tuanio commited on
Commit
f633d2f
1 Parent(s): a83a8cf

End of training

Browse files
Files changed (1) hide show
  1. trainer_state.json +1518 -0
trainer_state.json ADDED
@@ -0,0 +1,1518 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": 0.8174300254452926,
3
+ "best_model_checkpoint": "training_sentiment_analysis/checkpoint-8600",
4
+ "epoch": 20.0,
5
+ "eval_steps": 200,
6
+ "global_step": 18680,
7
+ "is_hyper_param_search": false,
8
+ "is_local_process_zero": true,
9
+ "is_world_process_zero": true,
10
+ "log_history": [
11
+ {
12
+ "epoch": 0.21,
13
+ "grad_norm": 1.3381836414337158,
14
+ "learning_rate": 3.2119914346895075e-05,
15
+ "loss": 0.9299,
16
+ "step": 200
17
+ },
18
+ {
19
+ "epoch": 0.21,
20
+ "eval_accuracy": 0.638676844783715,
21
+ "eval_loss": 0.827367901802063,
22
+ "eval_runtime": 3.055,
23
+ "eval_samples_per_second": 514.569,
24
+ "eval_steps_per_second": 16.367,
25
+ "step": 200
26
+ },
27
+ {
28
+ "epoch": 0.43,
29
+ "grad_norm": 1.0220164060592651,
30
+ "learning_rate": 6.423982869379015e-05,
31
+ "loss": 0.7793,
32
+ "step": 400
33
+ },
34
+ {
35
+ "epoch": 0.43,
36
+ "eval_accuracy": 0.7188295165394402,
37
+ "eval_loss": 0.6643335223197937,
38
+ "eval_runtime": 3.0013,
39
+ "eval_samples_per_second": 523.77,
40
+ "eval_steps_per_second": 16.659,
41
+ "step": 400
42
+ },
43
+ {
44
+ "epoch": 0.64,
45
+ "grad_norm": 1.7421491146087646,
46
+ "learning_rate": 9.635974304068522e-05,
47
+ "loss": 0.6574,
48
+ "step": 600
49
+ },
50
+ {
51
+ "epoch": 0.64,
52
+ "eval_accuracy": 0.7659033078880407,
53
+ "eval_loss": 0.5868020057678223,
54
+ "eval_runtime": 2.9749,
55
+ "eval_samples_per_second": 528.422,
56
+ "eval_steps_per_second": 16.807,
57
+ "step": 600
58
+ },
59
+ {
60
+ "epoch": 0.86,
61
+ "grad_norm": 1.8133894205093384,
62
+ "learning_rate": 0.0001284796573875803,
63
+ "loss": 0.6132,
64
+ "step": 800
65
+ },
66
+ {
67
+ "epoch": 0.86,
68
+ "eval_accuracy": 0.772264631043257,
69
+ "eval_loss": 0.5582301616668701,
70
+ "eval_runtime": 2.9908,
71
+ "eval_samples_per_second": 525.617,
72
+ "eval_steps_per_second": 16.718,
73
+ "step": 800
74
+ },
75
+ {
76
+ "epoch": 1.07,
77
+ "grad_norm": 1.3071078062057495,
78
+ "learning_rate": 0.00016059957173447537,
79
+ "loss": 0.5791,
80
+ "step": 1000
81
+ },
82
+ {
83
+ "epoch": 1.07,
84
+ "eval_accuracy": 0.7830788804071247,
85
+ "eval_loss": 0.5515692234039307,
86
+ "eval_runtime": 2.9665,
87
+ "eval_samples_per_second": 529.915,
88
+ "eval_steps_per_second": 16.855,
89
+ "step": 1000
90
+ },
91
+ {
92
+ "epoch": 1.28,
93
+ "grad_norm": 1.0445743799209595,
94
+ "learning_rate": 0.00019271948608137044,
95
+ "loss": 0.554,
96
+ "step": 1200
97
+ },
98
+ {
99
+ "epoch": 1.28,
100
+ "eval_accuracy": 0.7964376590330788,
101
+ "eval_loss": 0.5187413692474365,
102
+ "eval_runtime": 2.9846,
103
+ "eval_samples_per_second": 526.705,
104
+ "eval_steps_per_second": 16.753,
105
+ "step": 1200
106
+ },
107
+ {
108
+ "epoch": 1.5,
109
+ "grad_norm": 1.0763362646102905,
110
+ "learning_rate": 0.0002248394004282655,
111
+ "loss": 0.5258,
112
+ "step": 1400
113
+ },
114
+ {
115
+ "epoch": 1.5,
116
+ "eval_accuracy": 0.8034351145038168,
117
+ "eval_loss": 0.5125576257705688,
118
+ "eval_runtime": 2.9831,
119
+ "eval_samples_per_second": 526.967,
120
+ "eval_steps_per_second": 16.761,
121
+ "step": 1400
122
+ },
123
+ {
124
+ "epoch": 1.71,
125
+ "grad_norm": 0.8554897308349609,
126
+ "learning_rate": 0.0002569593147751606,
127
+ "loss": 0.5373,
128
+ "step": 1600
129
+ },
130
+ {
131
+ "epoch": 1.71,
132
+ "eval_accuracy": 0.8002544529262087,
133
+ "eval_loss": 0.51680988073349,
134
+ "eval_runtime": 2.9726,
135
+ "eval_samples_per_second": 528.823,
136
+ "eval_steps_per_second": 16.82,
137
+ "step": 1600
138
+ },
139
+ {
140
+ "epoch": 1.93,
141
+ "grad_norm": 1.538806438446045,
142
+ "learning_rate": 0.0002890792291220556,
143
+ "loss": 0.5266,
144
+ "step": 1800
145
+ },
146
+ {
147
+ "epoch": 1.93,
148
+ "eval_accuracy": 0.8027989821882952,
149
+ "eval_loss": 0.5283887982368469,
150
+ "eval_runtime": 2.9766,
151
+ "eval_samples_per_second": 528.12,
152
+ "eval_steps_per_second": 16.798,
153
+ "step": 1800
154
+ },
155
+ {
156
+ "epoch": 2.14,
157
+ "grad_norm": 1.1234441995620728,
158
+ "learning_rate": 0.000297644539614561,
159
+ "loss": 0.5076,
160
+ "step": 2000
161
+ },
162
+ {
163
+ "epoch": 2.14,
164
+ "eval_accuracy": 0.7977099236641222,
165
+ "eval_loss": 0.5178301334381104,
166
+ "eval_runtime": 2.9829,
167
+ "eval_samples_per_second": 526.996,
168
+ "eval_steps_per_second": 16.762,
169
+ "step": 2000
170
+ },
171
+ {
172
+ "epoch": 2.36,
173
+ "grad_norm": 1.6212774515151978,
174
+ "learning_rate": 0.0002940756602426838,
175
+ "loss": 0.5094,
176
+ "step": 2200
177
+ },
178
+ {
179
+ "epoch": 2.36,
180
+ "eval_accuracy": 0.8027989821882952,
181
+ "eval_loss": 0.5134572982788086,
182
+ "eval_runtime": 2.981,
183
+ "eval_samples_per_second": 527.334,
184
+ "eval_steps_per_second": 16.773,
185
+ "step": 2200
186
+ },
187
+ {
188
+ "epoch": 2.57,
189
+ "grad_norm": 1.4514294862747192,
190
+ "learning_rate": 0.00029050678087080655,
191
+ "loss": 0.5032,
192
+ "step": 2400
193
+ },
194
+ {
195
+ "epoch": 2.57,
196
+ "eval_accuracy": 0.8104325699745547,
197
+ "eval_loss": 0.5022692084312439,
198
+ "eval_runtime": 2.963,
199
+ "eval_samples_per_second": 530.535,
200
+ "eval_steps_per_second": 16.875,
201
+ "step": 2400
202
+ },
203
+ {
204
+ "epoch": 2.78,
205
+ "grad_norm": 1.826932668685913,
206
+ "learning_rate": 0.0002869379014989293,
207
+ "loss": 0.5034,
208
+ "step": 2600
209
+ },
210
+ {
211
+ "epoch": 2.78,
212
+ "eval_accuracy": 0.80470737913486,
213
+ "eval_loss": 0.5088226199150085,
214
+ "eval_runtime": 2.9831,
215
+ "eval_samples_per_second": 526.969,
216
+ "eval_steps_per_second": 16.761,
217
+ "step": 2600
218
+ },
219
+ {
220
+ "epoch": 3.0,
221
+ "grad_norm": 1.4404336214065552,
222
+ "learning_rate": 0.0002833690221270521,
223
+ "loss": 0.4923,
224
+ "step": 2800
225
+ },
226
+ {
227
+ "epoch": 3.0,
228
+ "eval_accuracy": 0.799618320610687,
229
+ "eval_loss": 0.5219257473945618,
230
+ "eval_runtime": 2.9722,
231
+ "eval_samples_per_second": 528.9,
232
+ "eval_steps_per_second": 16.823,
233
+ "step": 2800
234
+ },
235
+ {
236
+ "epoch": 3.21,
237
+ "grad_norm": 0.8795878291130066,
238
+ "learning_rate": 0.00027980014275517484,
239
+ "loss": 0.4934,
240
+ "step": 3000
241
+ },
242
+ {
243
+ "epoch": 3.21,
244
+ "eval_accuracy": 0.8129770992366412,
245
+ "eval_loss": 0.4905295968055725,
246
+ "eval_runtime": 2.9734,
247
+ "eval_samples_per_second": 528.696,
248
+ "eval_steps_per_second": 16.816,
249
+ "step": 3000
250
+ },
251
+ {
252
+ "epoch": 3.43,
253
+ "grad_norm": 1.6092537641525269,
254
+ "learning_rate": 0.0002762312633832976,
255
+ "loss": 0.4798,
256
+ "step": 3200
257
+ },
258
+ {
259
+ "epoch": 3.43,
260
+ "eval_accuracy": 0.8097964376590331,
261
+ "eval_loss": 0.4907812178134918,
262
+ "eval_runtime": 2.9897,
263
+ "eval_samples_per_second": 525.803,
264
+ "eval_steps_per_second": 16.724,
265
+ "step": 3200
266
+ },
267
+ {
268
+ "epoch": 3.64,
269
+ "grad_norm": 1.6475110054016113,
270
+ "learning_rate": 0.0002726623840114204,
271
+ "loss": 0.4831,
272
+ "step": 3400
273
+ },
274
+ {
275
+ "epoch": 3.64,
276
+ "eval_accuracy": 0.8072519083969466,
277
+ "eval_loss": 0.48748457431793213,
278
+ "eval_runtime": 2.9694,
279
+ "eval_samples_per_second": 529.396,
280
+ "eval_steps_per_second": 16.838,
281
+ "step": 3400
282
+ },
283
+ {
284
+ "epoch": 3.85,
285
+ "grad_norm": 1.1669467687606812,
286
+ "learning_rate": 0.00026909350463954313,
287
+ "loss": 0.4707,
288
+ "step": 3600
289
+ },
290
+ {
291
+ "epoch": 3.85,
292
+ "eval_accuracy": 0.8072519083969466,
293
+ "eval_loss": 0.4985896944999695,
294
+ "eval_runtime": 2.991,
295
+ "eval_samples_per_second": 525.579,
296
+ "eval_steps_per_second": 16.717,
297
+ "step": 3600
298
+ },
299
+ {
300
+ "epoch": 4.07,
301
+ "grad_norm": 0.9440352320671082,
302
+ "learning_rate": 0.00026552462526766593,
303
+ "loss": 0.4674,
304
+ "step": 3800
305
+ },
306
+ {
307
+ "epoch": 4.07,
308
+ "eval_accuracy": 0.8104325699745547,
309
+ "eval_loss": 0.5195557475090027,
310
+ "eval_runtime": 2.9789,
311
+ "eval_samples_per_second": 527.711,
312
+ "eval_steps_per_second": 16.785,
313
+ "step": 3800
314
+ },
315
+ {
316
+ "epoch": 4.28,
317
+ "grad_norm": 1.8151628971099854,
318
+ "learning_rate": 0.0002619557458957887,
319
+ "loss": 0.4535,
320
+ "step": 4000
321
+ },
322
+ {
323
+ "epoch": 4.28,
324
+ "eval_accuracy": 0.8097964376590331,
325
+ "eval_loss": 0.4896373152732849,
326
+ "eval_runtime": 2.9869,
327
+ "eval_samples_per_second": 526.295,
328
+ "eval_steps_per_second": 16.74,
329
+ "step": 4000
330
+ },
331
+ {
332
+ "epoch": 4.5,
333
+ "grad_norm": 3.0790090560913086,
334
+ "learning_rate": 0.0002583868665239115,
335
+ "loss": 0.464,
336
+ "step": 4200
337
+ },
338
+ {
339
+ "epoch": 4.5,
340
+ "eval_accuracy": 0.8078880407124682,
341
+ "eval_loss": 0.517495334148407,
342
+ "eval_runtime": 2.9986,
343
+ "eval_samples_per_second": 524.246,
344
+ "eval_steps_per_second": 16.674,
345
+ "step": 4200
346
+ },
347
+ {
348
+ "epoch": 4.71,
349
+ "grad_norm": 1.1520639657974243,
350
+ "learning_rate": 0.0002548179871520343,
351
+ "loss": 0.4715,
352
+ "step": 4400
353
+ },
354
+ {
355
+ "epoch": 4.71,
356
+ "eval_accuracy": 0.8027989821882952,
357
+ "eval_loss": 0.5001667737960815,
358
+ "eval_runtime": 2.9723,
359
+ "eval_samples_per_second": 528.885,
360
+ "eval_steps_per_second": 16.822,
361
+ "step": 4400
362
+ },
363
+ {
364
+ "epoch": 4.93,
365
+ "grad_norm": 0.8184943795204163,
366
+ "learning_rate": 0.000251249107780157,
367
+ "loss": 0.468,
368
+ "step": 4600
369
+ },
370
+ {
371
+ "epoch": 4.93,
372
+ "eval_accuracy": 0.8110687022900763,
373
+ "eval_loss": 0.4883332848548889,
374
+ "eval_runtime": 2.9769,
375
+ "eval_samples_per_second": 528.068,
376
+ "eval_steps_per_second": 16.796,
377
+ "step": 4600
378
+ },
379
+ {
380
+ "epoch": 5.14,
381
+ "grad_norm": 1.155013084411621,
382
+ "learning_rate": 0.00024768022840827977,
383
+ "loss": 0.4645,
384
+ "step": 4800
385
+ },
386
+ {
387
+ "epoch": 5.14,
388
+ "eval_accuracy": 0.8040712468193384,
389
+ "eval_loss": 0.5186554789543152,
390
+ "eval_runtime": 2.9698,
391
+ "eval_samples_per_second": 529.333,
392
+ "eval_steps_per_second": 16.836,
393
+ "step": 4800
394
+ },
395
+ {
396
+ "epoch": 5.35,
397
+ "grad_norm": 1.6959339380264282,
398
+ "learning_rate": 0.00024411134903640257,
399
+ "loss": 0.445,
400
+ "step": 5000
401
+ },
402
+ {
403
+ "epoch": 5.35,
404
+ "eval_accuracy": 0.806615776081425,
405
+ "eval_loss": 0.4928103983402252,
406
+ "eval_runtime": 2.9782,
407
+ "eval_samples_per_second": 527.83,
408
+ "eval_steps_per_second": 16.789,
409
+ "step": 5000
410
+ },
411
+ {
412
+ "epoch": 5.57,
413
+ "grad_norm": 1.0461735725402832,
414
+ "learning_rate": 0.00024054246966452532,
415
+ "loss": 0.4558,
416
+ "step": 5200
417
+ },
418
+ {
419
+ "epoch": 5.57,
420
+ "eval_accuracy": 0.8078880407124682,
421
+ "eval_loss": 0.48704999685287476,
422
+ "eval_runtime": 2.9838,
423
+ "eval_samples_per_second": 526.839,
424
+ "eval_steps_per_second": 16.757,
425
+ "step": 5200
426
+ },
427
+ {
428
+ "epoch": 5.78,
429
+ "grad_norm": 0.9599233269691467,
430
+ "learning_rate": 0.00023697359029264806,
431
+ "loss": 0.4405,
432
+ "step": 5400
433
+ },
434
+ {
435
+ "epoch": 5.78,
436
+ "eval_accuracy": 0.8104325699745547,
437
+ "eval_loss": 0.4985482692718506,
438
+ "eval_runtime": 3.0065,
439
+ "eval_samples_per_second": 522.862,
440
+ "eval_steps_per_second": 16.63,
441
+ "step": 5400
442
+ },
443
+ {
444
+ "epoch": 6.0,
445
+ "grad_norm": 1.4131615161895752,
446
+ "learning_rate": 0.00023340471092077086,
447
+ "loss": 0.4648,
448
+ "step": 5600
449
+ },
450
+ {
451
+ "epoch": 6.0,
452
+ "eval_accuracy": 0.8059796437659033,
453
+ "eval_loss": 0.48415422439575195,
454
+ "eval_runtime": 2.9786,
455
+ "eval_samples_per_second": 527.759,
456
+ "eval_steps_per_second": 16.786,
457
+ "step": 5600
458
+ },
459
+ {
460
+ "epoch": 6.21,
461
+ "grad_norm": 1.189572811126709,
462
+ "learning_rate": 0.0002298358315488936,
463
+ "loss": 0.435,
464
+ "step": 5800
465
+ },
466
+ {
467
+ "epoch": 6.21,
468
+ "eval_accuracy": 0.811704834605598,
469
+ "eval_loss": 0.4911487102508545,
470
+ "eval_runtime": 2.9997,
471
+ "eval_samples_per_second": 524.044,
472
+ "eval_steps_per_second": 16.668,
473
+ "step": 5800
474
+ },
475
+ {
476
+ "epoch": 6.42,
477
+ "grad_norm": 1.5198345184326172,
478
+ "learning_rate": 0.00022626695217701638,
479
+ "loss": 0.437,
480
+ "step": 6000
481
+ },
482
+ {
483
+ "epoch": 6.42,
484
+ "eval_accuracy": 0.8085241730279898,
485
+ "eval_loss": 0.48542749881744385,
486
+ "eval_runtime": 3.0042,
487
+ "eval_samples_per_second": 523.274,
488
+ "eval_steps_per_second": 16.644,
489
+ "step": 6000
490
+ },
491
+ {
492
+ "epoch": 6.64,
493
+ "grad_norm": 1.1990240812301636,
494
+ "learning_rate": 0.00022269807280513918,
495
+ "loss": 0.4588,
496
+ "step": 6200
497
+ },
498
+ {
499
+ "epoch": 6.64,
500
+ "eval_accuracy": 0.8085241730279898,
501
+ "eval_loss": 0.48791924118995667,
502
+ "eval_runtime": 3.0014,
503
+ "eval_samples_per_second": 523.758,
504
+ "eval_steps_per_second": 16.659,
505
+ "step": 6200
506
+ },
507
+ {
508
+ "epoch": 6.85,
509
+ "grad_norm": 1.346658706665039,
510
+ "learning_rate": 0.00021912919343326193,
511
+ "loss": 0.4342,
512
+ "step": 6400
513
+ },
514
+ {
515
+ "epoch": 6.85,
516
+ "eval_accuracy": 0.8104325699745547,
517
+ "eval_loss": 0.49220582842826843,
518
+ "eval_runtime": 3.0046,
519
+ "eval_samples_per_second": 523.193,
520
+ "eval_steps_per_second": 16.641,
521
+ "step": 6400
522
+ },
523
+ {
524
+ "epoch": 7.07,
525
+ "grad_norm": 1.8644700050354004,
526
+ "learning_rate": 0.00021556031406138473,
527
+ "loss": 0.4347,
528
+ "step": 6600
529
+ },
530
+ {
531
+ "epoch": 7.07,
532
+ "eval_accuracy": 0.8142493638676844,
533
+ "eval_loss": 0.49111655354499817,
534
+ "eval_runtime": 2.985,
535
+ "eval_samples_per_second": 526.634,
536
+ "eval_steps_per_second": 16.75,
537
+ "step": 6600
538
+ },
539
+ {
540
+ "epoch": 7.28,
541
+ "grad_norm": 1.9364045858383179,
542
+ "learning_rate": 0.00021199143468950748,
543
+ "loss": 0.4326,
544
+ "step": 6800
545
+ },
546
+ {
547
+ "epoch": 7.28,
548
+ "eval_accuracy": 0.8078880407124682,
549
+ "eval_loss": 0.491384893655777,
550
+ "eval_runtime": 2.9723,
551
+ "eval_samples_per_second": 528.882,
552
+ "eval_steps_per_second": 16.822,
553
+ "step": 6800
554
+ },
555
+ {
556
+ "epoch": 7.49,
557
+ "grad_norm": 0.9911957383155823,
558
+ "learning_rate": 0.00020842255531763022,
559
+ "loss": 0.4267,
560
+ "step": 7000
561
+ },
562
+ {
563
+ "epoch": 7.49,
564
+ "eval_accuracy": 0.8104325699745547,
565
+ "eval_loss": 0.4917159080505371,
566
+ "eval_runtime": 2.9808,
567
+ "eval_samples_per_second": 527.373,
568
+ "eval_steps_per_second": 16.774,
569
+ "step": 7000
570
+ },
571
+ {
572
+ "epoch": 7.71,
573
+ "grad_norm": 1.2186638116836548,
574
+ "learning_rate": 0.00020485367594575302,
575
+ "loss": 0.4241,
576
+ "step": 7200
577
+ },
578
+ {
579
+ "epoch": 7.71,
580
+ "eval_accuracy": 0.8136132315521628,
581
+ "eval_loss": 0.4887010455131531,
582
+ "eval_runtime": 2.9872,
583
+ "eval_samples_per_second": 526.253,
584
+ "eval_steps_per_second": 16.738,
585
+ "step": 7200
586
+ },
587
+ {
588
+ "epoch": 7.92,
589
+ "grad_norm": 1.1467108726501465,
590
+ "learning_rate": 0.0002012847965738758,
591
+ "loss": 0.4376,
592
+ "step": 7400
593
+ },
594
+ {
595
+ "epoch": 7.92,
596
+ "eval_accuracy": 0.8078880407124682,
597
+ "eval_loss": 0.5122085213661194,
598
+ "eval_runtime": 2.9829,
599
+ "eval_samples_per_second": 527.007,
600
+ "eval_steps_per_second": 16.762,
601
+ "step": 7400
602
+ },
603
+ {
604
+ "epoch": 8.14,
605
+ "grad_norm": 0.8427834510803223,
606
+ "learning_rate": 0.00019771591720199854,
607
+ "loss": 0.4323,
608
+ "step": 7600
609
+ },
610
+ {
611
+ "epoch": 8.14,
612
+ "eval_accuracy": 0.8097964376590331,
613
+ "eval_loss": 0.49093857407569885,
614
+ "eval_runtime": 2.9738,
615
+ "eval_samples_per_second": 528.625,
616
+ "eval_steps_per_second": 16.814,
617
+ "step": 7600
618
+ },
619
+ {
620
+ "epoch": 8.35,
621
+ "grad_norm": 1.2060902118682861,
622
+ "learning_rate": 0.00019414703783012134,
623
+ "loss": 0.4264,
624
+ "step": 7800
625
+ },
626
+ {
627
+ "epoch": 8.35,
628
+ "eval_accuracy": 0.8142493638676844,
629
+ "eval_loss": 0.48821595311164856,
630
+ "eval_runtime": 2.9836,
631
+ "eval_samples_per_second": 526.88,
632
+ "eval_steps_per_second": 16.758,
633
+ "step": 7800
634
+ },
635
+ {
636
+ "epoch": 8.57,
637
+ "grad_norm": 1.7033394575119019,
638
+ "learning_rate": 0.0001905781584582441,
639
+ "loss": 0.4175,
640
+ "step": 8000
641
+ },
642
+ {
643
+ "epoch": 8.57,
644
+ "eval_accuracy": 0.8053435114503816,
645
+ "eval_loss": 0.5090692043304443,
646
+ "eval_runtime": 2.9978,
647
+ "eval_samples_per_second": 524.393,
648
+ "eval_steps_per_second": 16.679,
649
+ "step": 8000
650
+ },
651
+ {
652
+ "epoch": 8.78,
653
+ "grad_norm": 1.3033976554870605,
654
+ "learning_rate": 0.0001870092790863669,
655
+ "loss": 0.4228,
656
+ "step": 8200
657
+ },
658
+ {
659
+ "epoch": 8.78,
660
+ "eval_accuracy": 0.8097964376590331,
661
+ "eval_loss": 0.5060204863548279,
662
+ "eval_runtime": 2.9975,
663
+ "eval_samples_per_second": 524.436,
664
+ "eval_steps_per_second": 16.681,
665
+ "step": 8200
666
+ },
667
+ {
668
+ "epoch": 8.99,
669
+ "grad_norm": 1.2635438442230225,
670
+ "learning_rate": 0.00018344039971448964,
671
+ "loss": 0.4189,
672
+ "step": 8400
673
+ },
674
+ {
675
+ "epoch": 8.99,
676
+ "eval_accuracy": 0.8091603053435115,
677
+ "eval_loss": 0.4940575361251831,
678
+ "eval_runtime": 2.9634,
679
+ "eval_samples_per_second": 530.468,
680
+ "eval_steps_per_second": 16.872,
681
+ "step": 8400
682
+ },
683
+ {
684
+ "epoch": 9.21,
685
+ "grad_norm": 1.496982455253601,
686
+ "learning_rate": 0.0001798715203426124,
687
+ "loss": 0.4161,
688
+ "step": 8600
689
+ },
690
+ {
691
+ "epoch": 9.21,
692
+ "eval_accuracy": 0.8174300254452926,
693
+ "eval_loss": 0.5010442137718201,
694
+ "eval_runtime": 2.973,
695
+ "eval_samples_per_second": 528.758,
696
+ "eval_steps_per_second": 16.818,
697
+ "step": 8600
698
+ },
699
+ {
700
+ "epoch": 9.42,
701
+ "grad_norm": 1.355362892150879,
702
+ "learning_rate": 0.00017630264097073518,
703
+ "loss": 0.4078,
704
+ "step": 8800
705
+ },
706
+ {
707
+ "epoch": 9.42,
708
+ "eval_accuracy": 0.8078880407124682,
709
+ "eval_loss": 0.4949406683444977,
710
+ "eval_runtime": 2.9901,
711
+ "eval_samples_per_second": 525.736,
712
+ "eval_steps_per_second": 16.722,
713
+ "step": 8800
714
+ },
715
+ {
716
+ "epoch": 9.64,
717
+ "grad_norm": 1.180076241493225,
718
+ "learning_rate": 0.00017273376159885795,
719
+ "loss": 0.4201,
720
+ "step": 9000
721
+ },
722
+ {
723
+ "epoch": 9.64,
724
+ "eval_accuracy": 0.8072519083969466,
725
+ "eval_loss": 0.5017107129096985,
726
+ "eval_runtime": 2.952,
727
+ "eval_samples_per_second": 532.525,
728
+ "eval_steps_per_second": 16.938,
729
+ "step": 9000
730
+ },
731
+ {
732
+ "epoch": 9.85,
733
+ "grad_norm": 1.1020286083221436,
734
+ "learning_rate": 0.0001691648822269807,
735
+ "loss": 0.4141,
736
+ "step": 9200
737
+ },
738
+ {
739
+ "epoch": 9.85,
740
+ "eval_accuracy": 0.8091603053435115,
741
+ "eval_loss": 0.4984731078147888,
742
+ "eval_runtime": 2.9633,
743
+ "eval_samples_per_second": 530.497,
744
+ "eval_steps_per_second": 16.873,
745
+ "step": 9200
746
+ },
747
+ {
748
+ "epoch": 10.06,
749
+ "grad_norm": 1.2666047811508179,
750
+ "learning_rate": 0.0001655960028551035,
751
+ "loss": 0.4132,
752
+ "step": 9400
753
+ },
754
+ {
755
+ "epoch": 10.06,
756
+ "eval_accuracy": 0.8053435114503816,
757
+ "eval_loss": 0.5031649470329285,
758
+ "eval_runtime": 2.9822,
759
+ "eval_samples_per_second": 527.133,
760
+ "eval_steps_per_second": 16.766,
761
+ "step": 9400
762
+ },
763
+ {
764
+ "epoch": 10.28,
765
+ "grad_norm": 0.6767197251319885,
766
+ "learning_rate": 0.00016202712348322625,
767
+ "loss": 0.4043,
768
+ "step": 9600
769
+ },
770
+ {
771
+ "epoch": 10.28,
772
+ "eval_accuracy": 0.8129770992366412,
773
+ "eval_loss": 0.5038406848907471,
774
+ "eval_runtime": 2.9816,
775
+ "eval_samples_per_second": 527.24,
776
+ "eval_steps_per_second": 16.77,
777
+ "step": 9600
778
+ },
779
+ {
780
+ "epoch": 10.49,
781
+ "grad_norm": 1.147275447845459,
782
+ "learning_rate": 0.00015845824411134902,
783
+ "loss": 0.4187,
784
+ "step": 9800
785
+ },
786
+ {
787
+ "epoch": 10.49,
788
+ "eval_accuracy": 0.8104325699745547,
789
+ "eval_loss": 0.4981047213077545,
790
+ "eval_runtime": 2.9858,
791
+ "eval_samples_per_second": 526.485,
792
+ "eval_steps_per_second": 16.746,
793
+ "step": 9800
794
+ },
795
+ {
796
+ "epoch": 10.71,
797
+ "grad_norm": 1.6172677278518677,
798
+ "learning_rate": 0.0001548893647394718,
799
+ "loss": 0.3827,
800
+ "step": 10000
801
+ },
802
+ {
803
+ "epoch": 10.71,
804
+ "eval_accuracy": 0.8072519083969466,
805
+ "eval_loss": 0.5126467943191528,
806
+ "eval_runtime": 2.9825,
807
+ "eval_samples_per_second": 527.072,
808
+ "eval_steps_per_second": 16.764,
809
+ "step": 10000
810
+ },
811
+ {
812
+ "epoch": 10.92,
813
+ "grad_norm": 1.8639923334121704,
814
+ "learning_rate": 0.00015132048536759457,
815
+ "loss": 0.4074,
816
+ "step": 10200
817
+ },
818
+ {
819
+ "epoch": 10.92,
820
+ "eval_accuracy": 0.8072519083969466,
821
+ "eval_loss": 0.5088323950767517,
822
+ "eval_runtime": 2.9816,
823
+ "eval_samples_per_second": 527.237,
824
+ "eval_steps_per_second": 16.77,
825
+ "step": 10200
826
+ },
827
+ {
828
+ "epoch": 11.13,
829
+ "grad_norm": 1.2519667148590088,
830
+ "learning_rate": 0.00014775160599571734,
831
+ "loss": 0.4013,
832
+ "step": 10400
833
+ },
834
+ {
835
+ "epoch": 11.13,
836
+ "eval_accuracy": 0.8072519083969466,
837
+ "eval_loss": 0.5061373114585876,
838
+ "eval_runtime": 2.9811,
839
+ "eval_samples_per_second": 527.316,
840
+ "eval_steps_per_second": 16.772,
841
+ "step": 10400
842
+ },
843
+ {
844
+ "epoch": 11.35,
845
+ "grad_norm": 1.1711052656173706,
846
+ "learning_rate": 0.0001441827266238401,
847
+ "loss": 0.3888,
848
+ "step": 10600
849
+ },
850
+ {
851
+ "epoch": 11.35,
852
+ "eval_accuracy": 0.8085241730279898,
853
+ "eval_loss": 0.5013065338134766,
854
+ "eval_runtime": 2.9847,
855
+ "eval_samples_per_second": 526.681,
856
+ "eval_steps_per_second": 16.752,
857
+ "step": 10600
858
+ },
859
+ {
860
+ "epoch": 11.56,
861
+ "grad_norm": 1.8078001737594604,
862
+ "learning_rate": 0.00014061384725196286,
863
+ "loss": 0.3855,
864
+ "step": 10800
865
+ },
866
+ {
867
+ "epoch": 11.56,
868
+ "eval_accuracy": 0.8059796437659033,
869
+ "eval_loss": 0.4992610514163971,
870
+ "eval_runtime": 2.9927,
871
+ "eval_samples_per_second": 525.27,
872
+ "eval_steps_per_second": 16.707,
873
+ "step": 10800
874
+ },
875
+ {
876
+ "epoch": 11.78,
877
+ "grad_norm": 1.1071592569351196,
878
+ "learning_rate": 0.00013704496788008563,
879
+ "loss": 0.3924,
880
+ "step": 11000
881
+ },
882
+ {
883
+ "epoch": 11.78,
884
+ "eval_accuracy": 0.8085241730279898,
885
+ "eval_loss": 0.5075262188911438,
886
+ "eval_runtime": 3.0066,
887
+ "eval_samples_per_second": 522.844,
888
+ "eval_steps_per_second": 16.63,
889
+ "step": 11000
890
+ },
891
+ {
892
+ "epoch": 11.99,
893
+ "grad_norm": 1.3704427480697632,
894
+ "learning_rate": 0.0001334760885082084,
895
+ "loss": 0.4046,
896
+ "step": 11200
897
+ },
898
+ {
899
+ "epoch": 11.99,
900
+ "eval_accuracy": 0.8027989821882952,
901
+ "eval_loss": 0.49990707635879517,
902
+ "eval_runtime": 3.0049,
903
+ "eval_samples_per_second": 523.149,
904
+ "eval_steps_per_second": 16.64,
905
+ "step": 11200
906
+ },
907
+ {
908
+ "epoch": 12.21,
909
+ "grad_norm": 1.40170419216156,
910
+ "learning_rate": 0.00012990720913633118,
911
+ "loss": 0.3957,
912
+ "step": 11400
913
+ },
914
+ {
915
+ "epoch": 12.21,
916
+ "eval_accuracy": 0.8034351145038168,
917
+ "eval_loss": 0.5089264512062073,
918
+ "eval_runtime": 2.9942,
919
+ "eval_samples_per_second": 525.011,
920
+ "eval_steps_per_second": 16.699,
921
+ "step": 11400
922
+ },
923
+ {
924
+ "epoch": 12.42,
925
+ "grad_norm": 1.1885521411895752,
926
+ "learning_rate": 0.00012633832976445395,
927
+ "loss": 0.381,
928
+ "step": 11600
929
+ },
930
+ {
931
+ "epoch": 12.42,
932
+ "eval_accuracy": 0.8072519083969466,
933
+ "eval_loss": 0.5207549929618835,
934
+ "eval_runtime": 2.9746,
935
+ "eval_samples_per_second": 528.479,
936
+ "eval_steps_per_second": 16.809,
937
+ "step": 11600
938
+ },
939
+ {
940
+ "epoch": 12.63,
941
+ "grad_norm": 0.8873888254165649,
942
+ "learning_rate": 0.00012276945039257673,
943
+ "loss": 0.3906,
944
+ "step": 11800
945
+ },
946
+ {
947
+ "epoch": 12.63,
948
+ "eval_accuracy": 0.806615776081425,
949
+ "eval_loss": 0.513671875,
950
+ "eval_runtime": 2.961,
951
+ "eval_samples_per_second": 530.901,
952
+ "eval_steps_per_second": 16.886,
953
+ "step": 11800
954
+ },
955
+ {
956
+ "epoch": 12.85,
957
+ "grad_norm": 1.6491570472717285,
958
+ "learning_rate": 0.0001192005710206995,
959
+ "loss": 0.3734,
960
+ "step": 12000
961
+ },
962
+ {
963
+ "epoch": 12.85,
964
+ "eval_accuracy": 0.8040712468193384,
965
+ "eval_loss": 0.5183374881744385,
966
+ "eval_runtime": 2.9533,
967
+ "eval_samples_per_second": 532.292,
968
+ "eval_steps_per_second": 16.93,
969
+ "step": 12000
970
+ },
971
+ {
972
+ "epoch": 13.06,
973
+ "grad_norm": 2.042646884918213,
974
+ "learning_rate": 0.00011563169164882227,
975
+ "loss": 0.3928,
976
+ "step": 12200
977
+ },
978
+ {
979
+ "epoch": 13.06,
980
+ "eval_accuracy": 0.806615776081425,
981
+ "eval_loss": 0.5069447159767151,
982
+ "eval_runtime": 2.959,
983
+ "eval_samples_per_second": 531.259,
984
+ "eval_steps_per_second": 16.898,
985
+ "step": 12200
986
+ },
987
+ {
988
+ "epoch": 13.28,
989
+ "grad_norm": 0.817425549030304,
990
+ "learning_rate": 0.00011206281227694502,
991
+ "loss": 0.3774,
992
+ "step": 12400
993
+ },
994
+ {
995
+ "epoch": 13.28,
996
+ "eval_accuracy": 0.8008905852417303,
997
+ "eval_loss": 0.5086419582366943,
998
+ "eval_runtime": 2.9547,
999
+ "eval_samples_per_second": 532.04,
1000
+ "eval_steps_per_second": 16.922,
1001
+ "step": 12400
1002
+ },
1003
+ {
1004
+ "epoch": 13.49,
1005
+ "grad_norm": 1.0988578796386719,
1006
+ "learning_rate": 0.0001084939329050678,
1007
+ "loss": 0.3892,
1008
+ "step": 12600
1009
+ },
1010
+ {
1011
+ "epoch": 13.49,
1012
+ "eval_accuracy": 0.8059796437659033,
1013
+ "eval_loss": 0.4966925382614136,
1014
+ "eval_runtime": 2.9538,
1015
+ "eval_samples_per_second": 532.194,
1016
+ "eval_steps_per_second": 16.927,
1017
+ "step": 12600
1018
+ },
1019
+ {
1020
+ "epoch": 13.7,
1021
+ "grad_norm": 1.312321662902832,
1022
+ "learning_rate": 0.00010492505353319058,
1023
+ "loss": 0.372,
1024
+ "step": 12800
1025
+ },
1026
+ {
1027
+ "epoch": 13.7,
1028
+ "eval_accuracy": 0.8040712468193384,
1029
+ "eval_loss": 0.5042534470558167,
1030
+ "eval_runtime": 2.9651,
1031
+ "eval_samples_per_second": 530.16,
1032
+ "eval_steps_per_second": 16.863,
1033
+ "step": 12800
1034
+ },
1035
+ {
1036
+ "epoch": 13.92,
1037
+ "grad_norm": 1.642741322517395,
1038
+ "learning_rate": 0.00010135617416131332,
1039
+ "loss": 0.388,
1040
+ "step": 13000
1041
+ },
1042
+ {
1043
+ "epoch": 13.92,
1044
+ "eval_accuracy": 0.8072519083969466,
1045
+ "eval_loss": 0.5095480680465698,
1046
+ "eval_runtime": 2.9526,
1047
+ "eval_samples_per_second": 532.404,
1048
+ "eval_steps_per_second": 16.934,
1049
+ "step": 13000
1050
+ },
1051
+ {
1052
+ "epoch": 14.13,
1053
+ "grad_norm": 1.10377836227417,
1054
+ "learning_rate": 9.778729478943611e-05,
1055
+ "loss": 0.3754,
1056
+ "step": 13200
1057
+ },
1058
+ {
1059
+ "epoch": 14.13,
1060
+ "eval_accuracy": 0.8021628498727735,
1061
+ "eval_loss": 0.5103972554206848,
1062
+ "eval_runtime": 2.9663,
1063
+ "eval_samples_per_second": 529.954,
1064
+ "eval_steps_per_second": 16.856,
1065
+ "step": 13200
1066
+ },
1067
+ {
1068
+ "epoch": 14.35,
1069
+ "grad_norm": 1.1614229679107666,
1070
+ "learning_rate": 9.421841541755888e-05,
1071
+ "loss": 0.3639,
1072
+ "step": 13400
1073
+ },
1074
+ {
1075
+ "epoch": 14.35,
1076
+ "eval_accuracy": 0.7983460559796438,
1077
+ "eval_loss": 0.5263165235519409,
1078
+ "eval_runtime": 2.9391,
1079
+ "eval_samples_per_second": 534.858,
1080
+ "eval_steps_per_second": 17.012,
1081
+ "step": 13400
1082
+ },
1083
+ {
1084
+ "epoch": 14.56,
1085
+ "grad_norm": 1.6049692630767822,
1086
+ "learning_rate": 9.064953604568166e-05,
1087
+ "loss": 0.3795,
1088
+ "step": 13600
1089
+ },
1090
+ {
1091
+ "epoch": 14.56,
1092
+ "eval_accuracy": 0.8015267175572519,
1093
+ "eval_loss": 0.5145931839942932,
1094
+ "eval_runtime": 2.9465,
1095
+ "eval_samples_per_second": 533.506,
1096
+ "eval_steps_per_second": 16.969,
1097
+ "step": 13600
1098
+ },
1099
+ {
1100
+ "epoch": 14.78,
1101
+ "grad_norm": 2.813002347946167,
1102
+ "learning_rate": 8.708065667380442e-05,
1103
+ "loss": 0.3792,
1104
+ "step": 13800
1105
+ },
1106
+ {
1107
+ "epoch": 14.78,
1108
+ "eval_accuracy": 0.8040712468193384,
1109
+ "eval_loss": 0.5066380500793457,
1110
+ "eval_runtime": 2.9409,
1111
+ "eval_samples_per_second": 534.523,
1112
+ "eval_steps_per_second": 17.001,
1113
+ "step": 13800
1114
+ },
1115
+ {
1116
+ "epoch": 14.99,
1117
+ "grad_norm": 1.2670201063156128,
1118
+ "learning_rate": 8.351177730192719e-05,
1119
+ "loss": 0.3589,
1120
+ "step": 14000
1121
+ },
1122
+ {
1123
+ "epoch": 14.99,
1124
+ "eval_accuracy": 0.8078880407124682,
1125
+ "eval_loss": 0.5135853886604309,
1126
+ "eval_runtime": 2.962,
1127
+ "eval_samples_per_second": 530.717,
1128
+ "eval_steps_per_second": 16.88,
1129
+ "step": 14000
1130
+ },
1131
+ {
1132
+ "epoch": 15.2,
1133
+ "grad_norm": 1.9681557416915894,
1134
+ "learning_rate": 7.994289793004996e-05,
1135
+ "loss": 0.3624,
1136
+ "step": 14200
1137
+ },
1138
+ {
1139
+ "epoch": 15.2,
1140
+ "eval_accuracy": 0.8021628498727735,
1141
+ "eval_loss": 0.5237164497375488,
1142
+ "eval_runtime": 2.9535,
1143
+ "eval_samples_per_second": 532.245,
1144
+ "eval_steps_per_second": 16.929,
1145
+ "step": 14200
1146
+ },
1147
+ {
1148
+ "epoch": 15.42,
1149
+ "grad_norm": 1.8548041582107544,
1150
+ "learning_rate": 7.637401855817274e-05,
1151
+ "loss": 0.3659,
1152
+ "step": 14400
1153
+ },
1154
+ {
1155
+ "epoch": 15.42,
1156
+ "eval_accuracy": 0.8059796437659033,
1157
+ "eval_loss": 0.5165674090385437,
1158
+ "eval_runtime": 2.9482,
1159
+ "eval_samples_per_second": 533.2,
1160
+ "eval_steps_per_second": 16.959,
1161
+ "step": 14400
1162
+ },
1163
+ {
1164
+ "epoch": 15.63,
1165
+ "grad_norm": 1.3727173805236816,
1166
+ "learning_rate": 7.28051391862955e-05,
1167
+ "loss": 0.3657,
1168
+ "step": 14600
1169
+ },
1170
+ {
1171
+ "epoch": 15.63,
1172
+ "eval_accuracy": 0.8002544529262087,
1173
+ "eval_loss": 0.5177738070487976,
1174
+ "eval_runtime": 2.9451,
1175
+ "eval_samples_per_second": 533.764,
1176
+ "eval_steps_per_second": 16.977,
1177
+ "step": 14600
1178
+ },
1179
+ {
1180
+ "epoch": 15.85,
1181
+ "grad_norm": 2.10198974609375,
1182
+ "learning_rate": 6.923625981441827e-05,
1183
+ "loss": 0.359,
1184
+ "step": 14800
1185
+ },
1186
+ {
1187
+ "epoch": 15.85,
1188
+ "eval_accuracy": 0.7983460559796438,
1189
+ "eval_loss": 0.5152426362037659,
1190
+ "eval_runtime": 2.9473,
1191
+ "eval_samples_per_second": 533.372,
1192
+ "eval_steps_per_second": 16.965,
1193
+ "step": 14800
1194
+ },
1195
+ {
1196
+ "epoch": 16.06,
1197
+ "grad_norm": 1.0453667640686035,
1198
+ "learning_rate": 6.566738044254104e-05,
1199
+ "loss": 0.3677,
1200
+ "step": 15000
1201
+ },
1202
+ {
1203
+ "epoch": 16.06,
1204
+ "eval_accuracy": 0.8034351145038168,
1205
+ "eval_loss": 0.5211815237998962,
1206
+ "eval_runtime": 2.9478,
1207
+ "eval_samples_per_second": 533.274,
1208
+ "eval_steps_per_second": 16.962,
1209
+ "step": 15000
1210
+ },
1211
+ {
1212
+ "epoch": 16.27,
1213
+ "grad_norm": 1.0645538568496704,
1214
+ "learning_rate": 6.20985010706638e-05,
1215
+ "loss": 0.3521,
1216
+ "step": 15200
1217
+ },
1218
+ {
1219
+ "epoch": 16.27,
1220
+ "eval_accuracy": 0.8002544529262087,
1221
+ "eval_loss": 0.5323696732521057,
1222
+ "eval_runtime": 2.9594,
1223
+ "eval_samples_per_second": 531.197,
1224
+ "eval_steps_per_second": 16.896,
1225
+ "step": 15200
1226
+ },
1227
+ {
1228
+ "epoch": 16.49,
1229
+ "grad_norm": 3.849015951156616,
1230
+ "learning_rate": 5.852962169878657e-05,
1231
+ "loss": 0.3589,
1232
+ "step": 15400
1233
+ },
1234
+ {
1235
+ "epoch": 16.49,
1236
+ "eval_accuracy": 0.8040712468193384,
1237
+ "eval_loss": 0.5237988829612732,
1238
+ "eval_runtime": 2.9364,
1239
+ "eval_samples_per_second": 535.357,
1240
+ "eval_steps_per_second": 17.028,
1241
+ "step": 15400
1242
+ },
1243
+ {
1244
+ "epoch": 16.7,
1245
+ "grad_norm": 1.3231987953186035,
1246
+ "learning_rate": 5.496074232690935e-05,
1247
+ "loss": 0.3695,
1248
+ "step": 15600
1249
+ },
1250
+ {
1251
+ "epoch": 16.7,
1252
+ "eval_accuracy": 0.7977099236641222,
1253
+ "eval_loss": 0.511340320110321,
1254
+ "eval_runtime": 2.969,
1255
+ "eval_samples_per_second": 529.468,
1256
+ "eval_steps_per_second": 16.841,
1257
+ "step": 15600
1258
+ },
1259
+ {
1260
+ "epoch": 16.92,
1261
+ "grad_norm": 1.7709985971450806,
1262
+ "learning_rate": 5.139186295503211e-05,
1263
+ "loss": 0.3606,
1264
+ "step": 15800
1265
+ },
1266
+ {
1267
+ "epoch": 16.92,
1268
+ "eval_accuracy": 0.7983460559796438,
1269
+ "eval_loss": 0.5136662721633911,
1270
+ "eval_runtime": 2.9594,
1271
+ "eval_samples_per_second": 531.193,
1272
+ "eval_steps_per_second": 16.895,
1273
+ "step": 15800
1274
+ },
1275
+ {
1276
+ "epoch": 17.13,
1277
+ "grad_norm": 1.5108495950698853,
1278
+ "learning_rate": 4.782298358315489e-05,
1279
+ "loss": 0.3581,
1280
+ "step": 16000
1281
+ },
1282
+ {
1283
+ "epoch": 17.13,
1284
+ "eval_accuracy": 0.799618320610687,
1285
+ "eval_loss": 0.5130853056907654,
1286
+ "eval_runtime": 2.9611,
1287
+ "eval_samples_per_second": 530.882,
1288
+ "eval_steps_per_second": 16.886,
1289
+ "step": 16000
1290
+ },
1291
+ {
1292
+ "epoch": 17.34,
1293
+ "grad_norm": 1.3634617328643799,
1294
+ "learning_rate": 4.4254104211277655e-05,
1295
+ "loss": 0.3488,
1296
+ "step": 16200
1297
+ },
1298
+ {
1299
+ "epoch": 17.34,
1300
+ "eval_accuracy": 0.7989821882951654,
1301
+ "eval_loss": 0.5270070433616638,
1302
+ "eval_runtime": 2.9953,
1303
+ "eval_samples_per_second": 524.824,
1304
+ "eval_steps_per_second": 16.693,
1305
+ "step": 16200
1306
+ },
1307
+ {
1308
+ "epoch": 17.56,
1309
+ "grad_norm": 1.0239213705062866,
1310
+ "learning_rate": 4.068522483940043e-05,
1311
+ "loss": 0.3499,
1312
+ "step": 16400
1313
+ },
1314
+ {
1315
+ "epoch": 17.56,
1316
+ "eval_accuracy": 0.7964376590330788,
1317
+ "eval_loss": 0.523576021194458,
1318
+ "eval_runtime": 2.9356,
1319
+ "eval_samples_per_second": 535.502,
1320
+ "eval_steps_per_second": 17.033,
1321
+ "step": 16400
1322
+ },
1323
+ {
1324
+ "epoch": 17.77,
1325
+ "grad_norm": 1.108484148979187,
1326
+ "learning_rate": 3.7116345467523195e-05,
1327
+ "loss": 0.3603,
1328
+ "step": 16600
1329
+ },
1330
+ {
1331
+ "epoch": 17.77,
1332
+ "eval_accuracy": 0.8002544529262087,
1333
+ "eval_loss": 0.5186541080474854,
1334
+ "eval_runtime": 2.9666,
1335
+ "eval_samples_per_second": 529.891,
1336
+ "eval_steps_per_second": 16.854,
1337
+ "step": 16600
1338
+ },
1339
+ {
1340
+ "epoch": 17.99,
1341
+ "grad_norm": 2.816092014312744,
1342
+ "learning_rate": 3.354746609564596e-05,
1343
+ "loss": 0.3578,
1344
+ "step": 16800
1345
+ },
1346
+ {
1347
+ "epoch": 17.99,
1348
+ "eval_accuracy": 0.8021628498727735,
1349
+ "eval_loss": 0.5223926901817322,
1350
+ "eval_runtime": 2.9355,
1351
+ "eval_samples_per_second": 535.521,
1352
+ "eval_steps_per_second": 17.033,
1353
+ "step": 16800
1354
+ },
1355
+ {
1356
+ "epoch": 18.2,
1357
+ "grad_norm": 1.5831489562988281,
1358
+ "learning_rate": 2.997858672376873e-05,
1359
+ "loss": 0.3449,
1360
+ "step": 17000
1361
+ },
1362
+ {
1363
+ "epoch": 18.2,
1364
+ "eval_accuracy": 0.7989821882951654,
1365
+ "eval_loss": 0.5227622389793396,
1366
+ "eval_runtime": 2.9602,
1367
+ "eval_samples_per_second": 531.048,
1368
+ "eval_steps_per_second": 16.891,
1369
+ "step": 17000
1370
+ },
1371
+ {
1372
+ "epoch": 18.42,
1373
+ "grad_norm": 1.0060327053070068,
1374
+ "learning_rate": 2.64097073518915e-05,
1375
+ "loss": 0.3418,
1376
+ "step": 17200
1377
+ },
1378
+ {
1379
+ "epoch": 18.42,
1380
+ "eval_accuracy": 0.8008905852417303,
1381
+ "eval_loss": 0.5287216901779175,
1382
+ "eval_runtime": 2.9537,
1383
+ "eval_samples_per_second": 532.21,
1384
+ "eval_steps_per_second": 16.928,
1385
+ "step": 17200
1386
+ },
1387
+ {
1388
+ "epoch": 18.63,
1389
+ "grad_norm": 1.8092093467712402,
1390
+ "learning_rate": 2.2840827980014274e-05,
1391
+ "loss": 0.3334,
1392
+ "step": 17400
1393
+ },
1394
+ {
1395
+ "epoch": 18.63,
1396
+ "eval_accuracy": 0.799618320610687,
1397
+ "eval_loss": 0.5322315096855164,
1398
+ "eval_runtime": 2.9745,
1399
+ "eval_samples_per_second": 528.484,
1400
+ "eval_steps_per_second": 16.809,
1401
+ "step": 17400
1402
+ },
1403
+ {
1404
+ "epoch": 18.84,
1405
+ "grad_norm": 1.4800430536270142,
1406
+ "learning_rate": 1.9271948608137044e-05,
1407
+ "loss": 0.3567,
1408
+ "step": 17600
1409
+ },
1410
+ {
1411
+ "epoch": 18.84,
1412
+ "eval_accuracy": 0.7983460559796438,
1413
+ "eval_loss": 0.5293812155723572,
1414
+ "eval_runtime": 2.9485,
1415
+ "eval_samples_per_second": 533.161,
1416
+ "eval_steps_per_second": 16.958,
1417
+ "step": 17600
1418
+ },
1419
+ {
1420
+ "epoch": 19.06,
1421
+ "grad_norm": 1.6271811723709106,
1422
+ "learning_rate": 1.5703069236259814e-05,
1423
+ "loss": 0.3541,
1424
+ "step": 17800
1425
+ },
1426
+ {
1427
+ "epoch": 19.06,
1428
+ "eval_accuracy": 0.8002544529262087,
1429
+ "eval_loss": 0.5250320434570312,
1430
+ "eval_runtime": 2.9479,
1431
+ "eval_samples_per_second": 533.268,
1432
+ "eval_steps_per_second": 16.961,
1433
+ "step": 17800
1434
+ },
1435
+ {
1436
+ "epoch": 19.27,
1437
+ "grad_norm": 0.7758527994155884,
1438
+ "learning_rate": 1.2134189864382584e-05,
1439
+ "loss": 0.365,
1440
+ "step": 18000
1441
+ },
1442
+ {
1443
+ "epoch": 19.27,
1444
+ "eval_accuracy": 0.7983460559796438,
1445
+ "eval_loss": 0.5246437788009644,
1446
+ "eval_runtime": 2.9363,
1447
+ "eval_samples_per_second": 535.369,
1448
+ "eval_steps_per_second": 17.028,
1449
+ "step": 18000
1450
+ },
1451
+ {
1452
+ "epoch": 19.49,
1453
+ "grad_norm": 0.9722337126731873,
1454
+ "learning_rate": 8.565310492505352e-06,
1455
+ "loss": 0.337,
1456
+ "step": 18200
1457
+ },
1458
+ {
1459
+ "epoch": 19.49,
1460
+ "eval_accuracy": 0.7977099236641222,
1461
+ "eval_loss": 0.527810275554657,
1462
+ "eval_runtime": 2.9383,
1463
+ "eval_samples_per_second": 535.006,
1464
+ "eval_steps_per_second": 17.017,
1465
+ "step": 18200
1466
+ },
1467
+ {
1468
+ "epoch": 19.7,
1469
+ "grad_norm": 1.5007203817367554,
1470
+ "learning_rate": 4.996431120628123e-06,
1471
+ "loss": 0.3301,
1472
+ "step": 18400
1473
+ },
1474
+ {
1475
+ "epoch": 19.7,
1476
+ "eval_accuracy": 0.7989821882951654,
1477
+ "eval_loss": 0.5283259153366089,
1478
+ "eval_runtime": 2.9603,
1479
+ "eval_samples_per_second": 531.035,
1480
+ "eval_steps_per_second": 16.89,
1481
+ "step": 18400
1482
+ },
1483
+ {
1484
+ "epoch": 19.91,
1485
+ "grad_norm": 1.1220752000808716,
1486
+ "learning_rate": 1.4275517487508921e-06,
1487
+ "loss": 0.3421,
1488
+ "step": 18600
1489
+ },
1490
+ {
1491
+ "epoch": 19.91,
1492
+ "eval_accuracy": 0.7977099236641222,
1493
+ "eval_loss": 0.5287136435508728,
1494
+ "eval_runtime": 2.9398,
1495
+ "eval_samples_per_second": 534.737,
1496
+ "eval_steps_per_second": 17.008,
1497
+ "step": 18600
1498
+ },
1499
+ {
1500
+ "epoch": 20.0,
1501
+ "step": 18680,
1502
+ "total_flos": 1.44512252251488e+16,
1503
+ "train_loss": 0.42864556159401346,
1504
+ "train_runtime": 2680.553,
1505
+ "train_samples_per_second": 222.82,
1506
+ "train_steps_per_second": 6.969
1507
+ }
1508
+ ],
1509
+ "logging_steps": 200,
1510
+ "max_steps": 18680,
1511
+ "num_input_tokens_seen": 0,
1512
+ "num_train_epochs": 20,
1513
+ "save_steps": 200,
1514
+ "total_flos": 1.44512252251488e+16,
1515
+ "train_batch_size": 32,
1516
+ "trial_name": null,
1517
+ "trial_params": null
1518
+ }