jbostock commited on
Commit
035559f
·
verified ·
1 Parent(s): 2459856

Upload mid-g1/trainer_state.json with huggingface_hub

Browse files
Files changed (1) hide show
  1. mid-g1/trainer_state.json +888 -0
mid-g1/trainer_state.json ADDED
@@ -0,0 +1,888 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": null,
3
+ "best_metric": null,
4
+ "best_model_checkpoint": null,
5
+ "epoch": 0.9964267483409903,
6
+ "eval_steps": 500,
7
+ "global_step": 61,
8
+ "is_hyper_param_search": false,
9
+ "is_local_process_zero": true,
10
+ "is_world_process_zero": true,
11
+ "log_history": [
12
+ {
13
+ "epoch": 0.01633486472690148,
14
+ "grad_norm": 3.71875,
15
+ "learning_rate": 0.0,
16
+ "loss": 2.0191650390625,
17
+ "memory/device_reserved (GiB)": 53.69,
18
+ "memory/max_active (GiB)": 47.42,
19
+ "memory/max_allocated (GiB)": 47.42,
20
+ "ppl": 7.53203,
21
+ "step": 1,
22
+ "tokens/total": 524288,
23
+ "tokens/train_per_sec_per_gpu": 285.34,
24
+ "tokens/trainable": 524065
25
+ },
26
+ {
27
+ "epoch": 0.03266972945380296,
28
+ "grad_norm": 4.84375,
29
+ "learning_rate": 3.3333333333333333e-06,
30
+ "loss": 1.9945068359375,
31
+ "memory/device_reserved (GiB)": 60.86,
32
+ "memory/max_active (GiB)": 54.65,
33
+ "memory/max_allocated (GiB)": 54.65,
34
+ "ppl": 7.34858,
35
+ "step": 2,
36
+ "tokens/total": 1048576,
37
+ "tokens/train_per_sec_per_gpu": 316.47,
38
+ "tokens/trainable": 1047778
39
+ },
40
+ {
41
+ "epoch": 0.04900459418070444,
42
+ "grad_norm": 3.6875,
43
+ "learning_rate": 6.666666666666667e-06,
44
+ "loss": 2.025390625,
45
+ "memory/device_reserved (GiB)": 60.86,
46
+ "memory/max_active (GiB)": 54.65,
47
+ "memory/max_allocated (GiB)": 54.65,
48
+ "ppl": 7.57907,
49
+ "step": 3,
50
+ "tokens/total": 1572864,
51
+ "tokens/train_per_sec_per_gpu": 314.31,
52
+ "tokens/trainable": 1571485
53
+ },
54
+ {
55
+ "epoch": 0.06533945890760592,
56
+ "grad_norm": 2.453125,
57
+ "learning_rate": 1e-05,
58
+ "loss": 1.98974609375,
59
+ "memory/device_reserved (GiB)": 60.86,
60
+ "memory/max_active (GiB)": 54.65,
61
+ "memory/max_allocated (GiB)": 54.65,
62
+ "ppl": 7.31368,
63
+ "step": 4,
64
+ "tokens/total": 2097152,
65
+ "tokens/train_per_sec_per_gpu": 314.27,
66
+ "tokens/trainable": 2095162
67
+ },
68
+ {
69
+ "epoch": 0.0816743236345074,
70
+ "grad_norm": 3.3125,
71
+ "learning_rate": 9.993400362330058e-06,
72
+ "loss": 1.9696044921875,
73
+ "memory/device_reserved (GiB)": 60.86,
74
+ "memory/max_active (GiB)": 54.65,
75
+ "memory/max_allocated (GiB)": 54.65,
76
+ "ppl": 7.16784,
77
+ "step": 5,
78
+ "tokens/total": 2621440,
79
+ "tokens/train_per_sec_per_gpu": 317.21,
80
+ "tokens/trainable": 2618849
81
+ },
82
+ {
83
+ "epoch": 0.09800918836140889,
84
+ "grad_norm": 2.59375,
85
+ "learning_rate": 9.97362080719462e-06,
86
+ "loss": 1.86279296875,
87
+ "memory/device_reserved (GiB)": 60.86,
88
+ "memory/max_active (GiB)": 54.65,
89
+ "memory/max_allocated (GiB)": 54.65,
90
+ "ppl": 6.4417,
91
+ "step": 6,
92
+ "tokens/total": 3145728,
93
+ "tokens/train_per_sec_per_gpu": 317.14,
94
+ "tokens/trainable": 3142609
95
+ },
96
+ {
97
+ "epoch": 0.11434405308831036,
98
+ "grad_norm": 2.140625,
99
+ "learning_rate": 9.94071935143687e-06,
100
+ "loss": 1.917724609375,
101
+ "memory/device_reserved (GiB)": 60.86,
102
+ "memory/max_active (GiB)": 54.65,
103
+ "memory/max_allocated (GiB)": 54.65,
104
+ "ppl": 6.80546,
105
+ "step": 7,
106
+ "tokens/total": 3670016,
107
+ "tokens/train_per_sec_per_gpu": 316.33,
108
+ "tokens/trainable": 3666263
109
+ },
110
+ {
111
+ "epoch": 0.13067891781521185,
112
+ "grad_norm": 1.8046875,
113
+ "learning_rate": 9.89479250069539e-06,
114
+ "loss": 1.9022216796875,
115
+ "memory/device_reserved (GiB)": 60.86,
116
+ "memory/max_active (GiB)": 54.65,
117
+ "memory/max_allocated (GiB)": 54.65,
118
+ "ppl": 6.70076,
119
+ "step": 8,
120
+ "tokens/total": 4194304,
121
+ "tokens/train_per_sec_per_gpu": 310.55,
122
+ "tokens/trainable": 4189918
123
+ },
124
+ {
125
+ "epoch": 0.14701378254211334,
126
+ "grad_norm": 1.5078125,
127
+ "learning_rate": 9.835974966336504e-06,
128
+ "loss": 1.912841796875,
129
+ "memory/device_reserved (GiB)": 60.86,
130
+ "memory/max_active (GiB)": 54.65,
131
+ "memory/max_allocated (GiB)": 54.65,
132
+ "ppl": 6.77231,
133
+ "step": 9,
134
+ "tokens/total": 4718592,
135
+ "tokens/train_per_sec_per_gpu": 316.98,
136
+ "tokens/trainable": 4713396
137
+ },
138
+ {
139
+ "epoch": 0.1633486472690148,
140
+ "grad_norm": 1.3359375,
141
+ "learning_rate": 9.764439270322612e-06,
142
+ "loss": 1.733642578125,
143
+ "memory/device_reserved (GiB)": 60.86,
144
+ "memory/max_active (GiB)": 54.65,
145
+ "memory/max_allocated (GiB)": 54.65,
146
+ "ppl": 5.66124,
147
+ "step": 10,
148
+ "tokens/total": 5242880,
149
+ "tokens/train_per_sec_per_gpu": 315.3,
150
+ "tokens/trainable": 5236996
151
+ },
152
+ {
153
+ "epoch": 0.17968351199591628,
154
+ "grad_norm": 1.359375,
155
+ "learning_rate": 9.680395239175563e-06,
156
+ "loss": 1.8359375,
157
+ "memory/device_reserved (GiB)": 60.86,
158
+ "memory/max_active (GiB)": 54.65,
159
+ "memory/max_allocated (GiB)": 54.65,
160
+ "ppl": 6.27101,
161
+ "step": 11,
162
+ "tokens/total": 5767168,
163
+ "tokens/train_per_sec_per_gpu": 313.04,
164
+ "tokens/trainable": 5760597
165
+ },
166
+ {
167
+ "epoch": 0.19601837672281777,
168
+ "grad_norm": 1.125,
169
+ "learning_rate": 9.584089388519307e-06,
170
+ "loss": 1.822265625,
171
+ "memory/device_reserved (GiB)": 60.86,
172
+ "memory/max_active (GiB)": 54.65,
173
+ "memory/max_allocated (GiB)": 54.65,
174
+ "ppl": 6.18586,
175
+ "step": 12,
176
+ "tokens/total": 6291456,
177
+ "tokens/train_per_sec_per_gpu": 316.94,
178
+ "tokens/trainable": 6284141
179
+ },
180
+ {
181
+ "epoch": 0.21235324144971923,
182
+ "grad_norm": 1.2578125,
183
+ "learning_rate": 9.475804200007104e-06,
184
+ "loss": 1.8668212890625,
185
+ "memory/device_reserved (GiB)": 60.86,
186
+ "memory/max_active (GiB)": 54.65,
187
+ "memory/max_allocated (GiB)": 54.65,
188
+ "ppl": 6.4677,
189
+ "step": 13,
190
+ "tokens/total": 6815744,
191
+ "tokens/train_per_sec_per_gpu": 312.91,
192
+ "tokens/trainable": 6807833
193
+ },
194
+ {
195
+ "epoch": 0.22868810617662072,
196
+ "grad_norm": 1.6796875,
197
+ "learning_rate": 9.355857292754152e-06,
198
+ "loss": 1.7608642578125,
199
+ "memory/device_reserved (GiB)": 60.86,
200
+ "memory/max_active (GiB)": 54.65,
201
+ "memory/max_allocated (GiB)": 54.65,
202
+ "ppl": 5.81746,
203
+ "step": 14,
204
+ "tokens/total": 7340032,
205
+ "tokens/train_per_sec_per_gpu": 311.53,
206
+ "tokens/trainable": 7331502
207
+ },
208
+ {
209
+ "epoch": 0.2450229709035222,
210
+ "grad_norm": 1.2265625,
211
+ "learning_rate": 9.224600491706009e-06,
212
+ "loss": 1.8038330078125,
213
+ "memory/device_reserved (GiB)": 60.86,
214
+ "memory/max_active (GiB)": 54.65,
215
+ "memory/max_allocated (GiB)": 54.65,
216
+ "ppl": 6.07288,
217
+ "step": 15,
218
+ "tokens/total": 7864320,
219
+ "tokens/train_per_sec_per_gpu": 317.7,
220
+ "tokens/trainable": 7855274
221
+ },
222
+ {
223
+ "epoch": 0.2613578356304237,
224
+ "grad_norm": 0.984375,
225
+ "learning_rate": 9.082418795675397e-06,
226
+ "loss": 1.68658447265625,
227
+ "memory/device_reserved (GiB)": 60.88,
228
+ "memory/max_active (GiB)": 54.65,
229
+ "memory/max_allocated (GiB)": 54.65,
230
+ "ppl": 5.401,
231
+ "step": 16,
232
+ "tokens/total": 8388608,
233
+ "tokens/train_per_sec_per_gpu": 310.73,
234
+ "tokens/trainable": 8379003
235
+ },
236
+ {
237
+ "epoch": 0.2776927003573252,
238
+ "grad_norm": 0.90234375,
239
+ "learning_rate": 8.929729248074364e-06,
240
+ "loss": 1.7650146484375,
241
+ "memory/device_reserved (GiB)": 60.88,
242
+ "memory/max_active (GiB)": 54.65,
243
+ "memory/max_allocated (GiB)": 54.65,
244
+ "ppl": 5.84166,
245
+ "step": 17,
246
+ "tokens/total": 8912896,
247
+ "tokens/train_per_sec_per_gpu": 317.48,
248
+ "tokens/trainable": 8902519
249
+ },
250
+ {
251
+ "epoch": 0.29402756508422667,
252
+ "grad_norm": 1.078125,
253
+ "learning_rate": 8.76697971365409e-06,
254
+ "loss": 1.766845703125,
255
+ "memory/device_reserved (GiB)": 60.88,
256
+ "memory/max_active (GiB)": 54.65,
257
+ "memory/max_allocated (GiB)": 54.65,
258
+ "ppl": 5.85236,
259
+ "step": 18,
260
+ "tokens/total": 9437184,
261
+ "tokens/train_per_sec_per_gpu": 316.29,
262
+ "tokens/trainable": 9426044
263
+ },
264
+ {
265
+ "epoch": 0.3103624298111281,
266
+ "grad_norm": 0.8828125,
267
+ "learning_rate": 8.594647564840407e-06,
268
+ "loss": 1.7130126953125,
269
+ "memory/device_reserved (GiB)": 60.88,
270
+ "memory/max_active (GiB)": 54.65,
271
+ "memory/max_allocated (GiB)": 54.65,
272
+ "ppl": 5.54564,
273
+ "step": 19,
274
+ "tokens/total": 9961472,
275
+ "tokens/train_per_sec_per_gpu": 318.79,
276
+ "tokens/trainable": 9949444
277
+ },
278
+ {
279
+ "epoch": 0.3266972945380296,
280
+ "grad_norm": 0.875,
281
+ "learning_rate": 8.413238281518225e-06,
282
+ "loss": 1.773193359375,
283
+ "memory/device_reserved (GiB)": 60.88,
284
+ "memory/max_active (GiB)": 54.65,
285
+ "memory/max_allocated (GiB)": 54.65,
286
+ "ppl": 5.88963,
287
+ "step": 20,
288
+ "tokens/total": 10485760,
289
+ "tokens/train_per_sec_per_gpu": 315.18,
290
+ "tokens/trainable": 10473247
291
+ },
292
+ {
293
+ "epoch": 0.3430321592649311,
294
+ "grad_norm": 0.96875,
295
+ "learning_rate": 8.223283968371945e-06,
296
+ "loss": 1.70068359375,
297
+ "memory/device_reserved (GiB)": 60.88,
298
+ "memory/max_active (GiB)": 54.65,
299
+ "memory/max_allocated (GiB)": 54.65,
300
+ "ppl": 5.47769,
301
+ "step": 21,
302
+ "tokens/total": 11010048,
303
+ "tokens/train_per_sec_per_gpu": 311.67,
304
+ "tokens/trainable": 10996956
305
+ },
306
+ {
307
+ "epoch": 0.35936702399183257,
308
+ "grad_norm": 0.96875,
309
+ "learning_rate": 8.025341794130722e-06,
310
+ "loss": 1.7213134765625,
311
+ "memory/device_reserved (GiB)": 60.88,
312
+ "memory/max_active (GiB)": 54.65,
313
+ "memory/max_allocated (GiB)": 54.65,
314
+ "ppl": 5.59187,
315
+ "step": 22,
316
+ "tokens/total": 11534336,
317
+ "tokens/train_per_sec_per_gpu": 314.2,
318
+ "tokens/trainable": 11520495
319
+ },
320
+ {
321
+ "epoch": 0.37570188871873406,
322
+ "grad_norm": 0.8515625,
323
+ "learning_rate": 7.8199923572966e-06,
324
+ "loss": 1.7293701171875,
325
+ "memory/device_reserved (GiB)": 60.88,
326
+ "memory/max_active (GiB)": 54.65,
327
+ "memory/max_allocated (GiB)": 54.65,
328
+ "ppl": 5.6371,
329
+ "step": 23,
330
+ "tokens/total": 12058624,
331
+ "tokens/train_per_sec_per_gpu": 313.92,
332
+ "tokens/trainable": 12044083
333
+ },
334
+ {
335
+ "epoch": 0.39203675344563554,
336
+ "grad_norm": 0.84375,
337
+ "learning_rate": 7.607837983149057e-06,
338
+ "loss": 1.727783203125,
339
+ "memory/device_reserved (GiB)": 60.88,
340
+ "memory/max_active (GiB)": 54.65,
341
+ "memory/max_allocated (GiB)": 54.65,
342
+ "ppl": 5.62816,
343
+ "step": 24,
344
+ "tokens/total": 12582912,
345
+ "tokens/train_per_sec_per_gpu": 315.87,
346
+ "tokens/trainable": 12567778
347
+ },
348
+ {
349
+ "epoch": 0.40837161817253703,
350
+ "grad_norm": 0.734375,
351
+ "learning_rate": 7.389500957021192e-06,
352
+ "loss": 1.659912109375,
353
+ "memory/device_reserved (GiB)": 60.88,
354
+ "memory/max_active (GiB)": 54.65,
355
+ "memory/max_allocated (GiB)": 54.65,
356
+ "ppl": 5.25885,
357
+ "step": 25,
358
+ "tokens/total": 13107200,
359
+ "tokens/train_per_sec_per_gpu": 315.52,
360
+ "tokens/trainable": 13091322
361
+ },
362
+ {
363
+ "epoch": 0.42470648289943846,
364
+ "grad_norm": 0.75390625,
365
+ "learning_rate": 7.165621699029615e-06,
366
+ "loss": 1.672119140625,
367
+ "memory/device_reserved (GiB)": 60.88,
368
+ "memory/max_active (GiB)": 54.65,
369
+ "memory/max_allocated (GiB)": 54.65,
370
+ "ppl": 5.32344,
371
+ "step": 26,
372
+ "tokens/total": 13631488,
373
+ "tokens/train_per_sec_per_gpu": 314.88,
374
+ "tokens/trainable": 13615015
375
+ },
376
+ {
377
+ "epoch": 0.44104134762633995,
378
+ "grad_norm": 0.8359375,
379
+ "learning_rate": 6.93685688561191e-06,
380
+ "loss": 1.697509765625,
381
+ "memory/device_reserved (GiB)": 60.88,
382
+ "memory/max_active (GiB)": 54.65,
383
+ "memory/max_allocated (GiB)": 54.65,
384
+ "ppl": 5.46033,
385
+ "step": 27,
386
+ "tokens/total": 14155776,
387
+ "tokens/train_per_sec_per_gpu": 316.84,
388
+ "tokens/trainable": 14138428
389
+ },
390
+ {
391
+ "epoch": 0.45737621235324144,
392
+ "grad_norm": 0.71875,
393
+ "learning_rate": 6.703877523381495e-06,
394
+ "loss": 1.72406005859375,
395
+ "memory/device_reserved (GiB)": 60.88,
396
+ "memory/max_active (GiB)": 54.65,
397
+ "memory/max_allocated (GiB)": 54.65,
398
+ "ppl": 5.60725,
399
+ "step": 28,
400
+ "tokens/total": 14680064,
401
+ "tokens/train_per_sec_per_gpu": 314.54,
402
+ "tokens/trainable": 14661921
403
+ },
404
+ {
405
+ "epoch": 0.47371107708014293,
406
+ "grad_norm": 0.71875,
407
+ "learning_rate": 6.46736698094961e-06,
408
+ "loss": 1.7232666015625,
409
+ "memory/device_reserved (GiB)": 60.88,
410
+ "memory/max_active (GiB)": 54.65,
411
+ "memory/max_allocated (GiB)": 54.65,
412
+ "ppl": 5.6028,
413
+ "step": 29,
414
+ "tokens/total": 15204352,
415
+ "tokens/train_per_sec_per_gpu": 314.53,
416
+ "tokens/trainable": 15185398
417
+ },
418
+ {
419
+ "epoch": 0.4900459418070444,
420
+ "grad_norm": 0.78125,
421
+ "learning_rate": 6.228018984487443e-06,
422
+ "loss": 1.7021484375,
423
+ "memory/device_reserved (GiB)": 60.88,
424
+ "memory/max_active (GiB)": 54.65,
425
+ "memory/max_allocated (GiB)": 54.65,
426
+ "ppl": 5.48572,
427
+ "step": 30,
428
+ "tokens/total": 15728640,
429
+ "tokens/train_per_sec_per_gpu": 315.82,
430
+ "tokens/trainable": 15709057
431
+ },
432
+ {
433
+ "epoch": 0.5063808065339459,
434
+ "grad_norm": 1.625,
435
+ "learning_rate": 5.986535582907739e-06,
436
+ "loss": 1.7518310546875,
437
+ "memory/device_reserved (GiB)": 60.88,
438
+ "memory/max_active (GiB)": 54.65,
439
+ "memory/max_allocated (GiB)": 54.65,
440
+ "ppl": 5.76515,
441
+ "step": 31,
442
+ "tokens/total": 16252928,
443
+ "tokens/train_per_sec_per_gpu": 312.53,
444
+ "tokens/trainable": 16232597
445
+ },
446
+ {
447
+ "epoch": 0.5227156712608474,
448
+ "grad_norm": 0.76171875,
449
+ "learning_rate": 5.74362508863438e-06,
450
+ "loss": 1.7025146484375,
451
+ "memory/device_reserved (GiB)": 60.88,
452
+ "memory/max_active (GiB)": 54.65,
453
+ "memory/max_allocated (GiB)": 54.65,
454
+ "ppl": 5.48773,
455
+ "step": 32,
456
+ "tokens/total": 16777216,
457
+ "tokens/train_per_sec_per_gpu": 310.53,
458
+ "tokens/trainable": 16756110
459
+ },
460
+ {
461
+ "epoch": 0.5390505359877489,
462
+ "grad_norm": 0.66015625,
463
+ "learning_rate": 5.500000000000001e-06,
464
+ "loss": 1.6651611328125,
465
+ "memory/device_reserved (GiB)": 60.88,
466
+ "memory/max_active (GiB)": 54.65,
467
+ "memory/max_allocated (GiB)": 54.65,
468
+ "ppl": 5.28653,
469
+ "step": 33,
470
+ "tokens/total": 17301504,
471
+ "tokens/train_per_sec_per_gpu": 317.49,
472
+ "tokens/trainable": 17279778
473
+ },
474
+ {
475
+ "epoch": 0.5553854007146504,
476
+ "grad_norm": 0.859375,
477
+ "learning_rate": 5.256374911365621e-06,
478
+ "loss": 1.649169921875,
479
+ "memory/device_reserved (GiB)": 60.88,
480
+ "memory/max_active (GiB)": 54.65,
481
+ "memory/max_allocated (GiB)": 54.65,
482
+ "ppl": 5.20266,
483
+ "step": 34,
484
+ "tokens/total": 17825792,
485
+ "tokens/train_per_sec_per_gpu": 317.37,
486
+ "tokens/trainable": 17803390
487
+ },
488
+ {
489
+ "epoch": 0.5717202654415519,
490
+ "grad_norm": 0.7578125,
491
+ "learning_rate": 5.013464417092263e-06,
492
+ "loss": 1.74072265625,
493
+ "memory/device_reserved (GiB)": 60.88,
494
+ "memory/max_active (GiB)": 54.65,
495
+ "memory/max_allocated (GiB)": 54.65,
496
+ "ppl": 5.70146,
497
+ "step": 35,
498
+ "tokens/total": 18350080,
499
+ "tokens/train_per_sec_per_gpu": 317.65,
500
+ "tokens/trainable": 18326908
501
+ },
502
+ {
503
+ "epoch": 0.5880551301684533,
504
+ "grad_norm": 0.6796875,
505
+ "learning_rate": 4.771981015512559e-06,
506
+ "loss": 1.7342529296875,
507
+ "memory/device_reserved (GiB)": 60.88,
508
+ "memory/max_active (GiB)": 54.65,
509
+ "memory/max_allocated (GiB)": 54.65,
510
+ "ppl": 5.66469,
511
+ "step": 36,
512
+ "tokens/total": 18874368,
513
+ "tokens/train_per_sec_per_gpu": 315.72,
514
+ "tokens/trainable": 18850560
515
+ },
516
+ {
517
+ "epoch": 0.6043899948953547,
518
+ "grad_norm": 0.6875,
519
+ "learning_rate": 4.532633019050392e-06,
520
+ "loss": 1.6907958984375,
521
+ "memory/device_reserved (GiB)": 60.88,
522
+ "memory/max_active (GiB)": 54.65,
523
+ "memory/max_allocated (GiB)": 54.65,
524
+ "ppl": 5.4238,
525
+ "step": 37,
526
+ "tokens/total": 19398656,
527
+ "tokens/train_per_sec_per_gpu": 314.03,
528
+ "tokens/trainable": 19374042
529
+ },
530
+ {
531
+ "epoch": 0.6207248596222562,
532
+ "grad_norm": 0.71875,
533
+ "learning_rate": 4.296122476618507e-06,
534
+ "loss": 1.689697265625,
535
+ "memory/device_reserved (GiB)": 60.88,
536
+ "memory/max_active (GiB)": 54.65,
537
+ "memory/max_allocated (GiB)": 54.65,
538
+ "ppl": 5.41784,
539
+ "step": 38,
540
+ "tokens/total": 19922944,
541
+ "tokens/train_per_sec_per_gpu": 315.95,
542
+ "tokens/trainable": 19897582
543
+ },
544
+ {
545
+ "epoch": 0.6370597243491577,
546
+ "grad_norm": 0.765625,
547
+ "learning_rate": 4.0631431143880915e-06,
548
+ "loss": 1.6766357421875,
549
+ "memory/device_reserved (GiB)": 60.88,
550
+ "memory/max_active (GiB)": 54.65,
551
+ "memory/max_allocated (GiB)": 54.65,
552
+ "ppl": 5.34754,
553
+ "step": 39,
554
+ "tokens/total": 20447232,
555
+ "tokens/train_per_sec_per_gpu": 316.88,
556
+ "tokens/trainable": 20421090
557
+ },
558
+ {
559
+ "epoch": 0.6533945890760592,
560
+ "grad_norm": 0.7578125,
561
+ "learning_rate": 3.834378300970385e-06,
562
+ "loss": 1.7320556640625,
563
+ "memory/device_reserved (GiB)": 60.88,
564
+ "memory/max_active (GiB)": 54.65,
565
+ "memory/max_allocated (GiB)": 54.65,
566
+ "ppl": 5.65226,
567
+ "step": 40,
568
+ "tokens/total": 20971520,
569
+ "tokens/train_per_sec_per_gpu": 316.49,
570
+ "tokens/trainable": 20944596
571
+ },
572
+ {
573
+ "epoch": 0.6697294538029607,
574
+ "grad_norm": 0.69140625,
575
+ "learning_rate": 3.6104990429788102e-06,
576
+ "loss": 1.72412109375,
577
+ "memory/device_reserved (GiB)": 60.88,
578
+ "memory/max_active (GiB)": 54.65,
579
+ "memory/max_allocated (GiB)": 54.65,
580
+ "ppl": 5.60759,
581
+ "step": 41,
582
+ "tokens/total": 21495808,
583
+ "tokens/train_per_sec_per_gpu": 316.25,
584
+ "tokens/trainable": 21468328
585
+ },
586
+ {
587
+ "epoch": 0.6860643185298622,
588
+ "grad_norm": 0.8203125,
589
+ "learning_rate": 3.392162016850945e-06,
590
+ "loss": 1.73095703125,
591
+ "memory/device_reserved (GiB)": 60.88,
592
+ "memory/max_active (GiB)": 54.65,
593
+ "memory/max_allocated (GiB)": 54.65,
594
+ "ppl": 5.64605,
595
+ "step": 42,
596
+ "tokens/total": 22020096,
597
+ "tokens/train_per_sec_per_gpu": 310.99,
598
+ "tokens/trainable": 21991928
599
+ },
600
+ {
601
+ "epoch": 0.7023991832567636,
602
+ "grad_norm": 0.73046875,
603
+ "learning_rate": 3.180007642703402e-06,
604
+ "loss": 1.6910400390625,
605
+ "memory/device_reserved (GiB)": 60.88,
606
+ "memory/max_active (GiB)": 54.65,
607
+ "memory/max_allocated (GiB)": 54.65,
608
+ "ppl": 5.42512,
609
+ "step": 43,
610
+ "tokens/total": 22544384,
611
+ "tokens/train_per_sec_per_gpu": 317.51,
612
+ "tokens/trainable": 22515572
613
+ },
614
+ {
615
+ "epoch": 0.7187340479836651,
616
+ "grad_norm": 0.734375,
617
+ "learning_rate": 2.9746582058692803e-06,
618
+ "loss": 1.7021484375,
619
+ "memory/device_reserved (GiB)": 60.88,
620
+ "memory/max_active (GiB)": 54.65,
621
+ "memory/max_allocated (GiB)": 54.65,
622
+ "ppl": 5.48572,
623
+ "step": 44,
624
+ "tokens/total": 23068672,
625
+ "tokens/train_per_sec_per_gpu": 316.96,
626
+ "tokens/trainable": 23039020
627
+ },
628
+ {
629
+ "epoch": 0.7350689127105666,
630
+ "grad_norm": 0.6640625,
631
+ "learning_rate": 2.7767160316280583e-06,
632
+ "loss": 1.6922607421875,
633
+ "memory/device_reserved (GiB)": 60.88,
634
+ "memory/max_active (GiB)": 54.65,
635
+ "memory/max_allocated (GiB)": 54.65,
636
+ "ppl": 5.43175,
637
+ "step": 45,
638
+ "tokens/total": 23592960,
639
+ "tokens/train_per_sec_per_gpu": 317.6,
640
+ "tokens/trainable": 23562636
641
+ },
642
+ {
643
+ "epoch": 0.7514037774374681,
644
+ "grad_norm": 0.67578125,
645
+ "learning_rate": 2.586761718481776e-06,
646
+ "loss": 1.7135009765625,
647
+ "memory/device_reserved (GiB)": 60.88,
648
+ "memory/max_active (GiB)": 54.65,
649
+ "memory/max_allocated (GiB)": 54.65,
650
+ "ppl": 5.54835,
651
+ "step": 46,
652
+ "tokens/total": 24117248,
653
+ "tokens/train_per_sec_per_gpu": 317.44,
654
+ "tokens/trainable": 24086080
655
+ },
656
+ {
657
+ "epoch": 0.7677386421643696,
658
+ "grad_norm": 0.671875,
659
+ "learning_rate": 2.405352435159595e-06,
660
+ "loss": 1.7366943359375,
661
+ "memory/device_reserved (GiB)": 60.88,
662
+ "memory/max_active (GiB)": 54.65,
663
+ "memory/max_allocated (GiB)": 54.65,
664
+ "ppl": 5.67854,
665
+ "step": 47,
666
+ "tokens/total": 24641536,
667
+ "tokens/train_per_sec_per_gpu": 309.64,
668
+ "tokens/trainable": 24609788
669
+ },
670
+ {
671
+ "epoch": 0.7840735068912711,
672
+ "grad_norm": 0.6484375,
673
+ "learning_rate": 2.2330202863459123e-06,
674
+ "loss": 1.6668701171875,
675
+ "memory/device_reserved (GiB)": 60.88,
676
+ "memory/max_active (GiB)": 54.65,
677
+ "memory/max_allocated (GiB)": 54.65,
678
+ "ppl": 5.29557,
679
+ "step": 48,
680
+ "tokens/total": 25165824,
681
+ "tokens/train_per_sec_per_gpu": 317.34,
682
+ "tokens/trainable": 25133492
683
+ },
684
+ {
685
+ "epoch": 0.8004083716181726,
686
+ "grad_norm": 0.76171875,
687
+ "learning_rate": 2.0702707519256365e-06,
688
+ "loss": 1.677978515625,
689
+ "memory/device_reserved (GiB)": 60.88,
690
+ "memory/max_active (GiB)": 54.65,
691
+ "memory/max_allocated (GiB)": 54.65,
692
+ "ppl": 5.35472,
693
+ "step": 49,
694
+ "tokens/total": 25690112,
695
+ "tokens/train_per_sec_per_gpu": 316.39,
696
+ "tokens/trainable": 25657004
697
+ },
698
+ {
699
+ "epoch": 0.8167432363450741,
700
+ "grad_norm": 1.1171875,
701
+ "learning_rate": 1.9175812043246034e-06,
702
+ "loss": 1.673583984375,
703
+ "memory/device_reserved (GiB)": 60.88,
704
+ "memory/max_active (GiB)": 54.65,
705
+ "memory/max_allocated (GiB)": 54.65,
706
+ "ppl": 5.33124,
707
+ "step": 50,
708
+ "tokens/total": 26214400,
709
+ "tokens/train_per_sec_per_gpu": 317.49,
710
+ "tokens/trainable": 26180508
711
+ },
712
+ {
713
+ "epoch": 0.8330781010719756,
714
+ "grad_norm": 0.703125,
715
+ "learning_rate": 1.7753995082939932e-06,
716
+ "loss": 1.7542724609375,
717
+ "memory/device_reserved (GiB)": 60.88,
718
+ "memory/max_active (GiB)": 54.65,
719
+ "memory/max_allocated (GiB)": 54.65,
720
+ "ppl": 5.77924,
721
+ "step": 51,
722
+ "tokens/total": 26738688,
723
+ "tokens/train_per_sec_per_gpu": 316.75,
724
+ "tokens/trainable": 26704180
725
+ },
726
+ {
727
+ "epoch": 0.8494129657988769,
728
+ "grad_norm": 0.62890625,
729
+ "learning_rate": 1.6441427072458493e-06,
730
+ "loss": 1.675048828125,
731
+ "memory/device_reserved (GiB)": 60.88,
732
+ "memory/max_active (GiB)": 54.65,
733
+ "memory/max_allocated (GiB)": 54.65,
734
+ "ppl": 5.33906,
735
+ "step": 52,
736
+ "tokens/total": 27262976,
737
+ "tokens/train_per_sec_per_gpu": 317.09,
738
+ "tokens/trainable": 27227688
739
+ },
740
+ {
741
+ "epoch": 0.8657478305257784,
742
+ "grad_norm": 0.62109375,
743
+ "learning_rate": 1.5241957999928974e-06,
744
+ "loss": 1.69140625,
745
+ "memory/device_reserved (GiB)": 60.88,
746
+ "memory/max_active (GiB)": 54.65,
747
+ "memory/max_allocated (GiB)": 54.65,
748
+ "ppl": 5.42711,
749
+ "step": 53,
750
+ "tokens/total": 27787264,
751
+ "tokens/train_per_sec_per_gpu": 318.12,
752
+ "tokens/trainable": 27751062
753
+ },
754
+ {
755
+ "epoch": 0.8820826952526799,
756
+ "grad_norm": 0.69140625,
757
+ "learning_rate": 1.4159106114806943e-06,
758
+ "loss": 1.634765625,
759
+ "memory/device_reserved (GiB)": 60.88,
760
+ "memory/max_active (GiB)": 54.65,
761
+ "memory/max_allocated (GiB)": 54.65,
762
+ "ppl": 5.12826,
763
+ "step": 54,
764
+ "tokens/total": 28311552,
765
+ "tokens/train_per_sec_per_gpu": 315.78,
766
+ "tokens/trainable": 28274560
767
+ },
768
+ {
769
+ "epoch": 0.8984175599795814,
770
+ "grad_norm": 0.78125,
771
+ "learning_rate": 1.319604760824439e-06,
772
+ "loss": 1.6998291015625,
773
+ "memory/device_reserved (GiB)": 60.88,
774
+ "memory/max_active (GiB)": 54.65,
775
+ "memory/max_allocated (GiB)": 54.65,
776
+ "ppl": 5.47301,
777
+ "step": 55,
778
+ "tokens/total": 28835840,
779
+ "tokens/train_per_sec_per_gpu": 317.77,
780
+ "tokens/trainable": 28798056
781
+ },
782
+ {
783
+ "epoch": 0.9147524247064829,
784
+ "grad_norm": 0.91015625,
785
+ "learning_rate": 1.2355607296773896e-06,
786
+ "loss": 1.701904296875,
787
+ "memory/device_reserved (GiB)": 60.88,
788
+ "memory/max_active (GiB)": 54.65,
789
+ "memory/max_allocated (GiB)": 54.65,
790
+ "ppl": 5.48438,
791
+ "step": 56,
792
+ "tokens/total": 29360128,
793
+ "tokens/train_per_sec_per_gpu": 314.84,
794
+ "tokens/trainable": 29321514
795
+ },
796
+ {
797
+ "epoch": 0.9310872894333844,
798
+ "grad_norm": 0.7109375,
799
+ "learning_rate": 1.164025033663497e-06,
800
+ "loss": 1.6883544921875,
801
+ "memory/device_reserved (GiB)": 60.88,
802
+ "memory/max_active (GiB)": 54.65,
803
+ "memory/max_allocated (GiB)": 54.65,
804
+ "ppl": 5.41057,
805
+ "step": 57,
806
+ "tokens/total": 29884416,
807
+ "tokens/train_per_sec_per_gpu": 316.78,
808
+ "tokens/trainable": 29845030
809
+ },
810
+ {
811
+ "epoch": 0.9474221541602859,
812
+ "grad_norm": 0.5859375,
813
+ "learning_rate": 1.1052074993046102e-06,
814
+ "loss": 1.707275390625,
815
+ "memory/device_reserved (GiB)": 60.88,
816
+ "memory/max_active (GiB)": 54.65,
817
+ "memory/max_allocated (GiB)": 54.65,
818
+ "ppl": 5.51392,
819
+ "step": 58,
820
+ "tokens/total": 30408704,
821
+ "tokens/train_per_sec_per_gpu": 318.2,
822
+ "tokens/trainable": 30368456
823
+ },
824
+ {
825
+ "epoch": 0.9637570188871873,
826
+ "grad_norm": 0.890625,
827
+ "learning_rate": 1.0592806485631326e-06,
828
+ "loss": 1.759765625,
829
+ "memory/device_reserved (GiB)": 60.88,
830
+ "memory/max_active (GiB)": 54.65,
831
+ "memory/max_allocated (GiB)": 54.65,
832
+ "ppl": 5.81108,
833
+ "step": 59,
834
+ "tokens/total": 30932992,
835
+ "tokens/train_per_sec_per_gpu": 316.06,
836
+ "tokens/trainable": 30891900
837
+ },
838
+ {
839
+ "epoch": 0.9800918836140888,
840
+ "grad_norm": 1.484375,
841
+ "learning_rate": 1.026379192805382e-06,
842
+ "loss": 1.66357421875,
843
+ "memory/device_reserved (GiB)": 60.88,
844
+ "memory/max_active (GiB)": 54.65,
845
+ "memory/max_allocated (GiB)": 54.65,
846
+ "ppl": 5.27814,
847
+ "step": 60,
848
+ "tokens/total": 31457280,
849
+ "tokens/train_per_sec_per_gpu": 317.36,
850
+ "tokens/trainable": 31415516
851
+ },
852
+ {
853
+ "epoch": 0.9964267483409903,
854
+ "grad_norm": 0.94921875,
855
+ "learning_rate": 1.006599637669943e-06,
856
+ "loss": 1.688720703125,
857
+ "memory/device_reserved (GiB)": 60.88,
858
+ "memory/max_active (GiB)": 54.65,
859
+ "memory/max_allocated (GiB)": 54.65,
860
+ "ppl": 5.41255,
861
+ "step": 61,
862
+ "tokens/total": 31981568,
863
+ "tokens/train_per_sec_per_gpu": 315.86,
864
+ "tokens/trainable": 31938924
865
+ }
866
+ ],
867
+ "logging_steps": 1,
868
+ "max_steps": 61,
869
+ "num_input_tokens_seen": 0,
870
+ "num_train_epochs": 1,
871
+ "save_steps": 500,
872
+ "stateful_callbacks": {
873
+ "TrainerControl": {
874
+ "args": {
875
+ "should_epoch_stop": false,
876
+ "should_evaluate": false,
877
+ "should_log": false,
878
+ "should_save": true,
879
+ "should_training_stop": true
880
+ },
881
+ "attributes": {}
882
+ }
883
+ },
884
+ "total_flos": 6.954280102867763e+17,
885
+ "train_batch_size": 1,
886
+ "trial_name": null,
887
+ "trial_params": null
888
+ }