arampacha commited on
Commit
d6eb9bc
1 Parent(s): 2d9955e
README.md CHANGED
@@ -2,7 +2,7 @@
2
  license: apache-2.0
3
  tags:
4
  - automatic-speech-recognition
5
- - /workspace/data/hy/noizy_student_2/
6
  - generated_from_trainer
7
  model-index:
8
  - name: ''
@@ -14,11 +14,11 @@ should probably proofread and complete it, then remove this comment. -->
14
 
15
  #
16
 
17
- This model is a fine-tuned version of [facebook/wav2vec2-xls-r-1b](https://huggingface.co/facebook/wav2vec2-xls-r-1b) on the /WORKSPACE/DATA/HY/NOIZY_STUDENT_2/ - NA dataset.
18
  It achieves the following results on the evaluation set:
19
- - Loss: 0.2249
20
- - Wer: 0.2783
21
- - Cer: 0.0508
22
 
23
  ## Model description
24
 
@@ -46,29 +46,29 @@ The following hyperparameters were used during training:
46
  - optimizer: Adam with betas=(0.9,0.98) and epsilon=1e-08
47
  - lr_scheduler_type: cosine
48
  - lr_scheduler_warmup_ratio: 0.1
49
- - training_steps: 1600
50
  - mixed_precision_training: Native AMP
51
 
52
  ### Training results
53
 
54
  | Training Loss | Epoch | Step | Validation Loss | Wer | Cer |
55
  |:-------------:|:-----:|:----:|:---------------:|:------:|:------:|
56
- | 4.9923 | 3.84 | 100 | 3.1562 | 1.0 | 1.0 |
57
- | 2.1775 | 7.69 | 200 | 0.4334 | 0.5804 | 0.1122 |
58
- | 1.3708 | 11.53 | 300 | 0.3106 | 0.4336 | 0.0797 |
59
- | 1.2266 | 15.38 | 400 | 0.2675 | 0.3673 | 0.0673 |
60
- | 1.093 | 19.23 | 500 | 0.2416 | 0.3501 | 0.0633 |
61
- | 0.989 | 23.08 | 600 | 0.2320 | 0.3251 | 0.0611 |
62
- | 0.9518 | 26.91 | 700 | 0.2413 | 0.3193 | 0.0584 |
63
- | 0.9075 | 30.76 | 800 | 0.2354 | 0.3201 | 0.0593 |
64
- | 0.878 | 34.61 | 900 | 0.2278 | 0.3126 | 0.0579 |
65
- | 0.8563 | 38.46 | 1000 | 0.2327 | 0.2963 | 0.0548 |
66
- | 0.8084 | 42.3 | 1100 | 0.2271 | 0.2923 | 0.0541 |
67
- | 0.7845 | 46.15 | 1200 | 0.2333 | 0.2951 | 0.0537 |
68
- | 0.7487 | 49.99 | 1300 | 0.2290 | 0.2888 | 0.0525 |
69
- | 0.7182 | 53.84 | 1400 | 0.2341 | 0.2877 | 0.0535 |
70
- | 0.7095 | 57.69 | 1500 | 0.2291 | 0.2818 | 0.0515 |
71
- | 0.6953 | 61.53 | 1600 | 0.2249 | 0.2783 | 0.0508 |
72
 
73
 
74
  ### Framework versions
 
2
  license: apache-2.0
3
  tags:
4
  - automatic-speech-recognition
5
+ - /workspace/data/hy/noizy_student_3/
6
  - generated_from_trainer
7
  model-index:
8
  - name: ''
 
14
 
15
  #
16
 
17
+ This model is a fine-tuned version of [facebook/wav2vec2-xls-r-1b](https://huggingface.co/facebook/wav2vec2-xls-r-1b) on the /WORKSPACE/DATA/HY/NOIZY_STUDENT_3/ - NA dataset.
18
  It achieves the following results on the evaluation set:
19
+ - Loss: 0.1827
20
+ - Wer: 0.2389
21
+ - Cer: 0.0427
22
 
23
  ## Model description
24
 
 
46
  - optimizer: Adam with betas=(0.9,0.98) and epsilon=1e-08
47
  - lr_scheduler_type: cosine
48
  - lr_scheduler_warmup_ratio: 0.1
49
+ - training_steps: 3200
50
  - mixed_precision_training: Native AMP
51
 
52
  ### Training results
53
 
54
  | Training Loss | Epoch | Step | Validation Loss | Wer | Cer |
55
  |:-------------:|:-----:|:----:|:---------------:|:------:|:------:|
56
+ | 4.0311 | 3.51 | 200 | 0.7943 | 0.8981 | 0.2374 |
57
+ | 1.4388 | 7.02 | 400 | 0.2546 | 0.3821 | 0.0658 |
58
+ | 1.0949 | 10.53 | 600 | 0.2201 | 0.3216 | 0.0573 |
59
+ | 1.0279 | 14.04 | 800 | 0.2250 | 0.3271 | 0.0583 |
60
+ | 0.9923 | 17.54 | 1000 | 0.2074 | 0.3111 | 0.0543 |
61
+ | 0.972 | 21.05 | 1200 | 0.2165 | 0.2955 | 0.0536 |
62
+ | 0.9587 | 24.56 | 1400 | 0.2064 | 0.3017 | 0.0535 |
63
+ | 0.9421 | 28.07 | 1600 | 0.2062 | 0.2884 | 0.0519 |
64
+ | 0.9189 | 31.58 | 1800 | 0.2014 | 0.2822 | 0.0507 |
65
+ | 0.8919 | 35.09 | 2000 | 0.1952 | 0.2689 | 0.0488 |
66
+ | 0.8615 | 38.6 | 2200 | 0.2020 | 0.2685 | 0.0480 |
67
+ | 0.834 | 42.11 | 2400 | 0.2001 | 0.2654 | 0.0467 |
68
+ | 0.8056 | 45.61 | 2600 | 0.1935 | 0.2498 | 0.0448 |
69
+ | 0.7888 | 49.12 | 2800 | 0.1892 | 0.2451 | 0.0446 |
70
+ | 0.761 | 52.63 | 3000 | 0.1884 | 0.2432 | 0.0441 |
71
+ | 0.742 | 56.14 | 3200 | 0.1827 | 0.2389 | 0.0427 |
72
 
73
 
74
  ### Framework versions
all_results.json CHANGED
@@ -1,15 +1,15 @@
1
  {
2
- "epoch": 61.53,
3
- "eval_cer": 0.05078401618614062,
4
- "eval_loss": 0.2248678207397461,
5
- "eval_runtime": 15.1655,
6
  "eval_samples": 335,
7
- "eval_samples_per_second": 22.09,
8
- "eval_steps_per_second": 0.396,
9
- "eval_wer": 0.2782982045277127,
10
- "train_loss": 1.2442097234725953,
11
- "train_runtime": 17605.3989,
12
- "train_samples": 3354,
13
- "train_samples_per_second": 11.633,
14
- "train_steps_per_second": 0.091
15
  }
 
1
  {
2
+ "epoch": 56.14,
3
+ "eval_cer": 0.04274152756702074,
4
+ "eval_loss": 0.18266724050045013,
5
+ "eval_runtime": 15.3733,
6
  "eval_samples": 335,
7
+ "eval_samples_per_second": 21.791,
8
+ "eval_steps_per_second": 0.39,
9
+ "eval_wer": 0.2388758782201405,
10
+ "train_loss": 1.1288447761535645,
11
+ "train_runtime": 34425.8492,
12
+ "train_samples": 7284,
13
+ "train_samples_per_second": 11.898,
14
+ "train_steps_per_second": 0.093
15
  }
eval_results.json CHANGED
@@ -1,10 +1,10 @@
1
  {
2
- "epoch": 61.53,
3
- "eval_cer": 0.05078401618614062,
4
- "eval_loss": 0.2248678207397461,
5
- "eval_runtime": 15.1655,
6
  "eval_samples": 335,
7
- "eval_samples_per_second": 22.09,
8
- "eval_steps_per_second": 0.396,
9
- "eval_wer": 0.2782982045277127
10
  }
 
1
  {
2
+ "epoch": 56.14,
3
+ "eval_cer": 0.04274152756702074,
4
+ "eval_loss": 0.18266724050045013,
5
+ "eval_runtime": 15.3733,
6
  "eval_samples": 335,
7
+ "eval_samples_per_second": 21.791,
8
+ "eval_steps_per_second": 0.39,
9
+ "eval_wer": 0.2388758782201405
10
  }
mozilla-foundation_common_voice_8_0_hy-AM_test_eval_results.txt CHANGED
@@ -1,2 +1,2 @@
1
- WER: 0.12724434035909446
2
- CER: 0.02716236722306525
 
1
+ WER: 0.1092896174863388
2
+ CER: 0.023773394031360646
pytorch_model.bin CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:dbfcbb3e2a2fa0ab89db84e59692c3f3bee66d7cee3154a41e036d1300fac796
3
  size 3850538161
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5841178926d310ff9b34929b7a12bd2a6884370660d7b8a36e5e68cad7b398b5
3
  size 3850538161
runs/Feb01_02-28-31_job-b1f4681b-d20d-47f2-af64-0c1734f4ff64/1643682600.7320218/events.out.tfevents.1643682600.job-b1f4681b-d20d-47f2-af64-0c1734f4ff64.56475.1 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9d7efa6feb747dd0cd32ce4ec27d06e70776b4b6973a356aa5d2e59ce98a0203
3
+ size 4772
runs/Feb01_02-28-31_job-b1f4681b-d20d-47f2-af64-0c1734f4ff64/events.out.tfevents.1643682600.job-b1f4681b-d20d-47f2-af64-0c1734f4ff64.56475.0 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:372caebe4aa87fa7bef8ad8a172467ae7445d0e019194704a802e54229f8c8b4
3
+ size 13409
runs/Feb01_02-28-31_job-b1f4681b-d20d-47f2-af64-0c1734f4ff64/events.out.tfevents.1643717046.job-b1f4681b-d20d-47f2-af64-0c1734f4ff64.56475.2 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:16cd520a9802ba9a4bae9e6a4bc6c34e2693a822e05be353a4545c592e412324
3
+ size 405
speech-recognition-community-v2_dev_data_hy_validation_eval_results.txt ADDED
@@ -0,0 +1,2 @@
 
 
 
1
+ WER: 0.19942816297355254
2
+ CER: 0.07332618465282714
train_results.json CHANGED
@@ -1,8 +1,8 @@
1
  {
2
- "epoch": 61.53,
3
- "train_loss": 1.2442097234725953,
4
- "train_runtime": 17605.3989,
5
- "train_samples": 3354,
6
- "train_samples_per_second": 11.633,
7
- "train_steps_per_second": 0.091
8
  }
 
1
  {
2
+ "epoch": 56.14,
3
+ "train_loss": 1.1288447761535645,
4
+ "train_runtime": 34425.8492,
5
+ "train_samples": 7284,
6
+ "train_samples_per_second": 11.898,
7
+ "train_steps_per_second": 0.093
8
  }
trainer_state.json CHANGED
@@ -1,281 +1,281 @@
1
  {
2
- "best_metric": 0.2248678207397461,
3
- "best_model_checkpoint": "./checkpoint-1600",
4
- "epoch": 61.53333333333333,
5
- "global_step": 1600,
6
  "is_hyper_param_search": false,
7
  "is_local_process_zero": true,
8
  "is_world_process_zero": true,
9
  "log_history": [
10
  {
11
- "epoch": 3.84,
12
  "learning_rate": 4.980500000000001e-05,
13
- "loss": 4.9923,
14
- "step": 100
15
  },
16
  {
17
- "epoch": 3.84,
18
- "eval_cer": 1.0,
19
- "eval_loss": 3.156188488006592,
20
- "eval_runtime": 15.7049,
21
- "eval_samples_per_second": 21.331,
22
  "eval_steps_per_second": 0.382,
23
- "eval_wer": 1.0,
24
- "step": 100
25
  },
26
  {
27
- "epoch": 7.69,
28
  "learning_rate": 8e-05,
29
- "loss": 2.1775,
30
- "step": 200
31
  },
32
  {
33
- "epoch": 7.69,
34
- "eval_cer": 0.11219018715225089,
35
- "eval_loss": 0.43336454033851624,
36
- "eval_runtime": 15.2789,
37
- "eval_samples_per_second": 21.926,
38
- "eval_steps_per_second": 0.393,
39
- "eval_wer": 0.5804059328649492,
40
- "step": 200
41
  },
42
  {
43
- "epoch": 11.53,
44
  "learning_rate": 8e-05,
45
- "loss": 1.3708,
46
- "step": 300
47
  },
48
  {
49
- "epoch": 11.53,
50
- "eval_cer": 0.07966616084977238,
51
- "eval_loss": 0.3105751574039459,
52
- "eval_runtime": 15.048,
53
- "eval_samples_per_second": 22.262,
54
- "eval_steps_per_second": 0.399,
55
- "eval_wer": 0.4336455893832943,
56
- "step": 300
57
  },
58
  {
59
- "epoch": 15.38,
60
  "learning_rate": 8e-05,
61
- "loss": 1.2266,
62
- "step": 400
63
  },
64
  {
65
- "epoch": 15.38,
66
- "eval_cer": 0.06732422862923622,
67
- "eval_loss": 0.26751142740249634,
68
- "eval_runtime": 15.232,
69
- "eval_samples_per_second": 21.993,
70
- "eval_steps_per_second": 0.394,
71
- "eval_wer": 0.3672911787665886,
72
- "step": 400
73
  },
74
  {
75
- "epoch": 19.23,
76
  "learning_rate": 8e-05,
77
- "loss": 1.093,
78
- "step": 500
79
  },
80
  {
81
- "epoch": 19.23,
82
- "eval_cer": 0.06327769347496207,
83
- "eval_loss": 0.24162611365318298,
84
- "eval_runtime": 14.995,
85
- "eval_samples_per_second": 22.341,
86
- "eval_steps_per_second": 0.4,
87
- "eval_wer": 0.35011709601873536,
88
- "step": 500
89
  },
90
  {
91
- "epoch": 23.08,
92
  "learning_rate": 8e-05,
93
- "loss": 0.989,
94
- "step": 600
95
  },
96
  {
97
- "epoch": 23.08,
98
- "eval_cer": 0.06105209914011128,
99
- "eval_loss": 0.23200440406799316,
100
- "eval_runtime": 15.3525,
101
- "eval_samples_per_second": 21.821,
102
- "eval_steps_per_second": 0.391,
103
- "eval_wer": 0.3251366120218579,
104
- "step": 600
105
  },
106
  {
107
- "epoch": 26.91,
108
  "learning_rate": 8e-05,
109
- "loss": 0.9518,
110
- "step": 700
111
  },
112
  {
113
- "epoch": 26.91,
114
- "eval_cer": 0.05842185128983308,
115
- "eval_loss": 0.2413272261619568,
116
- "eval_runtime": 15.2385,
117
- "eval_samples_per_second": 21.984,
118
- "eval_steps_per_second": 0.394,
119
- "eval_wer": 0.3192818110850898,
120
- "step": 700
121
  },
122
  {
123
- "epoch": 30.76,
124
  "learning_rate": 8e-05,
125
- "loss": 0.9075,
126
- "step": 800
127
  },
128
  {
129
- "epoch": 30.76,
130
- "eval_cer": 0.05933232169954476,
131
- "eval_loss": 0.23544833064079285,
132
- "eval_runtime": 15.1938,
133
- "eval_samples_per_second": 22.049,
134
- "eval_steps_per_second": 0.395,
135
- "eval_wer": 0.3200624512099922,
136
- "step": 800
137
  },
138
  {
139
- "epoch": 34.61,
140
  "learning_rate": 7.059500000000001e-05,
141
- "loss": 0.878,
142
- "step": 900
143
  },
144
  {
145
- "epoch": 34.61,
146
- "eval_cer": 0.057916034395548814,
147
- "eval_loss": 0.22777308523654938,
148
- "eval_runtime": 14.9728,
149
- "eval_samples_per_second": 22.374,
150
- "eval_steps_per_second": 0.401,
151
- "eval_wer": 0.3126463700234192,
152
- "step": 900
153
  },
154
  {
155
- "epoch": 38.46,
156
  "learning_rate": 6.109500000000001e-05,
157
- "loss": 0.8563,
158
- "step": 1000
159
  },
160
  {
161
- "epoch": 38.46,
162
- "eval_cer": 0.054779969650986346,
163
- "eval_loss": 0.2326740324497223,
164
- "eval_runtime": 15.1749,
165
- "eval_samples_per_second": 22.076,
166
- "eval_steps_per_second": 0.395,
167
- "eval_wer": 0.2962529274004684,
168
- "step": 1000
169
  },
170
  {
171
- "epoch": 42.3,
172
- "learning_rate": 5.169000000000001e-05,
173
- "loss": 0.8084,
174
- "step": 1100
175
  },
176
  {
177
- "epoch": 42.3,
178
- "eval_cer": 0.05407182599898837,
179
- "eval_loss": 0.22712552547454834,
180
- "eval_runtime": 15.3083,
181
- "eval_samples_per_second": 21.884,
182
- "eval_steps_per_second": 0.392,
183
- "eval_wer": 0.2923497267759563,
184
- "step": 1100
185
  },
186
  {
187
- "epoch": 46.15,
188
- "learning_rate": 4.219000000000001e-05,
189
- "loss": 0.7845,
190
- "step": 1200
191
  },
192
  {
193
- "epoch": 46.15,
194
- "eval_cer": 0.053667172483560954,
195
- "eval_loss": 0.23326420783996582,
196
- "eval_runtime": 15.1559,
197
- "eval_samples_per_second": 22.104,
198
- "eval_steps_per_second": 0.396,
199
- "eval_wer": 0.29508196721311475,
200
- "step": 1200
201
  },
202
  {
203
- "epoch": 49.99,
204
- "learning_rate": 3.269000000000001e-05,
205
- "loss": 0.7487,
206
- "step": 1300
207
  },
208
  {
209
- "epoch": 49.99,
210
- "eval_cer": 0.052453211937278706,
211
- "eval_loss": 0.22895006835460663,
212
- "eval_runtime": 15.373,
213
- "eval_samples_per_second": 21.791,
214
- "eval_steps_per_second": 0.39,
215
- "eval_wer": 0.2888368462138954,
216
- "step": 1300
217
  },
218
  {
219
- "epoch": 53.84,
220
- "learning_rate": 2.319e-05,
221
- "loss": 0.7182,
222
- "step": 1400
223
  },
224
  {
225
- "epoch": 53.84,
226
- "eval_cer": 0.05346484572584724,
227
- "eval_loss": 0.23406584560871124,
228
- "eval_runtime": 15.2056,
229
- "eval_samples_per_second": 22.031,
230
- "eval_steps_per_second": 0.395,
231
- "eval_wer": 0.28766588602654175,
232
- "step": 1400
233
  },
234
  {
235
- "epoch": 57.69,
236
- "learning_rate": 1.369e-05,
237
- "loss": 0.7095,
238
- "step": 1500
239
  },
240
  {
241
- "epoch": 57.69,
242
- "eval_cer": 0.05154274152756702,
243
- "eval_loss": 0.22908572852611542,
244
- "eval_runtime": 15.2684,
245
- "eval_samples_per_second": 21.941,
246
- "eval_steps_per_second": 0.393,
247
- "eval_wer": 0.2818110850897736,
248
- "step": 1500
249
  },
250
  {
251
- "epoch": 61.53,
252
- "learning_rate": 4.190000000000005e-06,
253
- "loss": 0.6953,
254
- "step": 1600
255
  },
256
  {
257
- "epoch": 61.53,
258
- "eval_cer": 0.05078401618614062,
259
- "eval_loss": 0.2248678207397461,
260
- "eval_runtime": 15.2139,
261
- "eval_samples_per_second": 22.019,
262
- "eval_steps_per_second": 0.394,
263
- "eval_wer": 0.2782982045277127,
264
- "step": 1600
265
  },
266
  {
267
- "epoch": 61.53,
268
- "step": 1600,
269
- "total_flos": 1.3126730002882698e+20,
270
- "train_loss": 1.2442097234725953,
271
- "train_runtime": 17605.3989,
272
- "train_samples_per_second": 11.633,
273
- "train_steps_per_second": 0.091
274
  }
275
  ],
276
- "max_steps": 1600,
277
- "num_train_epochs": 62,
278
- "total_flos": 1.3126730002882698e+20,
279
  "trial_name": null,
280
  "trial_params": null
281
  }
 
1
  {
2
+ "best_metric": 0.18266724050045013,
3
+ "best_model_checkpoint": "./checkpoint-3200",
4
+ "epoch": 56.14035087719298,
5
+ "global_step": 3200,
6
  "is_hyper_param_search": false,
7
  "is_local_process_zero": true,
8
  "is_world_process_zero": true,
9
  "log_history": [
10
  {
11
+ "epoch": 3.51,
12
  "learning_rate": 4.980500000000001e-05,
13
+ "loss": 4.0311,
14
+ "step": 200
15
  },
16
  {
17
+ "epoch": 3.51,
18
+ "eval_cer": 0.23737986848760748,
19
+ "eval_loss": 0.794314444065094,
20
+ "eval_runtime": 15.7,
21
+ "eval_samples_per_second": 21.338,
22
  "eval_steps_per_second": 0.382,
23
+ "eval_wer": 0.8981264637002342,
24
+ "step": 200
25
  },
26
  {
27
+ "epoch": 7.02,
28
  "learning_rate": 8e-05,
29
+ "loss": 1.4388,
30
+ "step": 400
31
  },
32
  {
33
+ "epoch": 7.02,
34
+ "eval_cer": 0.06575619625695499,
35
+ "eval_loss": 0.2545942962169647,
36
+ "eval_runtime": 14.891,
37
+ "eval_samples_per_second": 22.497,
38
+ "eval_steps_per_second": 0.403,
39
+ "eval_wer": 0.38212334113973456,
40
+ "step": 400
41
  },
42
  {
43
+ "epoch": 10.53,
44
  "learning_rate": 8e-05,
45
+ "loss": 1.0949,
46
+ "step": 600
47
  },
48
  {
49
+ "epoch": 10.53,
50
+ "eval_cer": 0.057309054122407686,
51
+ "eval_loss": 0.22006691992282867,
52
+ "eval_runtime": 15.2331,
53
+ "eval_samples_per_second": 21.992,
54
+ "eval_steps_per_second": 0.394,
55
+ "eval_wer": 0.32162373145979706,
56
+ "step": 600
57
  },
58
  {
59
+ "epoch": 14.04,
60
  "learning_rate": 8e-05,
61
+ "loss": 1.0279,
62
+ "step": 800
63
  },
64
  {
65
+ "epoch": 14.04,
66
+ "eval_cer": 0.0582701062215478,
67
+ "eval_loss": 0.22504645586013794,
68
+ "eval_runtime": 15.1767,
69
+ "eval_samples_per_second": 22.073,
70
+ "eval_steps_per_second": 0.395,
71
+ "eval_wer": 0.32708821233411395,
72
+ "step": 800
73
  },
74
  {
75
+ "epoch": 17.54,
76
  "learning_rate": 8e-05,
77
+ "loss": 0.9923,
78
+ "step": 1000
79
  },
80
  {
81
+ "epoch": 17.54,
82
+ "eval_cer": 0.0543247344461305,
83
+ "eval_loss": 0.2073642760515213,
84
+ "eval_runtime": 15.1305,
85
+ "eval_samples_per_second": 22.141,
86
+ "eval_steps_per_second": 0.397,
87
+ "eval_wer": 0.3110850897736144,
88
+ "step": 1000
89
  },
90
  {
91
+ "epoch": 21.05,
92
  "learning_rate": 8e-05,
93
+ "loss": 0.972,
94
+ "step": 1200
95
  },
96
  {
97
+ "epoch": 21.05,
98
+ "eval_cer": 0.05361659079413252,
99
+ "eval_loss": 0.21649114787578583,
100
+ "eval_runtime": 14.9827,
101
+ "eval_samples_per_second": 22.359,
102
+ "eval_steps_per_second": 0.4,
103
+ "eval_wer": 0.29547228727556596,
104
+ "step": 1200
105
  },
106
  {
107
+ "epoch": 24.56,
108
  "learning_rate": 8e-05,
109
+ "loss": 0.9587,
110
+ "step": 1400
111
  },
112
  {
113
+ "epoch": 24.56,
114
+ "eval_cer": 0.05351542741527567,
115
+ "eval_loss": 0.2064175009727478,
116
+ "eval_runtime": 15.1921,
117
+ "eval_samples_per_second": 22.051,
118
+ "eval_steps_per_second": 0.395,
119
+ "eval_wer": 0.3017174082747853,
120
+ "step": 1400
121
  },
122
  {
123
+ "epoch": 28.07,
124
  "learning_rate": 8e-05,
125
+ "loss": 0.9421,
126
+ "step": 1600
127
  },
128
  {
129
+ "epoch": 28.07,
130
+ "eval_cer": 0.051947395042994435,
131
+ "eval_loss": 0.2061864584684372,
132
+ "eval_runtime": 15.0418,
133
+ "eval_samples_per_second": 22.271,
134
+ "eval_steps_per_second": 0.399,
135
+ "eval_wer": 0.28844652615144417,
136
+ "step": 1600
137
  },
138
  {
139
+ "epoch": 31.58,
140
  "learning_rate": 7.059500000000001e-05,
141
+ "loss": 0.9189,
142
+ "step": 1800
143
  },
144
  {
145
+ "epoch": 31.58,
146
+ "eval_cer": 0.05073343449671219,
147
+ "eval_loss": 0.2014162391424179,
148
+ "eval_runtime": 15.125,
149
+ "eval_samples_per_second": 22.149,
150
+ "eval_steps_per_second": 0.397,
151
+ "eval_wer": 0.2822014051522248,
152
+ "step": 1800
153
  },
154
  {
155
+ "epoch": 35.09,
156
  "learning_rate": 6.109500000000001e-05,
157
+ "loss": 0.8919,
158
+ "step": 2000
159
  },
160
  {
161
+ "epoch": 35.09,
162
+ "eval_cer": 0.04881133029843197,
163
+ "eval_loss": 0.19518214464187622,
164
+ "eval_runtime": 15.0854,
165
+ "eval_samples_per_second": 22.207,
166
+ "eval_steps_per_second": 0.398,
167
+ "eval_wer": 0.2689305230288837,
168
+ "step": 2000
169
  },
170
  {
171
+ "epoch": 38.6,
172
+ "learning_rate": 5.1594999999999996e-05,
173
+ "loss": 0.8615,
174
+ "step": 2200
175
  },
176
  {
177
+ "epoch": 38.6,
178
+ "eval_cer": 0.04795144157814871,
179
+ "eval_loss": 0.20196911692619324,
180
+ "eval_runtime": 15.1604,
181
+ "eval_samples_per_second": 22.097,
182
+ "eval_steps_per_second": 0.396,
183
+ "eval_wer": 0.2685402029664325,
184
+ "step": 2200
185
  },
186
  {
187
+ "epoch": 42.11,
188
+ "learning_rate": 4.2095e-05,
189
+ "loss": 0.834,
190
+ "step": 2400
191
  },
192
  {
193
+ "epoch": 42.11,
194
+ "eval_cer": 0.04668689934243804,
195
+ "eval_loss": 0.2001034915447235,
196
+ "eval_runtime": 15.091,
197
+ "eval_samples_per_second": 22.199,
198
+ "eval_steps_per_second": 0.398,
199
+ "eval_wer": 0.2654176424668228,
200
+ "step": 2400
201
  },
202
  {
203
+ "epoch": 45.61,
204
+ "learning_rate": 3.2595e-05,
205
+ "loss": 0.8056,
206
+ "step": 2600
207
  },
208
  {
209
+ "epoch": 45.61,
210
+ "eval_cer": 0.04481537683358624,
211
+ "eval_loss": 0.1934908777475357,
212
+ "eval_runtime": 15.4173,
213
+ "eval_samples_per_second": 21.729,
214
+ "eval_steps_per_second": 0.389,
215
+ "eval_wer": 0.2498048399687744,
216
+ "step": 2600
217
  },
218
  {
219
+ "epoch": 49.12,
220
+ "learning_rate": 2.3095e-05,
221
+ "loss": 0.7888,
222
+ "step": 2800
223
  },
224
  {
225
+ "epoch": 49.12,
226
+ "eval_cer": 0.04461305007587253,
227
+ "eval_loss": 0.18915079534053802,
228
+ "eval_runtime": 15.29,
229
+ "eval_samples_per_second": 21.91,
230
+ "eval_steps_per_second": 0.392,
231
+ "eval_wer": 0.24512099921935987,
232
+ "step": 2800
233
  },
234
  {
235
+ "epoch": 52.63,
236
+ "learning_rate": 1.3595000000000008e-05,
237
+ "loss": 0.761,
238
+ "step": 3000
239
  },
240
  {
241
+ "epoch": 52.63,
242
+ "eval_cer": 0.044056651492159836,
243
+ "eval_loss": 0.18836112320423126,
244
+ "eval_runtime": 16.204,
245
+ "eval_samples_per_second": 20.674,
246
+ "eval_steps_per_second": 0.37,
247
+ "eval_wer": 0.24316939890710382,
248
+ "step": 3000
249
  },
250
  {
251
+ "epoch": 56.14,
252
+ "learning_rate": 4.095000000000005e-06,
253
+ "loss": 0.742,
254
+ "step": 3200
255
  },
256
  {
257
+ "epoch": 56.14,
258
+ "eval_cer": 0.04274152756702074,
259
+ "eval_loss": 0.18266724050045013,
260
+ "eval_runtime": 15.3463,
261
+ "eval_samples_per_second": 21.829,
262
+ "eval_steps_per_second": 0.391,
263
+ "eval_wer": 0.2388758782201405,
264
+ "step": 3200
265
  },
266
  {
267
+ "epoch": 56.14,
268
+ "step": 3200,
269
+ "total_flos": 2.6268075931237872e+20,
270
+ "train_loss": 1.1288447761535645,
271
+ "train_runtime": 34425.8492,
272
+ "train_samples_per_second": 11.898,
273
+ "train_steps_per_second": 0.093
274
  }
275
  ],
276
+ "max_steps": 3200,
277
+ "num_train_epochs": 57,
278
+ "total_flos": 2.6268075931237872e+20,
279
  "trial_name": null,
280
  "trial_params": null
281
  }
training_args.bin CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:c934f0ed6fa80bfe4f2228c9550d942f9d713597358d2f57a1375b6454c2d03d
3
  size 3055
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bfeb12863ac5de618f2b958c1b03e980705dda909c6bff76a929e0e5bfb2b372
3
  size 3055