107
+ "eval_wer": 43.17606786724215,
108
+ "step": 300
109
+ }
110
+ ],
111
+ "logging_steps": 25,
112
+ "max_steps": 10000,
113
+ "num_train_epochs": 40,
114
+ "save_steps": 100,
115
+ "total_flos": 4.84563745677312e+18,
116
+ "trial_name": null,
117
+ "trial_params": null
118
+ }
2/checkpoint-300/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c69a6da4c323af31ce3f4997ac3ae843fb847a604926bb5641ee181988d9e27c
3
+ size 4347
2/checkpoint-400/config.json ADDED
@@ -0,0 +1,52 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "makhataei/Whisper-Small-Common-Voice",
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
+ }
2/checkpoint-400/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
+ }
2/checkpoint-400/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3a9e76004a46445ebe6520898f5d4c9e94cd20ef79bc61395ffd6f61010c76d2
3
+ size 966995080
2/checkpoint-400/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:aff2c255e4690a9e1847d896a3b603c41ddcc52e7c06e0e177c7588476d46c9b
3
+ size 1925063607
2/checkpoint-400/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
+ }
2/checkpoint-400/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b4059183c172b67e00937a69636d827a4a9ad9d85c1ed69814be6cc45b436d38
3
+ size 14575
2/checkpoint-400/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9cc33c33ead1987b2ea72bb3035a2a996d683ea38f67d49c1eef7a83c9085a28
3
+ size 627
2/checkpoint-400/trainer_state.json ADDED
@@ -0,0 +1,151 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": 43.17606786724215,
3
+ "best_model_checkpoint": "/media/makhataei/Backups/Whisper-Small-Common-Voice/checkpoint-300",
4
+ "epoch": 1.5701668302257115,
5
+ "eval_steps": 100,
6
+ "global_step": 400,
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": 5e-06,
14
+ "loss": 0.3008,
15
+ "step": 25
16
+ },
17
+ {
18
+ "epoch": 0.2,
19
+ "learning_rate": 1e-05,
20
+ "loss": 0.2368,
21
+ "step": 50
22
+ },
23
+ {
24
+ "epoch": 0.29,
25
+ "learning_rate": 9.974874371859297e-06,
26
+ "loss": 0.2001,
27
+ "step": 75
28
+ },
29
+ {
30
+ "epoch": 0.39,
31
+ "learning_rate": 9.949748743718594e-06,
32
+ "loss": 0.1801,
33
+ "step": 100
34
+ },
35
+ {
36
+ "epoch": 0.39,
37
+ "eval_loss": 0.4975946843624115,
38
+ "eval_runtime": 1546.7344,
39
+ "eval_samples_per_second": 5.59,
40
+ "eval_steps_per_second": 0.559,
41
+ "eval_wer": 49.12602735195383,
42
+ "step": 100
43
+ },
44
+ {
45
+ "epoch": 0.49,
46
+ "learning_rate": 9.92462311557789e-06,
47
+ "loss": 0.1713,
48
+ "step": 125
49
+ },
50
+ {
51
+ "epoch": 0.59,
52
+ "learning_rate": 9.899497487437186e-06,
53
+ "loss": 0.1657,
54
+ "step": 150
55
+ },
56
+ {
57
+ "epoch": 0.69,
58
+ "learning_rate": 9.874371859296484e-06,
59
+ "loss": 0.1574,
60
+ "step": 175
61
+ },
62
+ {
63
+ "epoch": 0.79,
64
+ "learning_rate": 9.84924623115578e-06,
65
+ "loss": 0.1597,
66
+ "step": 200
67
+ },
68
+ {
69
+ "epoch": 0.79,
70
+ "eval_loss": 0.46244511008262634,
71
+ "eval_runtime": 1507.9912,
72
+ "eval_samples_per_second": 5.733,
73
+ "eval_steps_per_second": 0.574,
74
+ "eval_wer": 46.749681665591766,
75
+ "step": 200
76
+ },
77
+ {
78
+ "epoch": 0.88,
79
+ "learning_rate": 9.824120603015075e-06,
80
+ "loss": 0.1509,
81
+ "step": 225
82
+ },
83
+ {
84
+ "epoch": 0.98,
85
+ "learning_rate": 9.798994974874372e-06,
86
+ "loss": 0.1434,
87
+ "step": 250
88
+ },
89
+ {
90
+ "epoch": 1.08,
91
+ "learning_rate": 9.773869346733669e-06,
92
+ "loss": 0.0919,
93
+ "step": 275
94
+ },
95
+ {
96
+ "epoch": 1.18,
97
+ "learning_rate": 9.748743718592965e-06,
98
+ "loss": 0.0776,
99
+ "step": 300
100
+ },
101
+ {
102
+ "epoch": 1.18,
103
+ "eval_loss": 0.4794461131095886,
104
+ "eval_runtime": 1499.4378,
105
+ "eval_samples_per_second": 5.766,
106
+ "eval_steps_per_second": 0.577,
107
+ "eval_wer": 43.17606786724215,
108
+ "step": 300
109
+ },
110
+ {
111
+ "epoch": 1.28,
112
+ "learning_rate": 9.723618090452262e-06,
113
+ "loss": 0.0772,
114
+ "step": 325
115
+ },
116
+ {
117
+ "epoch": 1.37,
118
+ "learning_rate": 9.698492462311559e-06,
119
+ "loss": 0.0799,
120
+ "step": 350
121
+ },
122
+ {
123
+ "epoch": 1.47,
124
+ "learning_rate": 9.673366834170855e-06,
125
+ "loss": 0.0872,
126
+ "step": 375
127
+ },
128
+ {
129
+ "epoch": 1.57,
130
+ "learning_rate": 9.648241206030152e-06,
131
+ "loss": 0.083,
132
+ "step": 400
133
+ },
134
+ {
135
+ "epoch": 1.57,
136
+ "eval_loss": 0.48234912753105164,
137
+ "eval_runtime": 1500.1085,
138
+ "eval_samples_per_second": 5.764,
139
+ "eval_steps_per_second": 0.577,
140
+ "eval_wer": 43.802814572274315,
141
+ "step": 400
142
+ }
143
+ ],
144
+ "logging_steps": 25,
145
+ "max_steps": 10000,
146
+ "num_train_epochs": 40,
147
+ "save_steps": 100,
148
+ "total_flos": 6.46171569856512e+18,
149
+ "trial_name": null,
150
+ "trial_params": null
151
+ }
2/checkpoint-400/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c69a6da4c323af31ce3f4997ac3ae843fb847a604926bb5641ee181988d9e27c
3
+ size 4347
2/checkpoint-9900/config.json ADDED
@@ -0,0 +1,52 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "makhataei/Whisper-Small-Common-Voice",
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
+ }
2/checkpoint-9900/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
+ }
2/checkpoint-9900/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a3d949893127c0271f87b4e8cbc127a9a3c96f60f6fcda37d972429ca2c7032b
3
+ size 966995080
2/checkpoint-9900/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9090ffefe52d1d2b7644937ba3bb208ba29efb93c9b078b455f9a2ab70baab42
3
+ size 1925063607
2/checkpoint-9900/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
+ }