utakumi commited on
Commit
8565ad4
1 Parent(s): 5559bc3

End of training

Browse files
Files changed (5) hide show
  1. README.md +8 -4
  2. all_results.json +16 -0
  3. eval_results.json +10 -0
  4. train_results.json +9 -0
  5. trainer_state.json +1212 -0
README.md CHANGED
@@ -1,8 +1,12 @@
1
  ---
2
  library_name: transformers
 
 
3
  license: apache-2.0
4
  base_model: rinna/japanese-hubert-base
5
  tags:
 
 
6
  - generated_from_trainer
7
  metrics:
8
  - wer
@@ -16,11 +20,11 @@ should probably proofread and complete it, then remove this comment. -->
16
 
17
  # Hubert-common_voice_JSUT-ja-demo-japanese
18
 
19
- This model is a fine-tuned version of [rinna/japanese-hubert-base](https://huggingface.co/rinna/japanese-hubert-base) on the None dataset.
20
  It achieves the following results on the evaluation set:
21
- - Loss: 2.1762
22
- - Wer: 1.9914
23
- - Cer: 0.6413
24
 
25
  ## Model description
26
 
 
1
  ---
2
  library_name: transformers
3
+ language:
4
+ - ja
5
  license: apache-2.0
6
  base_model: rinna/japanese-hubert-base
7
  tags:
8
+ - automatic-speech-recognition
9
+ - mozilla-foundation/common_voice_13_0
10
  - generated_from_trainer
11
  metrics:
12
  - wer
 
20
 
21
  # Hubert-common_voice_JSUT-ja-demo-japanese
22
 
23
+ This model is a fine-tuned version of [rinna/japanese-hubert-base](https://huggingface.co/rinna/japanese-hubert-base) on the MOZILLA-FOUNDATION/COMMON_VOICE_13_0 - JA dataset.
24
  It achieves the following results on the evaluation set:
25
+ - Loss: 2.1568
26
+ - Wer: 1.9920
27
+ - Cer: 0.6415
28
 
29
  ## Model description
30
 
all_results.json ADDED
@@ -0,0 +1,16 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 20.0,
3
+ "eval_cer": 0.6415434821104824,
4
+ "eval_loss": 2.1568174362182617,
5
+ "eval_runtime": 247.3136,
6
+ "eval_samples": 5461,
7
+ "eval_samples_per_second": 22.081,
8
+ "eval_steps_per_second": 2.762,
9
+ "eval_wer": 1.9919502378338823,
10
+ "total_flos": 1.4857862206321902e+19,
11
+ "train_loss": 15.407913101757057,
12
+ "train_runtime": 65838.1292,
13
+ "train_samples": 16532,
14
+ "train_samples_per_second": 5.022,
15
+ "train_steps_per_second": 0.157
16
+ }
eval_results.json ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 20.0,
3
+ "eval_cer": 0.6415434821104824,
4
+ "eval_loss": 2.1568174362182617,
5
+ "eval_runtime": 247.3136,
6
+ "eval_samples": 5461,
7
+ "eval_samples_per_second": 22.081,
8
+ "eval_steps_per_second": 2.762,
9
+ "eval_wer": 1.9919502378338823
10
+ }
train_results.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 20.0,
3
+ "total_flos": 1.4857862206321902e+19,
4
+ "train_loss": 15.407913101757057,
5
+ "train_runtime": 65838.1292,
6
+ "train_samples": 16532,
7
+ "train_samples_per_second": 5.022,
8
+ "train_steps_per_second": 0.157
9
+ }
trainer_state.json ADDED
@@ -0,0 +1,1212 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": 2.176206588745117,
3
+ "best_model_checkpoint": "./Hubert-common_voice_JSUT-ja-demo-japanese/checkpoint-10300",
4
+ "epoch": 20.0,
5
+ "eval_steps": 100,
6
+ "global_step": 10340,
7
+ "is_hyper_param_search": false,
8
+ "is_local_process_zero": true,
9
+ "is_world_process_zero": true,
10
+ "log_history": [
11
+ {
12
+ "epoch": 0.19342359767891681,
13
+ "eval_cer": 8.48504498730911,
14
+ "eval_loss": 84.69584655761719,
15
+ "eval_runtime": 236.8633,
16
+ "eval_samples_per_second": 23.055,
17
+ "eval_steps_per_second": 2.884,
18
+ "eval_wer": 1.0115257958287596,
19
+ "step": 100
20
+ },
21
+ {
22
+ "epoch": 0.38684719535783363,
23
+ "eval_cer": 8.344315240030694,
24
+ "eval_loss": 83.78863525390625,
25
+ "eval_runtime": 232.6848,
26
+ "eval_samples_per_second": 23.47,
27
+ "eval_steps_per_second": 2.935,
28
+ "eval_wer": 1.008964507866813,
29
+ "step": 200
30
+ },
31
+ {
32
+ "epoch": 0.5802707930367504,
33
+ "eval_cer": 4.815741763569977,
34
+ "eval_loss": 81.74565124511719,
35
+ "eval_runtime": 231.6247,
36
+ "eval_samples_per_second": 23.577,
37
+ "eval_steps_per_second": 2.949,
38
+ "eval_wer": 1.000365898280278,
39
+ "step": 300
40
+ },
41
+ {
42
+ "epoch": 0.7736943907156673,
43
+ "eval_cer": 0.9907074011923535,
44
+ "eval_loss": 75.43038940429688,
45
+ "eval_runtime": 227.0361,
46
+ "eval_samples_per_second": 24.053,
47
+ "eval_steps_per_second": 3.008,
48
+ "eval_wer": 1.0,
49
+ "step": 400
50
+ },
51
+ {
52
+ "epoch": 0.9671179883945842,
53
+ "grad_norm": 140.793212890625,
54
+ "learning_rate": 1.188e-06,
55
+ "loss": 66.0277,
56
+ "step": 500
57
+ },
58
+ {
59
+ "epoch": 0.9671179883945842,
60
+ "eval_cer": 0.9907495636188854,
61
+ "eval_loss": 63.1251106262207,
62
+ "eval_runtime": 226.7903,
63
+ "eval_samples_per_second": 24.08,
64
+ "eval_steps_per_second": 3.012,
65
+ "eval_wer": 1.0,
66
+ "step": 500
67
+ },
68
+ {
69
+ "epoch": 1.1605415860735009,
70
+ "eval_cer": 0.9907074011923535,
71
+ "eval_loss": 57.10499572753906,
72
+ "eval_runtime": 255.6303,
73
+ "eval_samples_per_second": 21.363,
74
+ "eval_steps_per_second": 2.672,
75
+ "eval_wer": 1.0,
76
+ "step": 600
77
+ },
78
+ {
79
+ "epoch": 1.3539651837524178,
80
+ "eval_cer": 0.9907917260454173,
81
+ "eval_loss": 55.67994689941406,
82
+ "eval_runtime": 269.8181,
83
+ "eval_samples_per_second": 20.24,
84
+ "eval_steps_per_second": 2.531,
85
+ "eval_wer": 1.0,
86
+ "step": 700
87
+ },
88
+ {
89
+ "epoch": 1.5473887814313345,
90
+ "eval_cer": 0.9907495636188854,
91
+ "eval_loss": 55.04756546020508,
92
+ "eval_runtime": 241.9741,
93
+ "eval_samples_per_second": 22.569,
94
+ "eval_steps_per_second": 2.823,
95
+ "eval_wer": 1.0,
96
+ "step": 800
97
+ },
98
+ {
99
+ "epoch": 1.7408123791102514,
100
+ "eval_cer": 0.9907074011923535,
101
+ "eval_loss": 54.408546447753906,
102
+ "eval_runtime": 256.9455,
103
+ "eval_samples_per_second": 21.254,
104
+ "eval_steps_per_second": 2.658,
105
+ "eval_wer": 1.0,
106
+ "step": 900
107
+ },
108
+ {
109
+ "epoch": 1.9342359767891684,
110
+ "grad_norm": 141.13522338867188,
111
+ "learning_rate": 2.3855999999999997e-06,
112
+ "loss": 46.3141,
113
+ "step": 1000
114
+ },
115
+ {
116
+ "epoch": 1.9342359767891684,
117
+ "eval_cer": 0.9907917260454173,
118
+ "eval_loss": 53.68925857543945,
119
+ "eval_runtime": 260.4063,
120
+ "eval_samples_per_second": 20.971,
121
+ "eval_steps_per_second": 2.623,
122
+ "eval_wer": 1.0,
123
+ "step": 1000
124
+ },
125
+ {
126
+ "epoch": 2.127659574468085,
127
+ "eval_cer": 0.9907074011923535,
128
+ "eval_loss": 52.97110366821289,
129
+ "eval_runtime": 254.697,
130
+ "eval_samples_per_second": 21.441,
131
+ "eval_steps_per_second": 2.682,
132
+ "eval_wer": 1.0,
133
+ "step": 1100
134
+ },
135
+ {
136
+ "epoch": 2.3210831721470018,
137
+ "eval_cer": 0.9907495636188854,
138
+ "eval_loss": 52.13261032104492,
139
+ "eval_runtime": 246.5799,
140
+ "eval_samples_per_second": 22.147,
141
+ "eval_steps_per_second": 2.77,
142
+ "eval_wer": 1.0,
143
+ "step": 1200
144
+ },
145
+ {
146
+ "epoch": 2.514506769825919,
147
+ "eval_cer": 0.9907074011923535,
148
+ "eval_loss": 51.254920959472656,
149
+ "eval_runtime": 260.2756,
150
+ "eval_samples_per_second": 20.982,
151
+ "eval_steps_per_second": 2.624,
152
+ "eval_wer": 1.0,
153
+ "step": 1300
154
+ },
155
+ {
156
+ "epoch": 2.7079303675048356,
157
+ "eval_cer": 0.9907074011923535,
158
+ "eval_loss": 50.26494598388672,
159
+ "eval_runtime": 261.2096,
160
+ "eval_samples_per_second": 20.907,
161
+ "eval_steps_per_second": 2.615,
162
+ "eval_wer": 1.0,
163
+ "step": 1400
164
+ },
165
+ {
166
+ "epoch": 2.9013539651837523,
167
+ "grad_norm": 120.8663101196289,
168
+ "learning_rate": 3.5856e-06,
169
+ "loss": 42.8642,
170
+ "step": 1500
171
+ },
172
+ {
173
+ "epoch": 2.9013539651837523,
174
+ "eval_cer": 0.9907074011923535,
175
+ "eval_loss": 49.20811080932617,
176
+ "eval_runtime": 258.4268,
177
+ "eval_samples_per_second": 21.132,
178
+ "eval_steps_per_second": 2.643,
179
+ "eval_wer": 1.0,
180
+ "step": 1500
181
+ },
182
+ {
183
+ "epoch": 3.094777562862669,
184
+ "eval_cer": 0.9907495636188854,
185
+ "eval_loss": 48.105064392089844,
186
+ "eval_runtime": 247.347,
187
+ "eval_samples_per_second": 22.078,
188
+ "eval_steps_per_second": 2.761,
189
+ "eval_wer": 1.0,
190
+ "step": 1600
191
+ },
192
+ {
193
+ "epoch": 3.288201160541586,
194
+ "eval_cer": 0.9907074011923535,
195
+ "eval_loss": 46.87884521484375,
196
+ "eval_runtime": 262.3223,
197
+ "eval_samples_per_second": 20.818,
198
+ "eval_steps_per_second": 2.604,
199
+ "eval_wer": 1.0,
200
+ "step": 1700
201
+ },
202
+ {
203
+ "epoch": 3.481624758220503,
204
+ "eval_cer": 0.9907074011923535,
205
+ "eval_loss": 45.54108810424805,
206
+ "eval_runtime": 228.7269,
207
+ "eval_samples_per_second": 23.876,
208
+ "eval_steps_per_second": 2.986,
209
+ "eval_wer": 1.0,
210
+ "step": 1800
211
+ },
212
+ {
213
+ "epoch": 3.6750483558994196,
214
+ "eval_cer": 0.9907074011923535,
215
+ "eval_loss": 44.151554107666016,
216
+ "eval_runtime": 228.7844,
217
+ "eval_samples_per_second": 23.87,
218
+ "eval_steps_per_second": 2.985,
219
+ "eval_wer": 1.0,
220
+ "step": 1900
221
+ },
222
+ {
223
+ "epoch": 3.8684719535783367,
224
+ "grad_norm": 172.9249725341797,
225
+ "learning_rate": 4.7856e-06,
226
+ "loss": 38.3378,
227
+ "step": 2000
228
+ },
229
+ {
230
+ "epoch": 3.8684719535783367,
231
+ "eval_cer": 0.9907495636188854,
232
+ "eval_loss": 42.60865783691406,
233
+ "eval_runtime": 227.6693,
234
+ "eval_samples_per_second": 23.987,
235
+ "eval_steps_per_second": 3.0,
236
+ "eval_wer": 1.0,
237
+ "step": 2000
238
+ },
239
+ {
240
+ "epoch": 4.061895551257253,
241
+ "eval_cer": 0.9907074011923535,
242
+ "eval_loss": 40.9814567565918,
243
+ "eval_runtime": 230.7363,
244
+ "eval_samples_per_second": 23.668,
245
+ "eval_steps_per_second": 2.96,
246
+ "eval_wer": 1.0,
247
+ "step": 2100
248
+ },
249
+ {
250
+ "epoch": 4.25531914893617,
251
+ "eval_cer": 0.9907074011923535,
252
+ "eval_loss": 39.240055084228516,
253
+ "eval_runtime": 230.4939,
254
+ "eval_samples_per_second": 23.693,
255
+ "eval_steps_per_second": 2.963,
256
+ "eval_wer": 1.0,
257
+ "step": 2200
258
+ },
259
+ {
260
+ "epoch": 4.448742746615087,
261
+ "eval_cer": 0.9907917260454173,
262
+ "eval_loss": 37.40217590332031,
263
+ "eval_runtime": 228.8585,
264
+ "eval_samples_per_second": 23.862,
265
+ "eval_steps_per_second": 2.984,
266
+ "eval_wer": 1.0,
267
+ "step": 2300
268
+ },
269
+ {
270
+ "epoch": 4.6421663442940035,
271
+ "eval_cer": 0.9907495636188854,
272
+ "eval_loss": 35.430931091308594,
273
+ "eval_runtime": 229.2781,
274
+ "eval_samples_per_second": 23.818,
275
+ "eval_steps_per_second": 2.979,
276
+ "eval_wer": 1.0,
277
+ "step": 2400
278
+ },
279
+ {
280
+ "epoch": 4.835589941972921,
281
+ "grad_norm": 113.81422424316406,
282
+ "learning_rate": 5.9856e-06,
283
+ "loss": 31.9192,
284
+ "step": 2500
285
+ },
286
+ {
287
+ "epoch": 4.835589941972921,
288
+ "eval_cer": 0.9907074011923535,
289
+ "eval_loss": 33.417510986328125,
290
+ "eval_runtime": 233.5128,
291
+ "eval_samples_per_second": 23.386,
292
+ "eval_steps_per_second": 2.925,
293
+ "eval_wer": 1.0,
294
+ "step": 2500
295
+ },
296
+ {
297
+ "epoch": 5.029013539651838,
298
+ "eval_cer": 0.9907495636188854,
299
+ "eval_loss": 31.266008377075195,
300
+ "eval_runtime": 235.6664,
301
+ "eval_samples_per_second": 23.173,
302
+ "eval_steps_per_second": 2.898,
303
+ "eval_wer": 1.0,
304
+ "step": 2600
305
+ },
306
+ {
307
+ "epoch": 5.222437137330754,
308
+ "eval_cer": 0.9907917260454173,
309
+ "eval_loss": 29.01472282409668,
310
+ "eval_runtime": 238.0062,
311
+ "eval_samples_per_second": 22.945,
312
+ "eval_steps_per_second": 2.87,
313
+ "eval_wer": 1.0,
314
+ "step": 2700
315
+ },
316
+ {
317
+ "epoch": 5.415860735009671,
318
+ "eval_cer": 0.9907495636188854,
319
+ "eval_loss": 26.68852996826172,
320
+ "eval_runtime": 267.9224,
321
+ "eval_samples_per_second": 20.383,
322
+ "eval_steps_per_second": 2.549,
323
+ "eval_wer": 1.0,
324
+ "step": 2800
325
+ },
326
+ {
327
+ "epoch": 5.609284332688588,
328
+ "eval_cer": 0.9907495636188854,
329
+ "eval_loss": 24.301000595092773,
330
+ "eval_runtime": 235.2693,
331
+ "eval_samples_per_second": 23.212,
332
+ "eval_steps_per_second": 2.903,
333
+ "eval_wer": 1.0,
334
+ "step": 2900
335
+ },
336
+ {
337
+ "epoch": 5.802707930367505,
338
+ "grad_norm": 106.30471801757812,
339
+ "learning_rate": 7.1856e-06,
340
+ "loss": 23.4284,
341
+ "step": 3000
342
+ },
343
+ {
344
+ "epoch": 5.802707930367505,
345
+ "eval_cer": 0.9907495636188854,
346
+ "eval_loss": 21.88077735900879,
347
+ "eval_runtime": 232.75,
348
+ "eval_samples_per_second": 23.463,
349
+ "eval_steps_per_second": 2.934,
350
+ "eval_wer": 1.0,
351
+ "step": 3000
352
+ },
353
+ {
354
+ "epoch": 5.996131528046422,
355
+ "eval_cer": 0.9907917260454173,
356
+ "eval_loss": 19.473451614379883,
357
+ "eval_runtime": 233.7728,
358
+ "eval_samples_per_second": 23.36,
359
+ "eval_steps_per_second": 2.922,
360
+ "eval_wer": 1.0,
361
+ "step": 3100
362
+ },
363
+ {
364
+ "epoch": 6.189555125725338,
365
+ "eval_cer": 0.9908760508984813,
366
+ "eval_loss": 17.129289627075195,
367
+ "eval_runtime": 240.7161,
368
+ "eval_samples_per_second": 22.686,
369
+ "eval_steps_per_second": 2.837,
370
+ "eval_wer": 1.0,
371
+ "step": 3200
372
+ },
373
+ {
374
+ "epoch": 6.382978723404255,
375
+ "eval_cer": 0.9907917260454173,
376
+ "eval_loss": 14.863801002502441,
377
+ "eval_runtime": 233.6375,
378
+ "eval_samples_per_second": 23.374,
379
+ "eval_steps_per_second": 2.923,
380
+ "eval_wer": 1.0,
381
+ "step": 3300
382
+ },
383
+ {
384
+ "epoch": 6.576402321083172,
385
+ "eval_cer": 0.9907074011923535,
386
+ "eval_loss": 12.806206703186035,
387
+ "eval_runtime": 251.782,
388
+ "eval_samples_per_second": 21.689,
389
+ "eval_steps_per_second": 2.713,
390
+ "eval_wer": 1.0,
391
+ "step": 3400
392
+ },
393
+ {
394
+ "epoch": 6.769825918762089,
395
+ "grad_norm": 82.23467254638672,
396
+ "learning_rate": 8.3856e-06,
397
+ "loss": 13.9431,
398
+ "step": 3500
399
+ },
400
+ {
401
+ "epoch": 6.769825918762089,
402
+ "eval_cer": 0.9907074011923535,
403
+ "eval_loss": 10.964253425598145,
404
+ "eval_runtime": 233.2904,
405
+ "eval_samples_per_second": 23.409,
406
+ "eval_steps_per_second": 2.928,
407
+ "eval_wer": 1.0,
408
+ "step": 3500
409
+ },
410
+ {
411
+ "epoch": 6.963249516441006,
412
+ "eval_cer": 0.9907495636188854,
413
+ "eval_loss": 9.411906242370605,
414
+ "eval_runtime": 235.0547,
415
+ "eval_samples_per_second": 23.233,
416
+ "eval_steps_per_second": 2.906,
417
+ "eval_wer": 1.0,
418
+ "step": 3600
419
+ },
420
+ {
421
+ "epoch": 7.156673114119923,
422
+ "eval_cer": 0.9907495636188854,
423
+ "eval_loss": 8.164007186889648,
424
+ "eval_runtime": 287.0729,
425
+ "eval_samples_per_second": 19.023,
426
+ "eval_steps_per_second": 2.379,
427
+ "eval_wer": 1.0,
428
+ "step": 3700
429
+ },
430
+ {
431
+ "epoch": 7.350096711798839,
432
+ "eval_cer": 0.9907495636188854,
433
+ "eval_loss": 7.22973108291626,
434
+ "eval_runtime": 237.3392,
435
+ "eval_samples_per_second": 23.009,
436
+ "eval_steps_per_second": 2.878,
437
+ "eval_wer": 1.0,
438
+ "step": 3800
439
+ },
440
+ {
441
+ "epoch": 7.543520309477756,
442
+ "eval_cer": 0.9907074011923535,
443
+ "eval_loss": 6.571568489074707,
444
+ "eval_runtime": 232.6033,
445
+ "eval_samples_per_second": 23.478,
446
+ "eval_steps_per_second": 2.936,
447
+ "eval_wer": 1.0,
448
+ "step": 3900
449
+ },
450
+ {
451
+ "epoch": 7.7369439071566735,
452
+ "grad_norm": 4.932778358459473,
453
+ "learning_rate": 9.585600000000002e-06,
454
+ "loss": 7.4585,
455
+ "step": 4000
456
+ },
457
+ {
458
+ "epoch": 7.7369439071566735,
459
+ "eval_cer": 0.9907495636188854,
460
+ "eval_loss": 6.141250133514404,
461
+ "eval_runtime": 234.2968,
462
+ "eval_samples_per_second": 23.308,
463
+ "eval_steps_per_second": 2.915,
464
+ "eval_wer": 1.0,
465
+ "step": 4000
466
+ },
467
+ {
468
+ "epoch": 7.93036750483559,
469
+ "eval_cer": 0.9907074011923535,
470
+ "eval_loss": 5.885389804840088,
471
+ "eval_runtime": 236.1436,
472
+ "eval_samples_per_second": 23.126,
473
+ "eval_steps_per_second": 2.892,
474
+ "eval_wer": 1.0,
475
+ "step": 4100
476
+ },
477
+ {
478
+ "epoch": 8.123791102514506,
479
+ "eval_cer": 0.9907495636188854,
480
+ "eval_loss": 5.770660877227783,
481
+ "eval_runtime": 234.5762,
482
+ "eval_samples_per_second": 23.28,
483
+ "eval_steps_per_second": 2.912,
484
+ "eval_wer": 1.0,
485
+ "step": 4200
486
+ },
487
+ {
488
+ "epoch": 8.317214700193423,
489
+ "eval_cer": 0.9907074011923535,
490
+ "eval_loss": 5.680200576782227,
491
+ "eval_runtime": 231.827,
492
+ "eval_samples_per_second": 23.556,
493
+ "eval_steps_per_second": 2.946,
494
+ "eval_wer": 1.0,
495
+ "step": 4300
496
+ },
497
+ {
498
+ "epoch": 8.51063829787234,
499
+ "eval_cer": 0.9907495636188854,
500
+ "eval_loss": 5.597055435180664,
501
+ "eval_runtime": 232.7784,
502
+ "eval_samples_per_second": 23.46,
503
+ "eval_steps_per_second": 2.934,
504
+ "eval_wer": 1.0,
505
+ "step": 4400
506
+ },
507
+ {
508
+ "epoch": 8.704061895551257,
509
+ "grad_norm": 1.6451424360275269,
510
+ "learning_rate": 1.07856e-05,
511
+ "loss": 5.7398,
512
+ "step": 4500
513
+ },
514
+ {
515
+ "epoch": 8.704061895551257,
516
+ "eval_cer": 0.9907495636188854,
517
+ "eval_loss": 5.533324241638184,
518
+ "eval_runtime": 232.5892,
519
+ "eval_samples_per_second": 23.479,
520
+ "eval_steps_per_second": 2.937,
521
+ "eval_wer": 1.0,
522
+ "step": 4500
523
+ },
524
+ {
525
+ "epoch": 8.897485493230175,
526
+ "eval_cer": 0.9907074011923535,
527
+ "eval_loss": 5.475062370300293,
528
+ "eval_runtime": 233.1057,
529
+ "eval_samples_per_second": 23.427,
530
+ "eval_steps_per_second": 2.93,
531
+ "eval_wer": 1.0,
532
+ "step": 4600
533
+ },
534
+ {
535
+ "epoch": 9.090909090909092,
536
+ "eval_cer": 0.9907495636188854,
537
+ "eval_loss": 5.425434589385986,
538
+ "eval_runtime": 232.7779,
539
+ "eval_samples_per_second": 23.46,
540
+ "eval_steps_per_second": 2.934,
541
+ "eval_wer": 1.0,
542
+ "step": 4700
543
+ },
544
+ {
545
+ "epoch": 9.284332688588007,
546
+ "eval_cer": 0.9908423209572558,
547
+ "eval_loss": 5.377471923828125,
548
+ "eval_runtime": 234.4379,
549
+ "eval_samples_per_second": 23.294,
550
+ "eval_steps_per_second": 2.913,
551
+ "eval_wer": 1.1319063300402488,
552
+ "step": 4800
553
+ },
554
+ {
555
+ "epoch": 9.477756286266924,
556
+ "eval_cer": 0.9907495636188854,
557
+ "eval_loss": 5.343258380889893,
558
+ "eval_runtime": 239.0036,
559
+ "eval_samples_per_second": 22.849,
560
+ "eval_steps_per_second": 2.858,
561
+ "eval_wer": 1.33205268935236,
562
+ "step": 4900
563
+ },
564
+ {
565
+ "epoch": 9.671179883945841,
566
+ "grad_norm": 2.292585611343384,
567
+ "learning_rate": 1.19856e-05,
568
+ "loss": 5.4159,
569
+ "step": 5000
570
+ },
571
+ {
572
+ "epoch": 9.671179883945841,
573
+ "eval_cer": 0.9906146438539831,
574
+ "eval_loss": 5.311874866485596,
575
+ "eval_runtime": 235.3723,
576
+ "eval_samples_per_second": 23.202,
577
+ "eval_steps_per_second": 2.902,
578
+ "eval_wer": 1.686242224661544,
579
+ "step": 5000
580
+ },
581
+ {
582
+ "epoch": 9.864603481624759,
583
+ "eval_cer": 0.9909688082368516,
584
+ "eval_loss": 5.269064426422119,
585
+ "eval_runtime": 234.5017,
586
+ "eval_samples_per_second": 23.288,
587
+ "eval_steps_per_second": 2.913,
588
+ "eval_wer": 1.4255396999634102,
589
+ "step": 5100
590
+ },
591
+ {
592
+ "epoch": 10.058027079303676,
593
+ "eval_cer": 0.9909182133250133,
594
+ "eval_loss": 5.236879825592041,
595
+ "eval_runtime": 235.3148,
596
+ "eval_samples_per_second": 23.207,
597
+ "eval_steps_per_second": 2.902,
598
+ "eval_wer": 1.4043175997072814,
599
+ "step": 5200
600
+ },
601
+ {
602
+ "epoch": 10.251450676982591,
603
+ "eval_cer": 0.9909603757515453,
604
+ "eval_loss": 5.194947719573975,
605
+ "eval_runtime": 232.4545,
606
+ "eval_samples_per_second": 23.493,
607
+ "eval_steps_per_second": 2.938,
608
+ "eval_wer": 1.5686059275521405,
609
+ "step": 5300
610
+ },
611
+ {
612
+ "epoch": 10.444874274661508,
613
+ "eval_cer": 0.9908170235013366,
614
+ "eval_loss": 5.151918888092041,
615
+ "eval_runtime": 231.5581,
616
+ "eval_samples_per_second": 23.584,
617
+ "eval_steps_per_second": 2.95,
618
+ "eval_wer": 1.5166483717526527,
619
+ "step": 5400
620
+ },
621
+ {
622
+ "epoch": 10.638297872340425,
623
+ "grad_norm": 2.0632503032684326,
624
+ "learning_rate": 1.3185600000000001e-05,
625
+ "loss": 5.2163,
626
+ "step": 5500
627
+ },
628
+ {
629
+ "epoch": 10.638297872340425,
630
+ "eval_cer": 0.9909603757515453,
631
+ "eval_loss": 5.108114242553711,
632
+ "eval_runtime": 232.6627,
633
+ "eval_samples_per_second": 23.472,
634
+ "eval_steps_per_second": 2.936,
635
+ "eval_wer": 1.247713135748262,
636
+ "step": 5500
637
+ },
638
+ {
639
+ "epoch": 10.831721470019342,
640
+ "eval_cer": 0.9907664285894981,
641
+ "eval_loss": 5.055335998535156,
642
+ "eval_runtime": 231.9668,
643
+ "eval_samples_per_second": 23.542,
644
+ "eval_steps_per_second": 2.944,
645
+ "eval_wer": 1.5124405415294548,
646
+ "step": 5600
647
+ },
648
+ {
649
+ "epoch": 11.02514506769826,
650
+ "eval_cer": 0.9908760508984813,
651
+ "eval_loss": 5.012271881103516,
652
+ "eval_runtime": 231.643,
653
+ "eval_samples_per_second": 23.575,
654
+ "eval_steps_per_second": 2.949,
655
+ "eval_wer": 1.5495792169776803,
656
+ "step": 5700
657
+ },
658
+ {
659
+ "epoch": 11.218568665377177,
660
+ "eval_cer": 0.9885824148951421,
661
+ "eval_loss": 4.942389965057373,
662
+ "eval_runtime": 232.3058,
663
+ "eval_samples_per_second": 23.508,
664
+ "eval_steps_per_second": 2.94,
665
+ "eval_wer": 1.7621661178192463,
666
+ "step": 5800
667
+ },
668
+ {
669
+ "epoch": 11.411992263056092,
670
+ "eval_cer": 0.9830675695047602,
671
+ "eval_loss": 4.8753485679626465,
672
+ "eval_runtime": 233.7911,
673
+ "eval_samples_per_second": 23.358,
674
+ "eval_steps_per_second": 2.921,
675
+ "eval_wer": 1.5404317599707282,
676
+ "step": 5900
677
+ },
678
+ {
679
+ "epoch": 11.60541586073501,
680
+ "grad_norm": 1.9146788120269775,
681
+ "learning_rate": 1.43856e-05,
682
+ "loss": 4.9465,
683
+ "step": 6000
684
+ },
685
+ {
686
+ "epoch": 11.60541586073501,
687
+ "eval_cer": 0.9749808160959279,
688
+ "eval_loss": 4.77677059173584,
689
+ "eval_runtime": 232.0203,
690
+ "eval_samples_per_second": 23.537,
691
+ "eval_steps_per_second": 2.944,
692
+ "eval_wer": 1.853457738748628,
693
+ "step": 6000
694
+ },
695
+ {
696
+ "epoch": 11.798839458413926,
697
+ "eval_cer": 0.9713126849876464,
698
+ "eval_loss": 4.6841044425964355,
699
+ "eval_runtime": 231.7273,
700
+ "eval_samples_per_second": 23.566,
701
+ "eval_steps_per_second": 2.947,
702
+ "eval_wer": 1.8395536040980607,
703
+ "step": 6100
704
+ },
705
+ {
706
+ "epoch": 11.992263056092844,
707
+ "eval_cer": 0.9697020802941251,
708
+ "eval_loss": 4.582820892333984,
709
+ "eval_runtime": 232.9124,
710
+ "eval_samples_per_second": 23.447,
711
+ "eval_steps_per_second": 2.932,
712
+ "eval_wer": 1.7444200512257593,
713
+ "step": 6200
714
+ },
715
+ {
716
+ "epoch": 12.18568665377176,
717
+ "eval_cer": 0.968909426675324,
718
+ "eval_loss": 4.485307693481445,
719
+ "eval_runtime": 232.9842,
720
+ "eval_samples_per_second": 23.439,
721
+ "eval_steps_per_second": 2.932,
722
+ "eval_wer": 1.801317233809001,
723
+ "step": 6300
724
+ },
725
+ {
726
+ "epoch": 12.379110251450676,
727
+ "eval_cer": 0.9556029648618337,
728
+ "eval_loss": 4.395504951477051,
729
+ "eval_runtime": 233.0172,
730
+ "eval_samples_per_second": 23.436,
731
+ "eval_steps_per_second": 2.931,
732
+ "eval_wer": 1.827844859129162,
733
+ "step": 6400
734
+ },
735
+ {
736
+ "epoch": 12.572533849129593,
737
+ "grad_norm": 4.459765434265137,
738
+ "learning_rate": 1.55856e-05,
739
+ "loss": 4.5094,
740
+ "step": 6500
741
+ },
742
+ {
743
+ "epoch": 12.572533849129593,
744
+ "eval_cer": 0.9123021528134987,
745
+ "eval_loss": 4.284241676330566,
746
+ "eval_runtime": 230.5488,
747
+ "eval_samples_per_second": 23.687,
748
+ "eval_steps_per_second": 2.962,
749
+ "eval_wer": 1.8728503476033662,
750
+ "step": 6500
751
+ },
752
+ {
753
+ "epoch": 12.76595744680851,
754
+ "eval_cer": 0.8650380726711584,
755
+ "eval_loss": 4.181938171386719,
756
+ "eval_runtime": 232.0403,
757
+ "eval_samples_per_second": 23.535,
758
+ "eval_steps_per_second": 2.943,
759
+ "eval_wer": 1.9094401756311745,
760
+ "step": 6600
761
+ },
762
+ {
763
+ "epoch": 12.959381044487428,
764
+ "eval_cer": 0.848628456264915,
765
+ "eval_loss": 4.074057579040527,
766
+ "eval_runtime": 231.4468,
767
+ "eval_samples_per_second": 23.595,
768
+ "eval_steps_per_second": 2.951,
769
+ "eval_wer": 1.9134650567142335,
770
+ "step": 6700
771
+ },
772
+ {
773
+ "epoch": 13.152804642166345,
774
+ "eval_cer": 0.8385937987503057,
775
+ "eval_loss": 3.9648523330688477,
776
+ "eval_runtime": 231.3067,
777
+ "eval_samples_per_second": 23.609,
778
+ "eval_steps_per_second": 2.953,
779
+ "eval_wer": 1.9191364800585438,
780
+ "step": 6800
781
+ },
782
+ {
783
+ "epoch": 13.346228239845262,
784
+ "eval_cer": 0.8189292430157941,
785
+ "eval_loss": 3.8640658855438232,
786
+ "eval_runtime": 237.0711,
787
+ "eval_samples_per_second": 23.035,
788
+ "eval_steps_per_second": 2.881,
789
+ "eval_wer": 1.9195023783388219,
790
+ "step": 6900
791
+ },
792
+ {
793
+ "epoch": 13.539651837524177,
794
+ "grad_norm": 4.586193084716797,
795
+ "learning_rate": 1.67856e-05,
796
+ "loss": 4.0097,
797
+ "step": 7000
798
+ },
799
+ {
800
+ "epoch": 13.539651837524177,
801
+ "eval_cer": 0.8014487009756386,
802
+ "eval_loss": 3.7686750888824463,
803
+ "eval_runtime": 244.328,
804
+ "eval_samples_per_second": 22.351,
805
+ "eval_steps_per_second": 2.795,
806
+ "eval_wer": 1.9275521405049396,
807
+ "step": 7000
808
+ },
809
+ {
810
+ "epoch": 13.733075435203094,
811
+ "eval_cer": 0.7963301823946571,
812
+ "eval_loss": 3.680776596069336,
813
+ "eval_runtime": 230.3109,
814
+ "eval_samples_per_second": 23.711,
815
+ "eval_steps_per_second": 2.966,
816
+ "eval_wer": 1.9259055982436883,
817
+ "step": 7100
818
+ },
819
+ {
820
+ "epoch": 13.926499032882012,
821
+ "eval_cer": 0.779220669707983,
822
+ "eval_loss": 3.60208797454834,
823
+ "eval_runtime": 229.5487,
824
+ "eval_samples_per_second": 23.79,
825
+ "eval_steps_per_second": 2.975,
826
+ "eval_wer": 1.9275521405049396,
827
+ "step": 7200
828
+ },
829
+ {
830
+ "epoch": 14.119922630560929,
831
+ "eval_cer": 0.777466712764253,
832
+ "eval_loss": 3.55332088470459,
833
+ "eval_runtime": 231.4776,
834
+ "eval_samples_per_second": 23.592,
835
+ "eval_steps_per_second": 2.951,
836
+ "eval_wer": 1.9366995975118917,
837
+ "step": 7300
838
+ },
839
+ {
840
+ "epoch": 14.313346228239846,
841
+ "eval_cer": 0.775071886937237,
842
+ "eval_loss": 3.476841449737549,
843
+ "eval_runtime": 240.5885,
844
+ "eval_samples_per_second": 22.699,
845
+ "eval_steps_per_second": 2.839,
846
+ "eval_wer": 1.9321258690084158,
847
+ "step": 7400
848
+ },
849
+ {
850
+ "epoch": 14.506769825918763,
851
+ "grad_norm": 4.310749053955078,
852
+ "learning_rate": 1.7985600000000003e-05,
853
+ "loss": 3.5619,
854
+ "step": 7500
855
+ },
856
+ {
857
+ "epoch": 14.506769825918763,
858
+ "eval_cer": 0.7672043781463711,
859
+ "eval_loss": 3.4284844398498535,
860
+ "eval_runtime": 240.9512,
861
+ "eval_samples_per_second": 22.664,
862
+ "eval_steps_per_second": 2.835,
863
+ "eval_wer": 1.938529088913282,
864
+ "step": 7500
865
+ },
866
+ {
867
+ "epoch": 14.700193423597678,
868
+ "eval_cer": 0.7659816677769439,
869
+ "eval_loss": 3.362793445587158,
870
+ "eval_runtime": 230.1401,
871
+ "eval_samples_per_second": 23.729,
872
+ "eval_steps_per_second": 2.968,
873
+ "eval_wer": 1.9361507500914745,
874
+ "step": 7600
875
+ },
876
+ {
877
+ "epoch": 14.893617021276595,
878
+ "eval_cer": 0.7618919124033426,
879
+ "eval_loss": 3.2909657955169678,
880
+ "eval_runtime": 233.6336,
881
+ "eval_samples_per_second": 23.374,
882
+ "eval_steps_per_second": 2.923,
883
+ "eval_wer": 1.9313940724478595,
884
+ "step": 7700
885
+ },
886
+ {
887
+ "epoch": 15.087040618955513,
888
+ "eval_cer": 0.7486360455016907,
889
+ "eval_loss": 3.2242627143859863,
890
+ "eval_runtime": 241.9894,
891
+ "eval_samples_per_second": 22.567,
892
+ "eval_steps_per_second": 2.822,
893
+ "eval_wer": 1.928832784485913,
894
+ "step": 7800
895
+ },
896
+ {
897
+ "epoch": 15.28046421663443,
898
+ "eval_cer": 0.7431802275084536,
899
+ "eval_loss": 3.164518117904663,
900
+ "eval_runtime": 232.2322,
901
+ "eval_samples_per_second": 23.515,
902
+ "eval_steps_per_second": 2.941,
903
+ "eval_wer": 1.9308452250274424,
904
+ "step": 7900
905
+ },
906
+ {
907
+ "epoch": 15.473887814313347,
908
+ "grad_norm": 5.67767333984375,
909
+ "learning_rate": 1.91856e-05,
910
+ "loss": 3.2379,
911
+ "step": 8000
912
+ },
913
+ {
914
+ "epoch": 15.473887814313347,
915
+ "eval_cer": 0.7382556560895193,
916
+ "eval_loss": 3.1185944080352783,
917
+ "eval_runtime": 235.8886,
918
+ "eval_samples_per_second": 23.151,
919
+ "eval_steps_per_second": 2.895,
920
+ "eval_wer": 1.9332235638492499,
921
+ "step": 8000
922
+ },
923
+ {
924
+ "epoch": 15.667311411992262,
925
+ "eval_cer": 0.7374714349560246,
926
+ "eval_loss": 3.078275203704834,
927
+ "eval_runtime": 234.7574,
928
+ "eval_samples_per_second": 23.262,
929
+ "eval_steps_per_second": 2.909,
930
+ "eval_wer": 1.9348701061105014,
931
+ "step": 8100
932
+ },
933
+ {
934
+ "epoch": 15.86073500967118,
935
+ "eval_cer": 0.727942726559799,
936
+ "eval_loss": 3.0145950317382812,
937
+ "eval_runtime": 231.4339,
938
+ "eval_samples_per_second": 23.596,
939
+ "eval_steps_per_second": 2.951,
940
+ "eval_wer": 1.9321258690084158,
941
+ "step": 8200
942
+ },
943
+ {
944
+ "epoch": 16.054158607350097,
945
+ "eval_cer": 0.7300086854598656,
946
+ "eval_loss": 2.9523487091064453,
947
+ "eval_runtime": 235.1315,
948
+ "eval_samples_per_second": 23.225,
949
+ "eval_steps_per_second": 2.905,
950
+ "eval_wer": 1.9308452250274424,
951
+ "step": 8300
952
+ },
953
+ {
954
+ "epoch": 16.247582205029012,
955
+ "eval_cer": 0.7254467109091063,
956
+ "eval_loss": 2.918687105178833,
957
+ "eval_runtime": 238.3374,
958
+ "eval_samples_per_second": 22.913,
959
+ "eval_steps_per_second": 2.866,
960
+ "eval_wer": 1.9273691913648006,
961
+ "step": 8400
962
+ },
963
+ {
964
+ "epoch": 16.44100580270793,
965
+ "grad_norm": 8.755329132080078,
966
+ "learning_rate": 2.03856e-05,
967
+ "loss": 2.9448,
968
+ "step": 8500
969
+ },
970
+ {
971
+ "epoch": 16.44100580270793,
972
+ "eval_cer": 0.7177394193390618,
973
+ "eval_loss": 2.8671371936798096,
974
+ "eval_runtime": 229.0989,
975
+ "eval_samples_per_second": 23.837,
976
+ "eval_steps_per_second": 2.981,
977
+ "eval_wer": 1.929015733626052,
978
+ "step": 8500
979
+ },
980
+ {
981
+ "epoch": 16.634429400386846,
982
+ "eval_cer": 0.7115752725800876,
983
+ "eval_loss": 2.8188540935516357,
984
+ "eval_runtime": 230.6192,
985
+ "eval_samples_per_second": 23.68,
986
+ "eval_steps_per_second": 2.962,
987
+ "eval_wer": 1.9348701061105014,
988
+ "step": 8600
989
+ },
990
+ {
991
+ "epoch": 16.827852998065765,
992
+ "eval_cer": 0.70778065419221,
993
+ "eval_loss": 2.76908802986145,
994
+ "eval_runtime": 239.0894,
995
+ "eval_samples_per_second": 22.841,
996
+ "eval_steps_per_second": 2.857,
997
+ "eval_wer": 1.9365166483717526,
998
+ "step": 8700
999
+ },
1000
+ {
1001
+ "epoch": 17.02127659574468,
1002
+ "eval_cer": 0.7069205406909579,
1003
+ "eval_loss": 2.731661081314087,
1004
+ "eval_runtime": 231.1522,
1005
+ "eval_samples_per_second": 23.625,
1006
+ "eval_steps_per_second": 2.955,
1007
+ "eval_wer": 1.9420051225759238,
1008
+ "step": 8800
1009
+ },
1010
+ {
1011
+ "epoch": 17.214700193423596,
1012
+ "eval_cer": 0.7056388029243859,
1013
+ "eval_loss": 2.683185577392578,
1014
+ "eval_runtime": 229.2327,
1015
+ "eval_samples_per_second": 23.823,
1016
+ "eval_steps_per_second": 2.98,
1017
+ "eval_wer": 1.9489571899012075,
1018
+ "step": 8900
1019
+ },
1020
+ {
1021
+ "epoch": 17.408123791102515,
1022
+ "grad_norm": 4.358935832977295,
1023
+ "learning_rate": 2.1585600000000002e-05,
1024
+ "loss": 2.6749,
1025
+ "step": 9000
1026
+ },
1027
+ {
1028
+ "epoch": 17.408123791102515,
1029
+ "eval_cer": 0.7020128342426364,
1030
+ "eval_loss": 2.6420364379882812,
1031
+ "eval_runtime": 229.374,
1032
+ "eval_samples_per_second": 23.808,
1033
+ "eval_steps_per_second": 2.978,
1034
+ "eval_wer": 1.978412001463593,
1035
+ "step": 9000
1036
+ },
1037
+ {
1038
+ "epoch": 17.60154738878143,
1039
+ "eval_cer": 0.6991204917825431,
1040
+ "eval_loss": 2.601982831954956,
1041
+ "eval_runtime": 229.7146,
1042
+ "eval_samples_per_second": 23.773,
1043
+ "eval_steps_per_second": 2.973,
1044
+ "eval_wer": 1.9414562751555067,
1045
+ "step": 9100
1046
+ },
1047
+ {
1048
+ "epoch": 17.79497098646035,
1049
+ "eval_cer": 0.6994577911947989,
1050
+ "eval_loss": 2.5666821002960205,
1051
+ "eval_runtime": 234.5424,
1052
+ "eval_samples_per_second": 23.284,
1053
+ "eval_steps_per_second": 2.912,
1054
+ "eval_wer": 1.9762166117819246,
1055
+ "step": 9200
1056
+ },
1057
+ {
1058
+ "epoch": 17.988394584139265,
1059
+ "eval_cer": 0.6771285701034666,
1060
+ "eval_loss": 2.517096757888794,
1061
+ "eval_runtime": 247.5848,
1062
+ "eval_samples_per_second": 22.057,
1063
+ "eval_steps_per_second": 2.759,
1064
+ "eval_wer": 1.9857299670691548,
1065
+ "step": 9300
1066
+ },
1067
+ {
1068
+ "epoch": 18.181818181818183,
1069
+ "eval_cer": 0.6774658695157224,
1070
+ "eval_loss": 2.492238759994507,
1071
+ "eval_runtime": 229.5938,
1072
+ "eval_samples_per_second": 23.785,
1073
+ "eval_steps_per_second": 2.975,
1074
+ "eval_wer": 1.9890230515916576,
1075
+ "step": 9400
1076
+ },
1077
+ {
1078
+ "epoch": 18.3752417794971,
1079
+ "grad_norm": 6.681089878082275,
1080
+ "learning_rate": 2.27856e-05,
1081
+ "loss": 2.4473,
1082
+ "step": 9500
1083
+ },
1084
+ {
1085
+ "epoch": 18.3752417794971,
1086
+ "eval_cer": 0.6682575955611397,
1087
+ "eval_loss": 2.445500373840332,
1088
+ "eval_runtime": 229.8314,
1089
+ "eval_samples_per_second": 23.761,
1090
+ "eval_steps_per_second": 2.972,
1091
+ "eval_wer": 1.9882912550311014,
1092
+ "step": 9500
1093
+ },
1094
+ {
1095
+ "epoch": 18.568665377176014,
1096
+ "eval_cer": 0.6620681513462463,
1097
+ "eval_loss": 2.4191782474517822,
1098
+ "eval_runtime": 230.0128,
1099
+ "eval_samples_per_second": 23.742,
1100
+ "eval_steps_per_second": 2.969,
1101
+ "eval_wer": 1.9815221368459568,
1102
+ "step": 9600
1103
+ },
1104
+ {
1105
+ "epoch": 18.762088974854933,
1106
+ "eval_cer": 0.6523454957879736,
1107
+ "eval_loss": 2.386597156524658,
1108
+ "eval_runtime": 240.9084,
1109
+ "eval_samples_per_second": 22.668,
1110
+ "eval_steps_per_second": 2.835,
1111
+ "eval_wer": 1.9904866447127698,
1112
+ "step": 9700
1113
+ },
1114
+ {
1115
+ "epoch": 18.95551257253385,
1116
+ "eval_cer": 0.6539392355108822,
1117
+ "eval_loss": 2.335400342941284,
1118
+ "eval_runtime": 233.6249,
1119
+ "eval_samples_per_second": 23.375,
1120
+ "eval_steps_per_second": 2.923,
1121
+ "eval_wer": 1.991401390413465,
1122
+ "step": 9800
1123
+ },
1124
+ {
1125
+ "epoch": 19.148936170212767,
1126
+ "eval_cer": 0.6515612746544789,
1127
+ "eval_loss": 2.3113534450531006,
1128
+ "eval_runtime": 252.8489,
1129
+ "eval_samples_per_second": 21.598,
1130
+ "eval_steps_per_second": 2.701,
1131
+ "eval_wer": 1.9924990852542992,
1132
+ "step": 9900
1133
+ },
1134
+ {
1135
+ "epoch": 19.342359767891683,
1136
+ "grad_norm": 8.786458015441895,
1137
+ "learning_rate": 2.39856e-05,
1138
+ "loss": 2.2307,
1139
+ "step": 10000
1140
+ },
1141
+ {
1142
+ "epoch": 19.342359767891683,
1143
+ "eval_cer": 0.645447722807343,
1144
+ "eval_loss": 2.269487142562866,
1145
+ "eval_runtime": 236.1588,
1146
+ "eval_samples_per_second": 23.124,
1147
+ "eval_steps_per_second": 2.892,
1148
+ "eval_wer": 1.9903036955726308,
1149
+ "step": 10000
1150
+ },
1151
+ {
1152
+ "epoch": 19.535783365570598,
1153
+ "eval_cer": 0.6464258911028847,
1154
+ "eval_loss": 2.246647834777832,
1155
+ "eval_runtime": 241.6204,
1156
+ "eval_samples_per_second": 22.602,
1157
+ "eval_steps_per_second": 2.827,
1158
+ "eval_wer": 1.9924990852542992,
1159
+ "step": 10100
1160
+ },
1161
+ {
1162
+ "epoch": 19.729206963249517,
1163
+ "eval_cer": 0.6422939733027515,
1164
+ "eval_loss": 2.2167210578918457,
1165
+ "eval_runtime": 230.6902,
1166
+ "eval_samples_per_second": 23.672,
1167
+ "eval_steps_per_second": 2.961,
1168
+ "eval_wer": 1.9928649835345773,
1169
+ "step": 10200
1170
+ },
1171
+ {
1172
+ "epoch": 19.922630560928432,
1173
+ "eval_cer": 0.641256777610065,
1174
+ "eval_loss": 2.176206588745117,
1175
+ "eval_runtime": 236.5184,
1176
+ "eval_samples_per_second": 23.089,
1177
+ "eval_steps_per_second": 2.888,
1178
+ "eval_wer": 1.991401390413465,
1179
+ "step": 10300
1180
+ },
1181
+ {
1182
+ "epoch": 20.0,
1183
+ "step": 10340,
1184
+ "total_flos": 1.4857862206321902e+19,
1185
+ "train_loss": 15.407913101757057,
1186
+ "train_runtime": 65838.1292,
1187
+ "train_samples_per_second": 5.022,
1188
+ "train_steps_per_second": 0.157
1189
+ }
1190
+ ],
1191
+ "logging_steps": 500,
1192
+ "max_steps": 10340,
1193
+ "num_input_tokens_seen": 0,
1194
+ "num_train_epochs": 20,
1195
+ "save_steps": 400,
1196
+ "stateful_callbacks": {
1197
+ "TrainerControl": {
1198
+ "args": {
1199
+ "should_epoch_stop": false,
1200
+ "should_evaluate": false,
1201
+ "should_log": false,
1202
+ "should_save": true,
1203
+ "should_training_stop": true
1204
+ },
1205
+ "attributes": {}
1206
+ }
1207
+ },
1208
+ "total_flos": 1.4857862206321902e+19,
1209
+ "train_batch_size": 16,
1210
+ "trial_name": null,
1211
+ "trial_params": null
1212
+ }