update model with approx 1.6 epochs training
Browse files- config.json +1 -1
- latest +1 -0
- merges.txt +1 -1
- pytorch_model.bin +2 -2
- tokenizer_config.json +1 -1
- trainer_state.json +621 -1155
- training_args.bin +2 -2
- vocab.json +0 -0
- zero_to_fp32.py +484 -0
config.json
CHANGED
@@ -1,5 +1,5 @@
|
|
1 |
{
|
2 |
-
"_name_or_path": "pszemraj/opt-peter-1.3B
|
3 |
"activation_dropout": 0.0,
|
4 |
"activation_function": "relu",
|
5 |
"architectures": [
|
|
|
1 |
{
|
2 |
+
"_name_or_path": "pszemraj/opt-peter-1.3B",
|
3 |
"activation_dropout": 0.0,
|
4 |
"activation_function": "relu",
|
5 |
"architectures": [
|
latest
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
global_step1016
|
merges.txt
CHANGED
@@ -1,4 +1,4 @@
|
|
1 |
-
#version: 0.2
|
2 |
Ġ t
|
3 |
Ġ a
|
4 |
h e
|
|
|
1 |
+
#version: 0.2
|
2 |
Ġ t
|
3 |
Ġ a
|
4 |
h e
|
pytorch_model.bin
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
-
size
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:4cad2a7fef5856274323ae7e812eec18858a768fd7d3f83bcfc3c2e9d0d3d1de
|
3 |
+
size 5263006227
|
tokenizer_config.json
CHANGED
@@ -1 +1 @@
|
|
1 |
-
{"unk_token": {"content": "</s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true, "__type": "AddedToken"}, "bos_token": {"content": "</s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true, "__type": "AddedToken"}, "eos_token": {"content": "</s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true, "__type": "AddedToken"}, "
|
|
|
1 |
+
{"errors": "replace", "unk_token": {"content": "</s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true, "__type": "AddedToken"}, "bos_token": {"content": "</s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true, "__type": "AddedToken"}, "eos_token": {"content": "</s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true, "__type": "AddedToken"}, "pad_token": {"content": "<pad>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true, "__type": "AddedToken"}, "add_prefix_space": false, "add_bos_token": true, "special_tokens_map_file": null, "name_or_path": "pszemraj/opt-peter-1.3B", "model_max_length": 512, "tokenizer_class": "GPT2Tokenizer"}
|
trainer_state.json
CHANGED
@@ -1,1776 +1,1242 @@
|
|
1 |
{
|
2 |
"best_metric": null,
|
3 |
"best_model_checkpoint": null,
|
4 |
-
"epoch": 1.
|
5 |
-
"global_step":
|
6 |
"is_hyper_param_search": false,
|
7 |
"is_local_process_zero": true,
|
8 |
"is_world_process_zero": true,
|
9 |
"log_history": [
|
10 |
{
|
11 |
-
"epoch": 0.
|
12 |
-
"learning_rate":
|
13 |
-
"loss":
|
14 |
"step": 5
|
15 |
},
|
16 |
{
|
17 |
-
"epoch": 0.
|
18 |
-
"learning_rate":
|
19 |
-
"loss":
|
20 |
"step": 10
|
21 |
},
|
22 |
{
|
23 |
-
"epoch": 0.
|
24 |
-
"learning_rate":
|
25 |
-
"loss": 1.
|
26 |
"step": 15
|
27 |
},
|
28 |
{
|
29 |
-
"epoch": 0.
|
30 |
-
"learning_rate":
|
31 |
-
"loss": 1.
|
32 |
"step": 20
|
33 |
},
|
34 |
{
|
35 |
-
"epoch": 0.
|
36 |
-
"learning_rate":
|
37 |
-
"loss": 1.
|
38 |
"step": 25
|
39 |
},
|
40 |
{
|
41 |
-
"epoch": 0.
|
42 |
-
"learning_rate":
|
43 |
-
"loss": 1.
|
44 |
"step": 30
|
45 |
},
|
46 |
{
|
47 |
-
"epoch": 0.
|
48 |
-
"learning_rate":
|
49 |
-
"loss": 1.
|
50 |
"step": 35
|
51 |
},
|
52 |
{
|
53 |
-
"epoch": 0.
|
54 |
-
"learning_rate":
|
55 |
-
"loss": 1.
|
56 |
"step": 40
|
57 |
},
|
58 |
{
|
59 |
-
"epoch": 0.
|
60 |
-
"learning_rate": 2.
|
61 |
-
"loss": 1.
|
62 |
"step": 45
|
63 |
},
|
64 |
{
|
65 |
-
"epoch": 0.
|
66 |
-
"learning_rate": 2.
|
67 |
-
"loss": 1.
|
68 |
"step": 50
|
69 |
},
|
70 |
{
|
71 |
-
"epoch": 0.
|
72 |
-
"learning_rate": 2.
|
73 |
-
"loss": 1.
|
74 |
"step": 55
|
75 |
},
|
76 |
{
|
77 |
-
"epoch": 0.
|
78 |
-
"learning_rate": 2.
|
79 |
-
"loss": 1.
|
80 |
"step": 60
|
81 |
},
|
82 |
{
|
83 |
-
"epoch": 0.
|
84 |
-
"learning_rate": 2.
|
85 |
-
"loss": 1.
|
86 |
"step": 65
|
87 |
},
|
88 |
{
|
89 |
-
"epoch": 0.
|
90 |
-
"learning_rate": 2.
|
91 |
-
"loss": 1.
|
92 |
"step": 70
|
93 |
},
|
94 |
{
|
95 |
-
"epoch": 0.
|
96 |
-
"learning_rate": 2.
|
97 |
-
"loss": 1.
|
98 |
"step": 75
|
99 |
},
|
100 |
{
|
101 |
-
"epoch": 0.
|
102 |
-
"learning_rate": 2.
|
103 |
-
"loss": 1.
|
104 |
"step": 80
|
105 |
},
|
106 |
{
|
107 |
-
"epoch": 0.
|
108 |
-
"learning_rate": 2.
|
109 |
-
"loss": 1.
|
110 |
"step": 85
|
111 |
},
|
112 |
{
|
113 |
-
"epoch": 0.
|
114 |
-
"learning_rate": 2.
|
115 |
-
"loss": 1.
|
116 |
"step": 90
|
117 |
},
|
118 |
{
|
119 |
-
"epoch": 0.
|
120 |
-
"learning_rate": 2.
|
121 |
-
"loss": 1.
|
122 |
"step": 95
|
123 |
},
|
124 |
{
|
125 |
-
"epoch": 0.
|
126 |
-
"learning_rate": 2.
|
127 |
-
"loss": 1.
|
128 |
"step": 100
|
129 |
},
|
130 |
{
|
131 |
-
"epoch": 0.
|
132 |
-
"learning_rate": 2.
|
133 |
-
"loss": 1.
|
134 |
"step": 105
|
135 |
},
|
136 |
{
|
137 |
-
"epoch": 0.
|
138 |
-
"learning_rate": 2.
|
139 |
-
"loss": 1.
|
140 |
"step": 110
|
141 |
},
|
142 |
{
|
143 |
-
"epoch": 0.
|
144 |
-
"learning_rate": 2.
|
145 |
-
"loss": 1.
|
146 |
"step": 115
|
147 |
},
|
148 |
{
|
149 |
-
"epoch": 0.
|
150 |
-
"learning_rate": 2.
|
151 |
-
"loss": 1.
|
152 |
"step": 120
|
153 |
},
|
154 |
{
|
155 |
-
"epoch": 0.
|
156 |
-
"learning_rate": 2.
|
157 |
-
"loss": 1.
|
158 |
"step": 125
|
159 |
},
|
160 |
{
|
161 |
-
"epoch": 0.
|
162 |
-
"learning_rate": 2.
|
163 |
-
"loss": 1.
|
164 |
"step": 130
|
165 |
},
|
166 |
{
|
167 |
-
"epoch": 0.
|
168 |
-
"learning_rate": 2.
|
169 |
-
"loss": 1.
|
170 |
"step": 135
|
171 |
},
|
172 |
{
|
173 |
-
"epoch": 0.
|
174 |
-
"learning_rate": 2.
|
175 |
-
"loss": 1.
|
176 |
"step": 140
|
177 |
},
|
178 |
{
|
179 |
-
"epoch": 0.
|
180 |
-
"learning_rate": 2.
|
181 |
-
"loss": 1.
|
182 |
"step": 145
|
183 |
},
|
184 |
{
|
185 |
-
"epoch": 0.
|
186 |
-
"learning_rate": 2.
|
187 |
-
"loss": 1.
|
188 |
"step": 150
|
189 |
},
|
190 |
{
|
191 |
-
"epoch": 0.
|
192 |
-
"learning_rate": 2.
|
193 |
-
"loss": 1.
|
194 |
"step": 155
|
195 |
},
|
196 |
{
|
197 |
-
"epoch": 0.
|
198 |
-
"learning_rate": 2.
|
199 |
-
"loss": 1.
|
200 |
"step": 160
|
201 |
},
|
202 |
{
|
203 |
-
"epoch": 0.
|
204 |
-
"learning_rate": 2.
|
205 |
-
"loss": 1.
|
206 |
"step": 165
|
207 |
},
|
208 |
{
|
209 |
-
"epoch": 0.
|
210 |
-
"learning_rate": 2.
|
211 |
-
"loss": 1.
|
212 |
"step": 170
|
213 |
},
|
214 |
{
|
215 |
-
"epoch": 0.
|
216 |
-
"learning_rate": 2.
|
217 |
-
"loss": 1.
|
218 |
"step": 175
|
219 |
},
|
220 |
{
|
221 |
-
"epoch": 0.
|
222 |
-
"learning_rate": 2.
|
223 |
-
"loss": 1.
|
224 |
"step": 180
|
225 |
},
|
226 |
{
|
227 |
-
"epoch": 0.
|
228 |
-
"learning_rate": 2.
|
229 |
-
"loss": 1.
|
230 |
"step": 185
|
231 |
},
|
232 |
{
|
233 |
-
"epoch": 0.
|
234 |
-
"learning_rate": 2.
|
235 |
-
"loss": 1.
|
236 |
"step": 190
|
237 |
},
|
238 |
{
|
239 |
-
"epoch": 0.
|
240 |
-
"learning_rate": 2.
|
241 |
-
"loss": 1.
|
242 |
"step": 195
|
243 |
},
|
244 |
{
|
245 |
-
"epoch": 0.
|
246 |
-
"learning_rate": 2.
|
247 |
-
"loss": 1.
|
248 |
"step": 200
|
249 |
},
|
250 |
{
|
251 |
-
"epoch": 0.
|
252 |
-
"learning_rate": 2.
|
253 |
-
"loss": 1.
|
254 |
"step": 205
|
255 |
},
|
256 |
{
|
257 |
-
"epoch": 0.
|
258 |
-
"learning_rate": 2.
|
259 |
-
"loss": 1.
|
260 |
"step": 210
|
261 |
},
|
262 |
{
|
263 |
-
"epoch": 0.
|
264 |
-
"learning_rate": 2.
|
265 |
-
"loss": 1.
|
266 |
"step": 215
|
267 |
},
|
268 |
{
|
269 |
-
"epoch": 0.
|
270 |
-
"learning_rate": 2.
|
271 |
-
"loss": 1.
|
272 |
"step": 220
|
273 |
},
|
274 |
{
|
275 |
-
"epoch": 0.
|
276 |
-
"learning_rate": 2.
|
277 |
-
"loss": 1.
|
278 |
"step": 225
|
279 |
},
|
280 |
{
|
281 |
-
"epoch": 0.
|
282 |
-
"learning_rate": 2.
|
283 |
-
"loss": 1.
|
284 |
"step": 230
|
285 |
},
|
286 |
{
|
287 |
-
"epoch": 0.
|
288 |
-
"learning_rate": 2.
|
289 |
-
"loss": 1.
|
290 |
"step": 235
|
291 |
},
|
292 |
{
|
293 |
-
"epoch": 0.
|
294 |
-
"learning_rate": 2.
|
295 |
-
"loss": 1.
|
296 |
"step": 240
|
297 |
},
|
298 |
{
|
299 |
-
"epoch": 0.
|
300 |
-
"learning_rate": 2.
|
301 |
-
"loss": 1.
|
302 |
"step": 245
|
303 |
},
|
304 |
{
|
305 |
-
"epoch": 0.
|
306 |
-
"learning_rate": 2.
|
307 |
-
"loss": 1.
|
308 |
"step": 250
|
309 |
},
|
310 |
{
|
311 |
-
"epoch": 0.
|
312 |
-
"learning_rate": 2.
|
313 |
-
"loss": 1.
|
314 |
"step": 255
|
315 |
},
|
316 |
{
|
317 |
-
"epoch": 0.
|
318 |
-
"learning_rate": 2.
|
319 |
-
"loss": 1.
|
320 |
"step": 260
|
321 |
},
|
322 |
{
|
323 |
-
"epoch": 0.
|
324 |
-
"learning_rate": 2.
|
325 |
-
"loss": 1.
|
326 |
"step": 265
|
327 |
},
|
328 |
{
|
329 |
-
"epoch": 0.
|
330 |
-
"learning_rate": 2.
|
331 |
-
"loss": 1.
|
332 |
"step": 270
|
333 |
},
|
334 |
{
|
335 |
-
"epoch": 0.
|
336 |
-
"learning_rate": 2.
|
337 |
-
"loss": 1.
|
338 |
"step": 275
|
339 |
},
|
340 |
{
|
341 |
-
"epoch": 0.
|
342 |
-
"learning_rate": 2.
|
343 |
-
"loss": 1.
|
344 |
"step": 280
|
345 |
},
|
346 |
{
|
347 |
-
"epoch": 0.
|
348 |
-
"learning_rate": 2.
|
349 |
-
"loss": 1.
|
350 |
"step": 285
|
351 |
},
|
352 |
{
|
353 |
-
"epoch": 0.
|
354 |
-
"learning_rate": 2.
|
355 |
-
"loss": 1.
|
356 |
"step": 290
|
357 |
},
|
358 |
{
|
359 |
-
"epoch": 0.
|
360 |
-
"learning_rate": 2.
|
361 |
-
"loss": 1.
|
362 |
"step": 295
|
363 |
},
|
364 |
{
|
365 |
-
"epoch": 0.
|
366 |
-
"learning_rate": 2.
|
367 |
-
"loss": 1.
|
368 |
"step": 300
|
369 |
},
|
370 |
{
|
371 |
-
"epoch": 0.
|
372 |
-
"learning_rate": 2.
|
373 |
-
"loss": 1.
|
374 |
"step": 305
|
375 |
},
|
376 |
{
|
377 |
-
"epoch": 0.
|
378 |
-
"learning_rate": 2.
|
379 |
-
"loss": 1.
|
380 |
"step": 310
|
381 |
},
|
382 |
{
|
383 |
-
"epoch": 0.
|
384 |
-
"learning_rate": 2.
|
385 |
-
"loss": 1.
|
386 |
"step": 315
|
387 |
},
|
388 |
{
|
389 |
-
"epoch": 0.
|
390 |
-
"learning_rate": 2.
|
391 |
-
"loss": 1.
|
392 |
"step": 320
|
393 |
},
|
394 |
{
|
395 |
-
"epoch": 0.
|
396 |
-
"learning_rate": 2.
|
397 |
-
"loss": 1.
|
398 |
"step": 325
|
399 |
},
|
400 |
{
|
401 |
-
"epoch": 0.
|
402 |
-
"learning_rate": 2.
|
403 |
-
"loss": 1.
|
404 |
"step": 330
|
405 |
},
|
406 |
{
|
407 |
-
"epoch": 0.
|
408 |
-
"learning_rate": 2.
|
409 |
-
"loss": 1.
|
410 |
"step": 335
|
411 |
},
|
412 |
{
|
413 |
-
"epoch": 0.
|
414 |
-
"learning_rate": 2.
|
415 |
-
"loss": 1.
|
416 |
"step": 340
|
417 |
},
|
418 |
{
|
419 |
-
"epoch": 0.
|
420 |
-
"learning_rate": 2.
|
421 |
-
"loss": 1.
|
422 |
"step": 345
|
423 |
},
|
424 |
{
|
425 |
-
"epoch": 0.
|
426 |
-
"learning_rate": 2.
|
427 |
-
"loss": 1.
|
428 |
"step": 350
|
429 |
},
|
430 |
{
|
431 |
-
"epoch": 0.
|
432 |
-
"learning_rate": 2.
|
433 |
-
"loss": 1.
|
434 |
"step": 355
|
435 |
},
|
436 |
{
|
437 |
-
"epoch": 0.
|
438 |
-
"learning_rate": 2.
|
439 |
-
"loss": 1.
|
440 |
"step": 360
|
441 |
},
|
442 |
{
|
443 |
-
"epoch": 0.
|
444 |
-
"learning_rate": 2.
|
445 |
-
"loss": 1.
|
446 |
"step": 365
|
447 |
},
|
448 |
{
|
449 |
-
"epoch": 0.
|
450 |
-
"learning_rate": 2.
|
451 |
-
"loss": 1.
|
452 |
"step": 370
|
453 |
},
|
454 |
{
|
455 |
-
"epoch": 0.
|
456 |
-
"learning_rate": 2.
|
457 |
-
"loss": 1.
|
458 |
"step": 375
|
459 |
},
|
460 |
{
|
461 |
-
"epoch": 0.
|
462 |
-
"learning_rate": 2.
|
463 |
-
"loss": 1.
|
464 |
"step": 380
|
465 |
},
|
466 |
{
|
467 |
-
"epoch": 0.
|
468 |
-
"learning_rate": 2.
|
469 |
-
"loss": 1.
|
470 |
"step": 385
|
471 |
},
|
472 |
{
|
473 |
-
"epoch": 0.
|
474 |
-
"learning_rate": 2.
|
475 |
-
"loss": 1.
|
476 |
"step": 390
|
477 |
},
|
478 |
{
|
479 |
-
"epoch": 0.
|
480 |
-
"learning_rate": 2.
|
481 |
-
"loss": 1.
|
482 |
"step": 395
|
483 |
},
|
484 |
{
|
485 |
-
"epoch": 0.
|
486 |
-
"learning_rate": 2.
|
487 |
-
"loss": 1.
|
488 |
"step": 400
|
489 |
},
|
490 |
{
|
491 |
-
"epoch": 0.
|
492 |
-
"learning_rate": 2.
|
493 |
-
"loss": 1.
|
494 |
"step": 405
|
495 |
},
|
496 |
{
|
497 |
-
"epoch": 0.
|
498 |
-
"learning_rate": 2.
|
499 |
-
"loss": 1.
|
500 |
"step": 410
|
501 |
},
|
502 |
{
|
503 |
-
"epoch": 0.
|
504 |
-
"learning_rate": 2.
|
505 |
-
"loss": 1.
|
506 |
"step": 415
|
507 |
},
|
508 |
{
|
509 |
-
"epoch": 0.
|
510 |
-
"learning_rate": 2.
|
511 |
-
"loss": 1.
|
512 |
"step": 420
|
513 |
},
|
514 |
{
|
515 |
-
"epoch": 0.
|
516 |
-
"learning_rate": 2.
|
517 |
-
"loss": 1.
|
518 |
"step": 425
|
519 |
},
|
520 |
{
|
521 |
-
"epoch": 0.
|
522 |
-
"learning_rate": 2.
|
523 |
-
"loss": 1.
|
524 |
"step": 430
|
525 |
},
|
526 |
{
|
527 |
-
"epoch": 0.
|
528 |
-
"learning_rate": 2.
|
529 |
-
"loss": 1.
|
530 |
"step": 435
|
531 |
},
|
532 |
{
|
533 |
-
"epoch": 0.
|
534 |
-
"learning_rate": 2.
|
535 |
-
"loss": 1.
|
536 |
"step": 440
|
537 |
},
|
538 |
{
|
539 |
-
"epoch": 0.
|
540 |
-
"learning_rate": 2.
|
541 |
-
"loss": 1.
|
542 |
"step": 445
|
543 |
},
|
544 |
{
|
545 |
-
"epoch": 0.
|
546 |
-
"learning_rate": 2.
|
547 |
-
"loss": 1.
|
548 |
"step": 450
|
549 |
},
|
550 |
{
|
551 |
-
"epoch": 0.
|
552 |
-
"learning_rate": 2.
|
553 |
-
"loss": 1.
|
554 |
"step": 455
|
555 |
},
|
556 |
{
|
557 |
-
"epoch": 0.
|
558 |
-
"learning_rate": 2.
|
559 |
-
"loss": 1.
|
560 |
"step": 460
|
561 |
},
|
562 |
{
|
563 |
-
"epoch": 0.
|
564 |
-
"learning_rate": 2.
|
565 |
-
"loss": 1.
|
566 |
"step": 465
|
567 |
},
|
568 |
{
|
569 |
-
"epoch": 0.
|
570 |
-
"learning_rate": 2.
|
571 |
-
"loss": 1.
|
572 |
"step": 470
|
573 |
},
|
574 |
{
|
575 |
-
"epoch": 0.
|
576 |
-
"learning_rate": 2.
|
577 |
-
"loss": 1.
|
578 |
"step": 475
|
579 |
},
|
580 |
{
|
581 |
-
"epoch": 0.
|
582 |
-
"learning_rate": 2.
|
583 |
-
"loss": 1.
|
584 |
"step": 480
|
585 |
},
|
586 |
{
|
587 |
-
"epoch": 0.
|
588 |
-
"learning_rate": 2.
|
589 |
-
"loss": 1.
|
590 |
"step": 485
|
591 |
},
|
592 |
{
|
593 |
-
"epoch": 0.
|
594 |
-
"learning_rate": 2.
|
595 |
-
"loss": 1.
|
596 |
"step": 490
|
597 |
},
|
598 |
{
|
599 |
-
"epoch": 0.
|
600 |
-
"learning_rate": 2.
|
601 |
-
"loss": 1.
|
602 |
"step": 495
|
603 |
},
|
604 |
{
|
605 |
-
"epoch": 0.
|
606 |
-
"learning_rate": 2.
|
607 |
-
"loss": 1.
|
608 |
"step": 500
|
609 |
},
|
610 |
{
|
611 |
-
"epoch": 0.
|
612 |
-
"learning_rate": 2.
|
613 |
-
"loss": 1.
|
614 |
"step": 505
|
615 |
},
|
616 |
{
|
617 |
-
"epoch": 0.
|
618 |
-
"learning_rate": 2.
|
619 |
-
"loss": 1.
|
620 |
"step": 510
|
621 |
},
|
622 |
{
|
623 |
-
"epoch": 0.
|
624 |
-
"learning_rate": 2.
|
625 |
-
"loss": 1.
|
626 |
"step": 515
|
627 |
},
|
628 |
{
|
629 |
-
"epoch": 0.
|
630 |
-
"learning_rate": 2.
|
631 |
-
"loss": 1.
|
632 |
"step": 520
|
633 |
},
|
634 |
{
|
635 |
-
"epoch": 0.
|
636 |
-
"learning_rate":
|
637 |
-
"loss": 1.
|
638 |
"step": 525
|
639 |
},
|
640 |
{
|
641 |
-
"epoch": 0.
|
642 |
-
"learning_rate":
|
643 |
-
"loss": 1.
|
644 |
"step": 530
|
645 |
},
|
646 |
{
|
647 |
-
"epoch": 0.
|
648 |
-
"learning_rate":
|
649 |
-
"loss": 1.
|
650 |
"step": 535
|
651 |
},
|
652 |
{
|
653 |
-
"epoch": 0.
|
654 |
-
"learning_rate":
|
655 |
-
"loss": 1.
|
656 |
"step": 540
|
657 |
},
|
658 |
{
|
659 |
-
"epoch": 0.
|
660 |
-
"learning_rate":
|
661 |
-
"loss": 1.
|
662 |
"step": 545
|
663 |
},
|
664 |
{
|
665 |
-
"epoch": 0.
|
666 |
-
"learning_rate":
|
667 |
-
"loss": 1.
|
668 |
"step": 550
|
669 |
},
|
670 |
{
|
671 |
-
"epoch": 0.
|
672 |
-
"learning_rate":
|
673 |
-
"loss": 1.
|
674 |
"step": 555
|
675 |
},
|
676 |
{
|
677 |
-
"epoch": 0.
|
678 |
-
"learning_rate":
|
679 |
-
"loss": 1.
|
680 |
"step": 560
|
681 |
},
|
682 |
{
|
683 |
-
"epoch": 0.
|
684 |
-
"learning_rate":
|
685 |
-
"loss": 1.
|
686 |
"step": 565
|
687 |
},
|
688 |
{
|
689 |
-
"epoch": 0.
|
690 |
-
"learning_rate":
|
691 |
-
"loss": 1.
|
692 |
"step": 570
|
693 |
},
|
694 |
{
|
695 |
-
"epoch": 0.
|
696 |
-
"learning_rate":
|
697 |
-
"loss": 1.
|
698 |
"step": 575
|
699 |
},
|
700 |
{
|
701 |
-
"epoch": 0.
|
702 |
-
"learning_rate":
|
703 |
-
"loss": 1.
|
704 |
"step": 580
|
705 |
},
|
706 |
{
|
707 |
-
"epoch": 0.
|
708 |
-
"learning_rate":
|
709 |
-
"loss": 1.
|
710 |
"step": 585
|
711 |
},
|
712 |
{
|
713 |
-
"epoch": 0.
|
714 |
-
"learning_rate":
|
715 |
-
"loss": 1.
|
716 |
"step": 590
|
717 |
},
|
718 |
{
|
719 |
-
"epoch": 0.
|
720 |
-
"learning_rate":
|
721 |
-
"loss": 1.
|
722 |
"step": 595
|
723 |
},
|
724 |
{
|
725 |
-
"epoch": 0.
|
726 |
-
"learning_rate":
|
727 |
-
"loss": 1.
|
728 |
"step": 600
|
729 |
},
|
730 |
{
|
731 |
-
"epoch": 0.
|
732 |
-
"learning_rate":
|
733 |
-
"loss": 1.
|
734 |
"step": 605
|
735 |
},
|
736 |
{
|
737 |
-
"epoch": 0.
|
738 |
-
"learning_rate":
|
739 |
-
"loss": 1.
|
740 |
"step": 610
|
741 |
},
|
742 |
{
|
743 |
-
"epoch": 0
|
744 |
-
"learning_rate":
|
745 |
-
"loss": 1.
|
746 |
"step": 615
|
747 |
},
|
748 |
{
|
749 |
-
"epoch": 0
|
750 |
-
"
|
751 |
-
"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
752 |
"step": 620
|
753 |
},
|
754 |
{
|
755 |
-
"epoch":
|
756 |
-
"learning_rate":
|
757 |
-
"loss": 1.
|
758 |
"step": 625
|
759 |
},
|
760 |
{
|
761 |
-
"epoch":
|
762 |
-
"learning_rate":
|
763 |
-
"loss": 1.
|
764 |
"step": 630
|
765 |
},
|
766 |
{
|
767 |
-
"epoch":
|
768 |
-
"learning_rate":
|
769 |
-
"loss": 1.
|
770 |
"step": 635
|
771 |
},
|
772 |
{
|
773 |
-
"epoch":
|
774 |
-
"learning_rate":
|
775 |
-
"loss": 1.
|
776 |
"step": 640
|
777 |
},
|
778 |
{
|
779 |
-
"epoch":
|
780 |
-
"learning_rate":
|
781 |
-
"loss": 1.
|
782 |
"step": 645
|
783 |
},
|
784 |
{
|
785 |
-
"epoch":
|
786 |
-
"learning_rate":
|
787 |
-
"loss": 1.
|
788 |
"step": 650
|
789 |
},
|
790 |
{
|
791 |
-
"epoch":
|
792 |
-
"learning_rate":
|
793 |
-
"loss": 1.
|
794 |
"step": 655
|
795 |
},
|
796 |
{
|
797 |
-
"epoch":
|
798 |
-
"learning_rate":
|
799 |
-
"loss": 1.
|
800 |
"step": 660
|
801 |
},
|
802 |
{
|
803 |
-
"epoch":
|
804 |
-
"learning_rate":
|
805 |
-
"loss": 1.
|
806 |
"step": 665
|
807 |
},
|
808 |
{
|
809 |
-
"epoch":
|
810 |
-
"learning_rate":
|
811 |
-
"loss": 1.
|
812 |
"step": 670
|
813 |
},
|
814 |
{
|
815 |
-
"epoch":
|
816 |
-
"learning_rate":
|
817 |
-
"loss": 1.
|
818 |
"step": 675
|
819 |
},
|
820 |
{
|
821 |
-
"epoch":
|
822 |
-
"learning_rate":
|
823 |
-
"loss": 1.
|
824 |
"step": 680
|
825 |
},
|
826 |
{
|
827 |
-
"epoch":
|
828 |
-
"learning_rate":
|
829 |
-
"loss": 1.
|
830 |
"step": 685
|
831 |
},
|
832 |
{
|
833 |
-
"epoch":
|
834 |
-
"learning_rate":
|
835 |
-
"loss": 1.
|
836 |
"step": 690
|
837 |
},
|
838 |
{
|
839 |
-
"epoch":
|
840 |
-
"learning_rate":
|
841 |
-
"loss": 1.
|
842 |
"step": 695
|
843 |
},
|
844 |
{
|
845 |
-
"epoch":
|
846 |
-
"learning_rate":
|
847 |
-
"loss": 1.
|
848 |
"step": 700
|
849 |
},
|
850 |
{
|
851 |
-
"epoch":
|
852 |
-
"learning_rate":
|
853 |
-
"loss": 1.
|
854 |
"step": 705
|
855 |
},
|
856 |
{
|
857 |
-
"epoch":
|
858 |
-
"learning_rate":
|
859 |
-
"loss": 1.
|
860 |
"step": 710
|
861 |
},
|
862 |
{
|
863 |
-
"epoch":
|
864 |
-
"learning_rate":
|
865 |
-
"loss": 1.
|
866 |
"step": 715
|
867 |
},
|
868 |
{
|
869 |
-
"epoch":
|
870 |
-
"learning_rate":
|
871 |
-
"loss": 1.
|
872 |
"step": 720
|
873 |
},
|
874 |
{
|
875 |
-
"epoch":
|
876 |
-
"learning_rate":
|
877 |
-
"loss": 1.
|
878 |
"step": 725
|
879 |
},
|
880 |
{
|
881 |
-
"epoch":
|
882 |
-
"learning_rate":
|
883 |
-
"loss": 1.
|
884 |
"step": 730
|
885 |
},
|
886 |
{
|
887 |
-
"epoch":
|
888 |
-
"learning_rate":
|
889 |
-
"loss": 1.
|
890 |
"step": 735
|
891 |
},
|
892 |
{
|
893 |
-
"epoch":
|
894 |
-
"learning_rate":
|
895 |
-
"loss": 1.
|
896 |
"step": 740
|
897 |
},
|
898 |
{
|
899 |
-
"epoch":
|
900 |
-
"learning_rate":
|
901 |
-
"loss": 1.
|
902 |
"step": 745
|
903 |
},
|
904 |
{
|
905 |
-
"epoch":
|
906 |
-
"learning_rate":
|
907 |
-
"loss": 1.
|
908 |
"step": 750
|
909 |
},
|
910 |
{
|
911 |
-
"epoch":
|
912 |
-
"learning_rate":
|
913 |
-
"loss": 1.
|
914 |
"step": 755
|
915 |
},
|
916 |
{
|
917 |
-
"epoch":
|
918 |
-
"learning_rate":
|
919 |
-
"loss": 1.
|
920 |
"step": 760
|
921 |
},
|
922 |
{
|
923 |
-
"epoch":
|
924 |
-
"learning_rate":
|
925 |
-
"loss": 1.
|
926 |
"step": 765
|
927 |
},
|
928 |
{
|
929 |
-
"epoch":
|
930 |
-
"learning_rate":
|
931 |
-
"loss": 1.
|
932 |
"step": 770
|
933 |
},
|
934 |
{
|
935 |
-
"epoch":
|
936 |
-
"learning_rate":
|
937 |
-
"loss": 1.
|
938 |
"step": 775
|
939 |
},
|
940 |
{
|
941 |
-
"epoch":
|
942 |
-
"learning_rate":
|
943 |
-
"loss": 1.
|
944 |
"step": 780
|
945 |
},
|
946 |
{
|
947 |
-
"epoch":
|
948 |
-
"learning_rate":
|
949 |
-
"loss": 1.
|
950 |
"step": 785
|
951 |
},
|
952 |
{
|
953 |
-
"epoch":
|
954 |
-
"learning_rate":
|
955 |
-
"loss": 1.
|
956 |
"step": 790
|
957 |
},
|
958 |
{
|
959 |
-
"epoch":
|
960 |
-
"learning_rate":
|
961 |
-
"loss": 1.
|
962 |
"step": 795
|
963 |
},
|
964 |
{
|
965 |
-
"epoch":
|
966 |
-
"learning_rate":
|
967 |
-
"loss": 1.
|
968 |
"step": 800
|
969 |
},
|
970 |
{
|
971 |
-
"epoch":
|
972 |
-
"learning_rate":
|
973 |
-
"loss": 1.
|
974 |
"step": 805
|
975 |
},
|
976 |
{
|
977 |
-
"epoch":
|
978 |
-
"learning_rate":
|
979 |
-
"loss": 1.
|
980 |
"step": 810
|
981 |
},
|
982 |
{
|
983 |
-
"epoch":
|
984 |
-
"learning_rate":
|
985 |
-
"loss": 1.
|
986 |
"step": 815
|
987 |
},
|
988 |
{
|
989 |
-
"epoch":
|
990 |
-
"learning_rate":
|
991 |
-
"loss": 1.
|
992 |
"step": 820
|
993 |
},
|
994 |
{
|
995 |
-
"epoch":
|
996 |
-
"learning_rate":
|
997 |
-
"loss": 1.
|
998 |
"step": 825
|
999 |
},
|
1000 |
{
|
1001 |
-
"epoch":
|
1002 |
-
"learning_rate":
|
1003 |
-
"loss": 1.
|
1004 |
"step": 830
|
1005 |
},
|
1006 |
{
|
1007 |
-
"epoch":
|
1008 |
-
"learning_rate":
|
1009 |
-
"loss": 1.
|
1010 |
"step": 835
|
1011 |
},
|
1012 |
{
|
1013 |
-
"epoch":
|
1014 |
-
"learning_rate":
|
1015 |
-
"loss": 1.
|
1016 |
"step": 840
|
1017 |
},
|
1018 |
{
|
1019 |
-
"epoch":
|
1020 |
-
"learning_rate":
|
1021 |
-
"loss": 1.
|
1022 |
"step": 845
|
1023 |
},
|
1024 |
{
|
1025 |
-
"epoch":
|
1026 |
-
"learning_rate":
|
1027 |
-
"loss": 1.
|
1028 |
"step": 850
|
1029 |
},
|
1030 |
{
|
1031 |
-
"epoch":
|
1032 |
-
"learning_rate":
|
1033 |
-
"loss": 1.
|
1034 |
"step": 855
|
1035 |
},
|
1036 |
{
|
1037 |
-
"epoch":
|
1038 |
-
"learning_rate":
|
1039 |
-
"loss": 1.
|
1040 |
"step": 860
|
1041 |
},
|
1042 |
{
|
1043 |
-
"epoch":
|
1044 |
-
"learning_rate":
|
1045 |
-
"loss": 1.
|
1046 |
"step": 865
|
1047 |
},
|
1048 |
{
|
1049 |
-
"epoch":
|
1050 |
-
"learning_rate":
|
1051 |
-
"loss": 1.
|
1052 |
"step": 870
|
1053 |
},
|
1054 |
{
|
1055 |
-
"epoch":
|
1056 |
-
"learning_rate":
|
1057 |
-
"loss": 1.
|
1058 |
"step": 875
|
1059 |
},
|
1060 |
{
|
1061 |
-
"epoch":
|
1062 |
-
"learning_rate":
|
1063 |
-
"loss": 1.
|
1064 |
"step": 880
|
1065 |
},
|
1066 |
{
|
1067 |
-
"epoch":
|
1068 |
-
"learning_rate":
|
1069 |
-
"loss": 1.
|
1070 |
"step": 885
|
1071 |
},
|
1072 |
{
|
1073 |
-
"epoch":
|
1074 |
-
"learning_rate":
|
1075 |
-
"loss": 1.
|
1076 |
"step": 890
|
1077 |
},
|
1078 |
{
|
1079 |
-
"epoch":
|
1080 |
-
"learning_rate":
|
1081 |
-
"loss": 1.
|
1082 |
"step": 895
|
1083 |
},
|
1084 |
{
|
1085 |
-
"epoch":
|
1086 |
-
"learning_rate":
|
1087 |
-
"loss": 1.
|
1088 |
"step": 900
|
1089 |
},
|
1090 |
{
|
1091 |
-
"epoch":
|
1092 |
-
"learning_rate":
|
1093 |
-
"loss": 1.
|
1094 |
"step": 905
|
1095 |
},
|
1096 |
{
|
1097 |
-
"epoch":
|
1098 |
-
"learning_rate":
|
1099 |
-
"loss": 1.
|
1100 |
"step": 910
|
1101 |
},
|
1102 |
{
|
1103 |
-
"epoch":
|
1104 |
-
"learning_rate":
|
1105 |
-
"loss": 1.
|
1106 |
"step": 915
|
1107 |
},
|
1108 |
{
|
1109 |
-
"epoch":
|
1110 |
-
"learning_rate":
|
1111 |
-
"loss": 1.
|
1112 |
"step": 920
|
1113 |
},
|
1114 |
{
|
1115 |
-
"epoch":
|
1116 |
-
"learning_rate":
|
1117 |
-
"loss": 1.
|
1118 |
"step": 925
|
1119 |
},
|
1120 |
{
|
1121 |
-
"epoch":
|
1122 |
-
"learning_rate":
|
1123 |
-
"loss": 1.
|
1124 |
"step": 930
|
1125 |
},
|
1126 |
{
|
1127 |
-
"epoch":
|
1128 |
-
"learning_rate":
|
1129 |
-
"loss": 1.
|
1130 |
"step": 935
|
1131 |
},
|
1132 |
{
|
1133 |
-
"epoch":
|
1134 |
-
"learning_rate":
|
1135 |
-
"loss": 1.
|
1136 |
"step": 940
|
1137 |
},
|
1138 |
{
|
1139 |
-
"epoch":
|
1140 |
-
"learning_rate":
|
1141 |
-
"loss": 1.
|
1142 |
"step": 945
|
1143 |
},
|
1144 |
{
|
1145 |
-
"epoch":
|
1146 |
-
"learning_rate":
|
1147 |
-
"loss": 1.
|
1148 |
"step": 950
|
1149 |
},
|
1150 |
{
|
1151 |
-
"epoch":
|
1152 |
-
"learning_rate":
|
1153 |
-
"loss": 1.
|
1154 |
"step": 955
|
1155 |
},
|
1156 |
{
|
1157 |
-
"epoch":
|
1158 |
-
"learning_rate":
|
1159 |
-
"loss": 1.
|
1160 |
"step": 960
|
1161 |
},
|
1162 |
{
|
1163 |
-
"epoch":
|
1164 |
-
"learning_rate":
|
1165 |
-
"loss": 1.
|
1166 |
"step": 965
|
1167 |
},
|
1168 |
{
|
1169 |
-
"epoch":
|
1170 |
-
"learning_rate":
|
1171 |
-
"loss": 1.
|
1172 |
"step": 970
|
1173 |
},
|
1174 |
{
|
1175 |
-
"epoch":
|
1176 |
-
"learning_rate":
|
1177 |
-
"loss": 1.
|
1178 |
"step": 975
|
1179 |
},
|
1180 |
{
|
1181 |
-
"epoch":
|
1182 |
-
"learning_rate":
|
1183 |
-
"loss": 1.
|
1184 |
"step": 980
|
1185 |
},
|
1186 |
{
|
1187 |
-
"epoch":
|
1188 |
-
"learning_rate":
|
1189 |
-
"loss": 1.
|
1190 |
"step": 985
|
1191 |
},
|
1192 |
{
|
1193 |
-
"epoch":
|
1194 |
-
"learning_rate":
|
1195 |
-
"loss": 1.
|
1196 |
"step": 990
|
1197 |
},
|
1198 |
{
|
1199 |
-
"epoch":
|
1200 |
-
"learning_rate":
|
1201 |
-
"loss": 1.
|
1202 |
"step": 995
|
1203 |
},
|
1204 |
{
|
1205 |
-
"epoch":
|
1206 |
-
"learning_rate":
|
1207 |
-
"loss": 1.
|
1208 |
"step": 1000
|
1209 |
},
|
1210 |
{
|
1211 |
-
"epoch":
|
1212 |
-
"learning_rate":
|
1213 |
-
"loss": 1.
|
1214 |
"step": 1005
|
1215 |
},
|
1216 |
{
|
1217 |
-
"epoch":
|
1218 |
-
"learning_rate":
|
1219 |
-
"loss": 1.
|
1220 |
"step": 1010
|
1221 |
},
|
1222 |
{
|
1223 |
-
"epoch":
|
1224 |
-
"learning_rate":
|
1225 |
-
"loss": 1.
|
1226 |
"step": 1015
|
1227 |
-
},
|
1228 |
-
{
|
1229 |
-
"epoch": 0.83,
|
1230 |
-
"learning_rate": 1.9322559524149603e-05,
|
1231 |
-
"loss": 1.5966,
|
1232 |
-
"step": 1020
|
1233 |
-
},
|
1234 |
-
{
|
1235 |
-
"epoch": 0.83,
|
1236 |
-
"learning_rate": 1.92301910949537e-05,
|
1237 |
-
"loss": 1.5624,
|
1238 |
-
"step": 1025
|
1239 |
-
},
|
1240 |
-
{
|
1241 |
-
"epoch": 0.83,
|
1242 |
-
"learning_rate": 1.9137648067253087e-05,
|
1243 |
-
"loss": 1.6706,
|
1244 |
-
"step": 1030
|
1245 |
-
},
|
1246 |
-
{
|
1247 |
-
"epoch": 0.84,
|
1248 |
-
"learning_rate": 1.9044934260703623e-05,
|
1249 |
-
"loss": 1.566,
|
1250 |
-
"step": 1035
|
1251 |
-
},
|
1252 |
-
{
|
1253 |
-
"epoch": 0.84,
|
1254 |
-
"learning_rate": 1.895205350200998e-05,
|
1255 |
-
"loss": 1.508,
|
1256 |
-
"step": 1040
|
1257 |
-
},
|
1258 |
-
{
|
1259 |
-
"epoch": 0.85,
|
1260 |
-
"learning_rate": 1.885900962476767e-05,
|
1261 |
-
"loss": 1.6144,
|
1262 |
-
"step": 1045
|
1263 |
-
},
|
1264 |
-
{
|
1265 |
-
"epoch": 0.85,
|
1266 |
-
"learning_rate": 1.8765806469304814e-05,
|
1267 |
-
"loss": 1.6291,
|
1268 |
-
"step": 1050
|
1269 |
-
},
|
1270 |
-
{
|
1271 |
-
"epoch": 0.85,
|
1272 |
-
"learning_rate": 1.8672447882523644e-05,
|
1273 |
-
"loss": 1.6088,
|
1274 |
-
"step": 1055
|
1275 |
-
},
|
1276 |
-
{
|
1277 |
-
"epoch": 0.86,
|
1278 |
-
"learning_rate": 1.8578937717741727e-05,
|
1279 |
-
"loss": 1.5956,
|
1280 |
-
"step": 1060
|
1281 |
-
},
|
1282 |
-
{
|
1283 |
-
"epoch": 0.86,
|
1284 |
-
"learning_rate": 1.8485279834532923e-05,
|
1285 |
-
"loss": 1.6554,
|
1286 |
-
"step": 1065
|
1287 |
-
},
|
1288 |
-
{
|
1289 |
-
"epoch": 0.87,
|
1290 |
-
"learning_rate": 1.839147809856807e-05,
|
1291 |
-
"loss": 1.6333,
|
1292 |
-
"step": 1070
|
1293 |
-
},
|
1294 |
-
{
|
1295 |
-
"epoch": 0.87,
|
1296 |
-
"learning_rate": 1.8297536381455434e-05,
|
1297 |
-
"loss": 1.6016,
|
1298 |
-
"step": 1075
|
1299 |
-
},
|
1300 |
-
{
|
1301 |
-
"epoch": 0.87,
|
1302 |
-
"learning_rate": 1.8203458560580934e-05,
|
1303 |
-
"loss": 1.5765,
|
1304 |
-
"step": 1080
|
1305 |
-
},
|
1306 |
-
{
|
1307 |
-
"epoch": 0.88,
|
1308 |
-
"learning_rate": 1.810924851894807e-05,
|
1309 |
-
"loss": 1.4411,
|
1310 |
-
"step": 1085
|
1311 |
-
},
|
1312 |
-
{
|
1313 |
-
"epoch": 0.88,
|
1314 |
-
"learning_rate": 1.801491014501768e-05,
|
1315 |
-
"loss": 1.6064,
|
1316 |
-
"step": 1090
|
1317 |
-
},
|
1318 |
-
{
|
1319 |
-
"epoch": 0.89,
|
1320 |
-
"learning_rate": 1.7920447332547423e-05,
|
1321 |
-
"loss": 1.5165,
|
1322 |
-
"step": 1095
|
1323 |
-
},
|
1324 |
-
{
|
1325 |
-
"epoch": 0.89,
|
1326 |
-
"learning_rate": 1.7825863980431106e-05,
|
1327 |
-
"loss": 1.628,
|
1328 |
-
"step": 1100
|
1329 |
-
},
|
1330 |
-
{
|
1331 |
-
"epoch": 0.89,
|
1332 |
-
"learning_rate": 1.773116399253772e-05,
|
1333 |
-
"loss": 1.5792,
|
1334 |
-
"step": 1105
|
1335 |
-
},
|
1336 |
-
{
|
1337 |
-
"epoch": 0.9,
|
1338 |
-
"learning_rate": 1.7636351277550324e-05,
|
1339 |
-
"loss": 1.5908,
|
1340 |
-
"step": 1110
|
1341 |
-
},
|
1342 |
-
{
|
1343 |
-
"epoch": 0.9,
|
1344 |
-
"learning_rate": 1.7541429748804722e-05,
|
1345 |
-
"loss": 1.5477,
|
1346 |
-
"step": 1115
|
1347 |
-
},
|
1348 |
-
{
|
1349 |
-
"epoch": 0.91,
|
1350 |
-
"learning_rate": 1.744640332412794e-05,
|
1351 |
-
"loss": 1.6118,
|
1352 |
-
"step": 1120
|
1353 |
-
},
|
1354 |
-
{
|
1355 |
-
"epoch": 0.91,
|
1356 |
-
"learning_rate": 1.7351275925676517e-05,
|
1357 |
-
"loss": 1.6034,
|
1358 |
-
"step": 1125
|
1359 |
-
},
|
1360 |
-
{
|
1361 |
-
"epoch": 0.91,
|
1362 |
-
"learning_rate": 1.725605147977461e-05,
|
1363 |
-
"loss": 1.6053,
|
1364 |
-
"step": 1130
|
1365 |
-
},
|
1366 |
-
{
|
1367 |
-
"epoch": 0.92,
|
1368 |
-
"learning_rate": 1.716073391675197e-05,
|
1369 |
-
"loss": 1.5638,
|
1370 |
-
"step": 1135
|
1371 |
-
},
|
1372 |
-
{
|
1373 |
-
"epoch": 0.92,
|
1374 |
-
"learning_rate": 1.7065327170781678e-05,
|
1375 |
-
"loss": 1.5809,
|
1376 |
-
"step": 1140
|
1377 |
-
},
|
1378 |
-
{
|
1379 |
-
"epoch": 0.93,
|
1380 |
-
"learning_rate": 1.69698351797178e-05,
|
1381 |
-
"loss": 1.5822,
|
1382 |
-
"step": 1145
|
1383 |
-
},
|
1384 |
-
{
|
1385 |
-
"epoch": 0.93,
|
1386 |
-
"learning_rate": 1.6874261884932842e-05,
|
1387 |
-
"loss": 1.5981,
|
1388 |
-
"step": 1150
|
1389 |
-
},
|
1390 |
-
{
|
1391 |
-
"epoch": 0.93,
|
1392 |
-
"learning_rate": 1.677861123115506e-05,
|
1393 |
-
"loss": 1.5995,
|
1394 |
-
"step": 1155
|
1395 |
-
},
|
1396 |
-
{
|
1397 |
-
"epoch": 0.94,
|
1398 |
-
"learning_rate": 1.668288716630566e-05,
|
1399 |
-
"loss": 1.5323,
|
1400 |
-
"step": 1160
|
1401 |
-
},
|
1402 |
-
{
|
1403 |
-
"epoch": 0.94,
|
1404 |
-
"learning_rate": 1.6587093641335844e-05,
|
1405 |
-
"loss": 1.5981,
|
1406 |
-
"step": 1165
|
1407 |
-
},
|
1408 |
-
{
|
1409 |
-
"epoch": 0.95,
|
1410 |
-
"learning_rate": 1.6491234610063732e-05,
|
1411 |
-
"loss": 1.6448,
|
1412 |
-
"step": 1170
|
1413 |
-
},
|
1414 |
-
{
|
1415 |
-
"epoch": 0.95,
|
1416 |
-
"learning_rate": 1.6395314029011184e-05,
|
1417 |
-
"loss": 1.63,
|
1418 |
-
"step": 1175
|
1419 |
-
},
|
1420 |
-
{
|
1421 |
-
"epoch": 0.95,
|
1422 |
-
"learning_rate": 1.6299335857240484e-05,
|
1423 |
-
"loss": 1.6529,
|
1424 |
-
"step": 1180
|
1425 |
-
},
|
1426 |
-
{
|
1427 |
-
"epoch": 0.96,
|
1428 |
-
"learning_rate": 1.6203304056190938e-05,
|
1429 |
-
"loss": 1.6172,
|
1430 |
-
"step": 1185
|
1431 |
-
},
|
1432 |
-
{
|
1433 |
-
"epoch": 0.96,
|
1434 |
-
"learning_rate": 1.6107222589515376e-05,
|
1435 |
-
"loss": 1.5247,
|
1436 |
-
"step": 1190
|
1437 |
-
},
|
1438 |
-
{
|
1439 |
-
"epoch": 0.97,
|
1440 |
-
"learning_rate": 1.601109542291652e-05,
|
1441 |
-
"loss": 1.5932,
|
1442 |
-
"step": 1195
|
1443 |
-
},
|
1444 |
-
{
|
1445 |
-
"epoch": 0.97,
|
1446 |
-
"learning_rate": 1.591492652398336e-05,
|
1447 |
-
"loss": 1.6186,
|
1448 |
-
"step": 1200
|
1449 |
-
},
|
1450 |
-
{
|
1451 |
-
"epoch": 0.98,
|
1452 |
-
"learning_rate": 1.5818719862027338e-05,
|
1453 |
-
"loss": 1.6524,
|
1454 |
-
"step": 1205
|
1455 |
-
},
|
1456 |
-
{
|
1457 |
-
"epoch": 0.98,
|
1458 |
-
"learning_rate": 1.572247940791856e-05,
|
1459 |
-
"loss": 1.611,
|
1460 |
-
"step": 1210
|
1461 |
-
},
|
1462 |
-
{
|
1463 |
-
"epoch": 0.98,
|
1464 |
-
"learning_rate": 1.5626209133921863e-05,
|
1465 |
-
"loss": 1.5913,
|
1466 |
-
"step": 1215
|
1467 |
-
},
|
1468 |
-
{
|
1469 |
-
"epoch": 0.99,
|
1470 |
-
"learning_rate": 1.5529913013532894e-05,
|
1471 |
-
"loss": 1.7348,
|
1472 |
-
"step": 1220
|
1473 |
-
},
|
1474 |
-
{
|
1475 |
-
"epoch": 0.99,
|
1476 |
-
"learning_rate": 1.543359502131409e-05,
|
1477 |
-
"loss": 1.719,
|
1478 |
-
"step": 1225
|
1479 |
-
},
|
1480 |
-
{
|
1481 |
-
"epoch": 1.0,
|
1482 |
-
"learning_rate": 1.533725913273064e-05,
|
1483 |
-
"loss": 1.6274,
|
1484 |
-
"step": 1230
|
1485 |
-
},
|
1486 |
-
{
|
1487 |
-
"epoch": 1.0,
|
1488 |
-
"learning_rate": 1.5240909323986384e-05,
|
1489 |
-
"loss": 1.5266,
|
1490 |
-
"step": 1235
|
1491 |
-
},
|
1492 |
-
{
|
1493 |
-
"epoch": 1.0,
|
1494 |
-
"eval_loss": 3.4220237731933594,
|
1495 |
-
"eval_runtime": 954.9678,
|
1496 |
-
"eval_samples_per_second": 9.114,
|
1497 |
-
"eval_steps_per_second": 2.279,
|
1498 |
-
"step": 1235
|
1499 |
-
},
|
1500 |
-
{
|
1501 |
-
"epoch": 1.0,
|
1502 |
-
"learning_rate": 1.5144549571859711e-05,
|
1503 |
-
"loss": 1.3304,
|
1504 |
-
"step": 1240
|
1505 |
-
},
|
1506 |
-
{
|
1507 |
-
"epoch": 1.01,
|
1508 |
-
"learning_rate": 1.5048183853539425e-05,
|
1509 |
-
"loss": 1.0536,
|
1510 |
-
"step": 1245
|
1511 |
-
},
|
1512 |
-
{
|
1513 |
-
"epoch": 1.01,
|
1514 |
-
"learning_rate": 1.4951816146460574e-05,
|
1515 |
-
"loss": 1.0564,
|
1516 |
-
"step": 1250
|
1517 |
-
},
|
1518 |
-
{
|
1519 |
-
"epoch": 1.02,
|
1520 |
-
"learning_rate": 1.4855450428140291e-05,
|
1521 |
-
"loss": 1.0105,
|
1522 |
-
"step": 1255
|
1523 |
-
},
|
1524 |
-
{
|
1525 |
-
"epoch": 1.02,
|
1526 |
-
"learning_rate": 1.4759090676013616e-05,
|
1527 |
-
"loss": 1.0252,
|
1528 |
-
"step": 1260
|
1529 |
-
},
|
1530 |
-
{
|
1531 |
-
"epoch": 1.02,
|
1532 |
-
"learning_rate": 1.4662740867269361e-05,
|
1533 |
-
"loss": 1.0029,
|
1534 |
-
"step": 1265
|
1535 |
-
},
|
1536 |
-
{
|
1537 |
-
"epoch": 1.03,
|
1538 |
-
"learning_rate": 1.4566404978685912e-05,
|
1539 |
-
"loss": 0.9685,
|
1540 |
-
"step": 1270
|
1541 |
-
},
|
1542 |
-
{
|
1543 |
-
"epoch": 1.03,
|
1544 |
-
"learning_rate": 1.447008698646711e-05,
|
1545 |
-
"loss": 1.0095,
|
1546 |
-
"step": 1275
|
1547 |
-
},
|
1548 |
-
{
|
1549 |
-
"epoch": 1.04,
|
1550 |
-
"learning_rate": 1.4373790866078143e-05,
|
1551 |
-
"loss": 0.9759,
|
1552 |
-
"step": 1280
|
1553 |
-
},
|
1554 |
-
{
|
1555 |
-
"epoch": 1.04,
|
1556 |
-
"learning_rate": 1.4277520592081442e-05,
|
1557 |
-
"loss": 1.0217,
|
1558 |
-
"step": 1285
|
1559 |
-
},
|
1560 |
-
{
|
1561 |
-
"epoch": 1.04,
|
1562 |
-
"learning_rate": 1.4181280137972666e-05,
|
1563 |
-
"loss": 0.9472,
|
1564 |
-
"step": 1290
|
1565 |
-
},
|
1566 |
-
{
|
1567 |
-
"epoch": 1.05,
|
1568 |
-
"learning_rate": 1.4085073476016642e-05,
|
1569 |
-
"loss": 1.0432,
|
1570 |
-
"step": 1295
|
1571 |
-
},
|
1572 |
-
{
|
1573 |
-
"epoch": 1.05,
|
1574 |
-
"learning_rate": 1.3988904577083481e-05,
|
1575 |
-
"loss": 1.0355,
|
1576 |
-
"step": 1300
|
1577 |
-
},
|
1578 |
-
{
|
1579 |
-
"epoch": 1.06,
|
1580 |
-
"learning_rate": 1.389277741048463e-05,
|
1581 |
-
"loss": 0.9971,
|
1582 |
-
"step": 1305
|
1583 |
-
},
|
1584 |
-
{
|
1585 |
-
"epoch": 1.06,
|
1586 |
-
"learning_rate": 1.3796695943809063e-05,
|
1587 |
-
"loss": 0.9947,
|
1588 |
-
"step": 1310
|
1589 |
-
},
|
1590 |
-
{
|
1591 |
-
"epoch": 1.06,
|
1592 |
-
"learning_rate": 1.3700664142759521e-05,
|
1593 |
-
"loss": 1.0532,
|
1594 |
-
"step": 1315
|
1595 |
-
},
|
1596 |
-
{
|
1597 |
-
"epoch": 1.07,
|
1598 |
-
"learning_rate": 1.3604685970988817e-05,
|
1599 |
-
"loss": 0.9791,
|
1600 |
-
"step": 1320
|
1601 |
-
},
|
1602 |
-
{
|
1603 |
-
"epoch": 1.07,
|
1604 |
-
"learning_rate": 1.3508765389936272e-05,
|
1605 |
-
"loss": 1.0299,
|
1606 |
-
"step": 1325
|
1607 |
-
},
|
1608 |
-
{
|
1609 |
-
"epoch": 1.08,
|
1610 |
-
"learning_rate": 1.341290635866416e-05,
|
1611 |
-
"loss": 1.036,
|
1612 |
-
"step": 1330
|
1613 |
-
},
|
1614 |
-
{
|
1615 |
-
"epoch": 1.08,
|
1616 |
-
"learning_rate": 1.3317112833694344e-05,
|
1617 |
-
"loss": 1.0534,
|
1618 |
-
"step": 1335
|
1619 |
-
},
|
1620 |
-
{
|
1621 |
-
"epoch": 1.08,
|
1622 |
-
"learning_rate": 1.3221388768844937e-05,
|
1623 |
-
"loss": 1.0144,
|
1624 |
-
"step": 1340
|
1625 |
-
},
|
1626 |
-
{
|
1627 |
-
"epoch": 1.09,
|
1628 |
-
"learning_rate": 1.3125738115067159e-05,
|
1629 |
-
"loss": 1.0696,
|
1630 |
-
"step": 1345
|
1631 |
-
},
|
1632 |
-
{
|
1633 |
-
"epoch": 1.09,
|
1634 |
-
"learning_rate": 1.30301648202822e-05,
|
1635 |
-
"loss": 1.0651,
|
1636 |
-
"step": 1350
|
1637 |
-
},
|
1638 |
-
{
|
1639 |
-
"epoch": 1.1,
|
1640 |
-
"learning_rate": 1.2934672829218327e-05,
|
1641 |
-
"loss": 1.0598,
|
1642 |
-
"step": 1355
|
1643 |
-
},
|
1644 |
-
{
|
1645 |
-
"epoch": 1.1,
|
1646 |
-
"learning_rate": 1.283926608324804e-05,
|
1647 |
-
"loss": 1.0514,
|
1648 |
-
"step": 1360
|
1649 |
-
},
|
1650 |
-
{
|
1651 |
-
"epoch": 1.11,
|
1652 |
-
"learning_rate": 1.2743948520225391e-05,
|
1653 |
-
"loss": 0.9963,
|
1654 |
-
"step": 1365
|
1655 |
-
},
|
1656 |
-
{
|
1657 |
-
"epoch": 1.11,
|
1658 |
-
"learning_rate": 1.2648724074323492e-05,
|
1659 |
-
"loss": 1.0307,
|
1660 |
-
"step": 1370
|
1661 |
-
},
|
1662 |
-
{
|
1663 |
-
"epoch": 1.11,
|
1664 |
-
"learning_rate": 1.255359667587206e-05,
|
1665 |
-
"loss": 1.0354,
|
1666 |
-
"step": 1375
|
1667 |
-
},
|
1668 |
-
{
|
1669 |
-
"epoch": 1.12,
|
1670 |
-
"learning_rate": 1.2458570251195279e-05,
|
1671 |
-
"loss": 1.0965,
|
1672 |
-
"step": 1380
|
1673 |
-
},
|
1674 |
-
{
|
1675 |
-
"epoch": 1.12,
|
1676 |
-
"learning_rate": 1.2363648722449679e-05,
|
1677 |
-
"loss": 1.0484,
|
1678 |
-
"step": 1385
|
1679 |
-
},
|
1680 |
-
{
|
1681 |
-
"epoch": 1.13,
|
1682 |
-
"learning_rate": 1.2268836007462284e-05,
|
1683 |
-
"loss": 1.0421,
|
1684 |
-
"step": 1390
|
1685 |
-
},
|
1686 |
-
{
|
1687 |
-
"epoch": 1.13,
|
1688 |
-
"learning_rate": 1.21741360195689e-05,
|
1689 |
-
"loss": 1.0695,
|
1690 |
-
"step": 1395
|
1691 |
-
},
|
1692 |
-
{
|
1693 |
-
"epoch": 1.13,
|
1694 |
-
"learning_rate": 1.207955266745258e-05,
|
1695 |
-
"loss": 1.0323,
|
1696 |
-
"step": 1400
|
1697 |
-
},
|
1698 |
-
{
|
1699 |
-
"epoch": 1.14,
|
1700 |
-
"learning_rate": 1.198508985498233e-05,
|
1701 |
-
"loss": 1.0874,
|
1702 |
-
"step": 1405
|
1703 |
-
},
|
1704 |
-
{
|
1705 |
-
"epoch": 1.14,
|
1706 |
-
"learning_rate": 1.1890751481051933e-05,
|
1707 |
-
"loss": 1.0702,
|
1708 |
-
"step": 1410
|
1709 |
-
},
|
1710 |
-
{
|
1711 |
-
"epoch": 1.15,
|
1712 |
-
"learning_rate": 1.179654143941907e-05,
|
1713 |
-
"loss": 1.0211,
|
1714 |
-
"step": 1415
|
1715 |
-
},
|
1716 |
-
{
|
1717 |
-
"epoch": 1.15,
|
1718 |
-
"learning_rate": 1.1702463618544562e-05,
|
1719 |
-
"loss": 1.0252,
|
1720 |
-
"step": 1420
|
1721 |
-
},
|
1722 |
-
{
|
1723 |
-
"epoch": 1.15,
|
1724 |
-
"learning_rate": 1.1608521901431932e-05,
|
1725 |
-
"loss": 1.0588,
|
1726 |
-
"step": 1425
|
1727 |
-
},
|
1728 |
-
{
|
1729 |
-
"epoch": 1.16,
|
1730 |
-
"learning_rate": 1.1514720165467076e-05,
|
1731 |
-
"loss": 1.0644,
|
1732 |
-
"step": 1430
|
1733 |
-
},
|
1734 |
-
{
|
1735 |
-
"epoch": 1.16,
|
1736 |
-
"learning_rate": 1.1421062282258276e-05,
|
1737 |
-
"loss": 1.0831,
|
1738 |
-
"step": 1435
|
1739 |
-
},
|
1740 |
-
{
|
1741 |
-
"epoch": 1.17,
|
1742 |
-
"learning_rate": 1.1327552117476363e-05,
|
1743 |
-
"loss": 1.0262,
|
1744 |
-
"step": 1440
|
1745 |
-
},
|
1746 |
-
{
|
1747 |
-
"epoch": 1.17,
|
1748 |
-
"learning_rate": 1.1234193530695189e-05,
|
1749 |
-
"loss": 1.0775,
|
1750 |
-
"step": 1445
|
1751 |
-
},
|
1752 |
-
{
|
1753 |
-
"epoch": 1.17,
|
1754 |
-
"learning_rate": 1.1140990375232336e-05,
|
1755 |
-
"loss": 1.0476,
|
1756 |
-
"step": 1450
|
1757 |
-
},
|
1758 |
-
{
|
1759 |
-
"epoch": 1.18,
|
1760 |
-
"learning_rate": 1.1047946497990016e-05,
|
1761 |
-
"loss": 1.0373,
|
1762 |
-
"step": 1455
|
1763 |
-
},
|
1764 |
-
{
|
1765 |
-
"epoch": 1.18,
|
1766 |
-
"learning_rate": 1.0955065739296378e-05,
|
1767 |
-
"loss": 1.0734,
|
1768 |
-
"step": 1460
|
1769 |
}
|
1770 |
],
|
1771 |
-
"max_steps":
|
1772 |
"num_train_epochs": 2,
|
1773 |
-
"total_flos":
|
1774 |
"trial_name": null,
|
1775 |
"trial_params": null
|
1776 |
}
|
|
|
1 |
{
|
2 |
"best_metric": null,
|
3 |
"best_model_checkpoint": null,
|
4 |
+
"epoch": 1.6457290244272493,
|
5 |
+
"global_step": 1016,
|
6 |
"is_hyper_param_search": false,
|
7 |
"is_local_process_zero": true,
|
8 |
"is_world_process_zero": true,
|
9 |
"log_history": [
|
10 |
{
|
11 |
+
"epoch": 0.01,
|
12 |
+
"learning_rate": 2.4193548387096776e-06,
|
13 |
+
"loss": 3.4647,
|
14 |
"step": 5
|
15 |
},
|
16 |
{
|
17 |
+
"epoch": 0.02,
|
18 |
+
"learning_rate": 4.838709677419355e-06,
|
19 |
+
"loss": 2.4108,
|
20 |
"step": 10
|
21 |
},
|
22 |
{
|
23 |
+
"epoch": 0.02,
|
24 |
+
"learning_rate": 7.258064516129032e-06,
|
25 |
+
"loss": 1.5747,
|
26 |
"step": 15
|
27 |
},
|
28 |
{
|
29 |
+
"epoch": 0.03,
|
30 |
+
"learning_rate": 9.67741935483871e-06,
|
31 |
+
"loss": 1.5005,
|
32 |
"step": 20
|
33 |
},
|
34 |
{
|
35 |
+
"epoch": 0.04,
|
36 |
+
"learning_rate": 1.2096774193548387e-05,
|
37 |
+
"loss": 1.4694,
|
38 |
"step": 25
|
39 |
},
|
40 |
{
|
41 |
+
"epoch": 0.05,
|
42 |
+
"learning_rate": 1.4516129032258065e-05,
|
43 |
+
"loss": 1.4399,
|
44 |
"step": 30
|
45 |
},
|
46 |
{
|
47 |
+
"epoch": 0.06,
|
48 |
+
"learning_rate": 1.6935483870967744e-05,
|
49 |
+
"loss": 1.4215,
|
50 |
"step": 35
|
51 |
},
|
52 |
{
|
53 |
+
"epoch": 0.06,
|
54 |
+
"learning_rate": 1.935483870967742e-05,
|
55 |
+
"loss": 1.4526,
|
56 |
"step": 40
|
57 |
},
|
58 |
{
|
59 |
+
"epoch": 0.07,
|
60 |
+
"learning_rate": 2.1774193548387097e-05,
|
61 |
+
"loss": 1.4458,
|
62 |
"step": 45
|
63 |
},
|
64 |
{
|
65 |
+
"epoch": 0.08,
|
66 |
+
"learning_rate": 2.4193548387096773e-05,
|
67 |
+
"loss": 1.4237,
|
68 |
"step": 50
|
69 |
},
|
70 |
{
|
71 |
+
"epoch": 0.09,
|
72 |
+
"learning_rate": 2.6612903225806453e-05,
|
73 |
+
"loss": 1.4339,
|
74 |
"step": 55
|
75 |
},
|
76 |
{
|
77 |
+
"epoch": 0.1,
|
78 |
+
"learning_rate": 2.903225806451613e-05,
|
79 |
+
"loss": 1.4294,
|
80 |
"step": 60
|
81 |
},
|
82 |
{
|
83 |
+
"epoch": 0.11,
|
84 |
+
"learning_rate": 2.999951499529191e-05,
|
85 |
+
"loss": 1.4968,
|
86 |
"step": 65
|
87 |
},
|
88 |
{
|
89 |
+
"epoch": 0.11,
|
90 |
+
"learning_rate": 2.9996551191211948e-05,
|
91 |
+
"loss": 1.4261,
|
92 |
"step": 70
|
93 |
},
|
94 |
{
|
95 |
+
"epoch": 0.12,
|
96 |
+
"learning_rate": 2.9990893561853812e-05,
|
97 |
+
"loss": 1.4371,
|
98 |
"step": 75
|
99 |
},
|
100 |
{
|
101 |
+
"epoch": 0.13,
|
102 |
+
"learning_rate": 2.9982543123495507e-05,
|
103 |
+
"loss": 1.4412,
|
104 |
"step": 80
|
105 |
},
|
106 |
{
|
107 |
+
"epoch": 0.14,
|
108 |
+
"learning_rate": 2.9971501376123366e-05,
|
109 |
+
"loss": 1.4638,
|
110 |
"step": 85
|
111 |
},
|
112 |
{
|
113 |
+
"epoch": 0.15,
|
114 |
+
"learning_rate": 2.9957770303162634e-05,
|
115 |
+
"loss": 1.4498,
|
116 |
"step": 90
|
117 |
},
|
118 |
{
|
119 |
+
"epoch": 0.15,
|
120 |
+
"learning_rate": 2.9941352371121173e-05,
|
121 |
+
"loss": 1.4393,
|
122 |
"step": 95
|
123 |
},
|
124 |
{
|
125 |
+
"epoch": 0.16,
|
126 |
+
"learning_rate": 2.992225052914641e-05,
|
127 |
+
"loss": 1.4291,
|
128 |
"step": 100
|
129 |
},
|
130 |
{
|
131 |
+
"epoch": 0.17,
|
132 |
+
"learning_rate": 2.990046820849558e-05,
|
133 |
+
"loss": 1.4587,
|
134 |
"step": 105
|
135 |
},
|
136 |
{
|
137 |
+
"epoch": 0.18,
|
138 |
+
"learning_rate": 2.9876009321919372e-05,
|
139 |
+
"loss": 1.4272,
|
140 |
"step": 110
|
141 |
},
|
142 |
{
|
143 |
+
"epoch": 0.19,
|
144 |
+
"learning_rate": 2.9848878262959076e-05,
|
145 |
+
"loss": 1.4622,
|
146 |
"step": 115
|
147 |
},
|
148 |
{
|
149 |
+
"epoch": 0.19,
|
150 |
+
"learning_rate": 2.981907990515739e-05,
|
151 |
+
"loss": 1.4863,
|
152 |
"step": 120
|
153 |
},
|
154 |
{
|
155 |
+
"epoch": 0.2,
|
156 |
+
"learning_rate": 2.9786619601182965e-05,
|
157 |
+
"loss": 1.4274,
|
158 |
"step": 125
|
159 |
},
|
160 |
{
|
161 |
+
"epoch": 0.21,
|
162 |
+
"learning_rate": 2.975150318186892e-05,
|
163 |
+
"loss": 1.4382,
|
164 |
"step": 130
|
165 |
},
|
166 |
{
|
167 |
+
"epoch": 0.22,
|
168 |
+
"learning_rate": 2.9713736955165456e-05,
|
169 |
+
"loss": 1.4413,
|
170 |
"step": 135
|
171 |
},
|
172 |
{
|
173 |
+
"epoch": 0.23,
|
174 |
+
"learning_rate": 2.9673327705006745e-05,
|
175 |
+
"loss": 1.431,
|
176 |
"step": 140
|
177 |
},
|
178 |
{
|
179 |
+
"epoch": 0.23,
|
180 |
+
"learning_rate": 2.963028269009235e-05,
|
181 |
+
"loss": 1.4659,
|
182 |
"step": 145
|
183 |
},
|
184 |
{
|
185 |
+
"epoch": 0.24,
|
186 |
+
"learning_rate": 2.9584609642583337e-05,
|
187 |
+
"loss": 1.4426,
|
188 |
"step": 150
|
189 |
},
|
190 |
{
|
191 |
+
"epoch": 0.25,
|
192 |
+
"learning_rate": 2.9536316766713357e-05,
|
193 |
+
"loss": 1.4733,
|
194 |
"step": 155
|
195 |
},
|
196 |
{
|
197 |
+
"epoch": 0.26,
|
198 |
+
"learning_rate": 2.9485412737314923e-05,
|
199 |
+
"loss": 1.4595,
|
200 |
"step": 160
|
201 |
},
|
202 |
{
|
203 |
+
"epoch": 0.27,
|
204 |
+
"learning_rate": 2.9431906698261136e-05,
|
205 |
+
"loss": 1.4845,
|
206 |
"step": 165
|
207 |
},
|
208 |
{
|
209 |
+
"epoch": 0.28,
|
210 |
+
"learning_rate": 2.9375808260823192e-05,
|
211 |
+
"loss": 1.5219,
|
212 |
"step": 170
|
213 |
},
|
214 |
{
|
215 |
+
"epoch": 0.28,
|
216 |
+
"learning_rate": 2.931712750194392e-05,
|
217 |
+
"loss": 1.532,
|
218 |
"step": 175
|
219 |
},
|
220 |
{
|
221 |
+
"epoch": 0.29,
|
222 |
+
"learning_rate": 2.9255874962427638e-05,
|
223 |
+
"loss": 1.4632,
|
224 |
"step": 180
|
225 |
},
|
226 |
{
|
227 |
+
"epoch": 0.3,
|
228 |
+
"learning_rate": 2.9192061645046724e-05,
|
229 |
+
"loss": 1.5057,
|
230 |
"step": 185
|
231 |
},
|
232 |
{
|
233 |
+
"epoch": 0.31,
|
234 |
+
"learning_rate": 2.9125699012565204e-05,
|
235 |
+
"loss": 1.5023,
|
236 |
"step": 190
|
237 |
},
|
238 |
{
|
239 |
+
"epoch": 0.32,
|
240 |
+
"learning_rate": 2.9056798985679688e-05,
|
241 |
+
"loss": 1.4423,
|
242 |
"step": 195
|
243 |
},
|
244 |
{
|
245 |
+
"epoch": 0.32,
|
246 |
+
"learning_rate": 2.8985373940878053e-05,
|
247 |
+
"loss": 1.4968,
|
248 |
"step": 200
|
249 |
},
|
250 |
{
|
251 |
+
"epoch": 0.33,
|
252 |
+
"learning_rate": 2.8911436708216276e-05,
|
253 |
+
"loss": 1.4767,
|
254 |
"step": 205
|
255 |
},
|
256 |
{
|
257 |
+
"epoch": 0.34,
|
258 |
+
"learning_rate": 2.883500056901376e-05,
|
259 |
+
"loss": 1.4968,
|
260 |
"step": 210
|
261 |
},
|
262 |
{
|
263 |
+
"epoch": 0.35,
|
264 |
+
"learning_rate": 2.875607925346762e-05,
|
265 |
+
"loss": 1.4842,
|
266 |
"step": 215
|
267 |
},
|
268 |
{
|
269 |
+
"epoch": 0.36,
|
270 |
+
"learning_rate": 2.867468693818634e-05,
|
271 |
+
"loss": 1.4921,
|
272 |
"step": 220
|
273 |
},
|
274 |
{
|
275 |
+
"epoch": 0.36,
|
276 |
+
"learning_rate": 2.859083824364323e-05,
|
277 |
+
"loss": 1.4969,
|
278 |
"step": 225
|
279 |
},
|
280 |
{
|
281 |
+
"epoch": 0.37,
|
282 |
+
"learning_rate": 2.8504548231550143e-05,
|
283 |
+
"loss": 1.4423,
|
284 |
"step": 230
|
285 |
},
|
286 |
{
|
287 |
+
"epoch": 0.38,
|
288 |
+
"learning_rate": 2.8415832402151956e-05,
|
289 |
+
"loss": 1.5165,
|
290 |
"step": 235
|
291 |
},
|
292 |
{
|
293 |
+
"epoch": 0.39,
|
294 |
+
"learning_rate": 2.832470669144227e-05,
|
295 |
+
"loss": 1.4715,
|
296 |
"step": 240
|
297 |
},
|
298 |
{
|
299 |
+
"epoch": 0.4,
|
300 |
+
"learning_rate": 2.8231187468300836e-05,
|
301 |
+
"loss": 1.4747,
|
302 |
"step": 245
|
303 |
},
|
304 |
{
|
305 |
+
"epoch": 0.4,
|
306 |
+
"learning_rate": 2.8135291531553192e-05,
|
307 |
+
"loss": 1.466,
|
308 |
"step": 250
|
309 |
},
|
310 |
{
|
311 |
+
"epoch": 0.41,
|
312 |
+
"learning_rate": 2.8037036106953134e-05,
|
313 |
+
"loss": 1.4609,
|
314 |
"step": 255
|
315 |
},
|
316 |
{
|
317 |
+
"epoch": 0.42,
|
318 |
+
"learning_rate": 2.793643884408843e-05,
|
319 |
+
"loss": 1.4649,
|
320 |
"step": 260
|
321 |
},
|
322 |
{
|
323 |
+
"epoch": 0.43,
|
324 |
+
"learning_rate": 2.7833517813210437e-05,
|
325 |
+
"loss": 1.4923,
|
326 |
"step": 265
|
327 |
},
|
328 |
{
|
329 |
+
"epoch": 0.44,
|
330 |
+
"learning_rate": 2.7728291501988173e-05,
|
331 |
+
"loss": 1.4968,
|
332 |
"step": 270
|
333 |
},
|
334 |
{
|
335 |
+
"epoch": 0.45,
|
336 |
+
"learning_rate": 2.7620778812187338e-05,
|
337 |
+
"loss": 1.4545,
|
338 |
"step": 275
|
339 |
},
|
340 |
{
|
341 |
+
"epoch": 0.45,
|
342 |
+
"learning_rate": 2.7510999056275038e-05,
|
343 |
+
"loss": 1.4791,
|
344 |
"step": 280
|
345 |
},
|
346 |
{
|
347 |
+
"epoch": 0.46,
|
348 |
+
"learning_rate": 2.739897195395067e-05,
|
349 |
+
"loss": 1.4849,
|
350 |
"step": 285
|
351 |
},
|
352 |
{
|
353 |
+
"epoch": 0.47,
|
354 |
+
"learning_rate": 2.728471762860369e-05,
|
355 |
+
"loss": 1.4492,
|
356 |
"step": 290
|
357 |
},
|
358 |
{
|
359 |
+
"epoch": 0.48,
|
360 |
+
"learning_rate": 2.716825660369885e-05,
|
361 |
+
"loss": 1.5052,
|
362 |
"step": 295
|
363 |
},
|
364 |
{
|
365 |
+
"epoch": 0.49,
|
366 |
+
"learning_rate": 2.704960979908957e-05,
|
367 |
+
"loss": 1.4701,
|
368 |
"step": 300
|
369 |
},
|
370 |
{
|
371 |
+
"epoch": 0.49,
|
372 |
+
"learning_rate": 2.6928798527260127e-05,
|
373 |
+
"loss": 1.48,
|
374 |
"step": 305
|
375 |
},
|
376 |
{
|
377 |
+
"epoch": 0.5,
|
378 |
+
"learning_rate": 2.680584448949729e-05,
|
379 |
+
"loss": 1.5158,
|
380 |
"step": 310
|
381 |
},
|
382 |
{
|
383 |
+
"epoch": 0.51,
|
384 |
+
"learning_rate": 2.6680769771992136e-05,
|
385 |
+
"loss": 1.5047,
|
386 |
"step": 315
|
387 |
},
|
388 |
{
|
389 |
+
"epoch": 0.52,
|
390 |
+
"learning_rate": 2.6553596841872682e-05,
|
391 |
+
"loss": 1.5023,
|
392 |
"step": 320
|
393 |
},
|
394 |
{
|
395 |
+
"epoch": 0.53,
|
396 |
+
"learning_rate": 2.6424348543168177e-05,
|
397 |
+
"loss": 1.4932,
|
398 |
"step": 325
|
399 |
},
|
400 |
{
|
401 |
+
"epoch": 0.53,
|
402 |
+
"learning_rate": 2.6293048092705586e-05,
|
403 |
+
"loss": 1.5241,
|
404 |
"step": 330
|
405 |
},
|
406 |
{
|
407 |
+
"epoch": 0.54,
|
408 |
+
"learning_rate": 2.6159719075939196e-05,
|
409 |
+
"loss": 1.4969,
|
410 |
"step": 335
|
411 |
},
|
412 |
{
|
413 |
+
"epoch": 0.55,
|
414 |
+
"learning_rate": 2.602438544271395e-05,
|
415 |
+
"loss": 1.4983,
|
416 |
"step": 340
|
417 |
},
|
418 |
{
|
419 |
+
"epoch": 0.56,
|
420 |
+
"learning_rate": 2.5887071502963338e-05,
|
421 |
+
"loss": 1.4801,
|
422 |
"step": 345
|
423 |
},
|
424 |
{
|
425 |
+
"epoch": 0.57,
|
426 |
+
"learning_rate": 2.574780192234264e-05,
|
427 |
+
"loss": 1.4595,
|
428 |
"step": 350
|
429 |
},
|
430 |
{
|
431 |
+
"epoch": 0.57,
|
432 |
+
"learning_rate": 2.5606601717798212e-05,
|
433 |
+
"loss": 1.4901,
|
434 |
"step": 355
|
435 |
},
|
436 |
{
|
437 |
+
"epoch": 0.58,
|
438 |
+
"learning_rate": 2.5463496253073726e-05,
|
439 |
+
"loss": 1.4939,
|
440 |
"step": 360
|
441 |
},
|
442 |
{
|
443 |
+
"epoch": 0.59,
|
444 |
+
"learning_rate": 2.531851123415406e-05,
|
445 |
+
"loss": 1.4791,
|
446 |
"step": 365
|
447 |
},
|
448 |
{
|
449 |
+
"epoch": 0.6,
|
450 |
+
"learning_rate": 2.5171672704647785e-05,
|
451 |
+
"loss": 1.448,
|
452 |
"step": 370
|
453 |
},
|
454 |
{
|
455 |
+
"epoch": 0.61,
|
456 |
+
"learning_rate": 2.502300704110891e-05,
|
457 |
+
"loss": 1.4857,
|
458 |
"step": 375
|
459 |
},
|
460 |
{
|
461 |
+
"epoch": 0.61,
|
462 |
+
"learning_rate": 2.4872540948298913e-05,
|
463 |
+
"loss": 1.4829,
|
464 |
"step": 380
|
465 |
},
|
466 |
{
|
467 |
+
"epoch": 0.62,
|
468 |
+
"learning_rate": 2.472030145438974e-05,
|
469 |
+
"loss": 1.4786,
|
470 |
"step": 385
|
471 |
},
|
472 |
{
|
473 |
+
"epoch": 0.63,
|
474 |
+
"learning_rate": 2.4566315906108772e-05,
|
475 |
+
"loss": 1.4118,
|
476 |
"step": 390
|
477 |
},
|
478 |
{
|
479 |
+
"epoch": 0.64,
|
480 |
+
"learning_rate": 2.4410611963826522e-05,
|
481 |
+
"loss": 1.429,
|
482 |
"step": 395
|
483 |
},
|
484 |
{
|
485 |
+
"epoch": 0.65,
|
486 |
+
"learning_rate": 2.4253217596588036e-05,
|
487 |
+
"loss": 1.4719,
|
488 |
"step": 400
|
489 |
},
|
490 |
{
|
491 |
+
"epoch": 0.66,
|
492 |
+
"learning_rate": 2.4094161077088784e-05,
|
493 |
+
"loss": 1.5054,
|
494 |
"step": 405
|
495 |
},
|
496 |
{
|
497 |
+
"epoch": 0.66,
|
498 |
+
"learning_rate": 2.3933470976596088e-05,
|
499 |
+
"loss": 1.4493,
|
500 |
"step": 410
|
501 |
},
|
502 |
{
|
503 |
+
"epoch": 0.67,
|
504 |
+
"learning_rate": 2.3771176159816846e-05,
|
505 |
+
"loss": 1.4957,
|
506 |
"step": 415
|
507 |
},
|
508 |
{
|
509 |
+
"epoch": 0.68,
|
510 |
+
"learning_rate": 2.360730577971259e-05,
|
511 |
+
"loss": 1.4764,
|
512 |
"step": 420
|
513 |
},
|
514 |
{
|
515 |
+
"epoch": 0.69,
|
516 |
+
"learning_rate": 2.3441889272262742e-05,
|
517 |
+
"loss": 1.5056,
|
518 |
"step": 425
|
519 |
},
|
520 |
{
|
521 |
+
"epoch": 0.7,
|
522 |
+
"learning_rate": 2.3274956351177037e-05,
|
523 |
+
"loss": 1.4732,
|
524 |
"step": 430
|
525 |
},
|
526 |
{
|
527 |
+
"epoch": 0.7,
|
528 |
+
"learning_rate": 2.3106537002558074e-05,
|
529 |
+
"loss": 1.5047,
|
530 |
"step": 435
|
531 |
},
|
532 |
{
|
533 |
+
"epoch": 0.71,
|
534 |
+
"learning_rate": 2.293666147951491e-05,
|
535 |
+
"loss": 1.5098,
|
536 |
"step": 440
|
537 |
},
|
538 |
{
|
539 |
+
"epoch": 0.72,
|
540 |
+
"learning_rate": 2.2765360296728697e-05,
|
541 |
+
"loss": 1.504,
|
542 |
"step": 445
|
543 |
},
|
544 |
{
|
545 |
+
"epoch": 0.73,
|
546 |
+
"learning_rate": 2.259266422497137e-05,
|
547 |
+
"loss": 1.4742,
|
548 |
"step": 450
|
549 |
},
|
550 |
{
|
551 |
+
"epoch": 0.74,
|
552 |
+
"learning_rate": 2.2418604285578273e-05,
|
553 |
+
"loss": 1.482,
|
554 |
"step": 455
|
555 |
},
|
556 |
{
|
557 |
+
"epoch": 0.74,
|
558 |
+
"learning_rate": 2.2243211744875818e-05,
|
559 |
+
"loss": 1.4702,
|
560 |
"step": 460
|
561 |
},
|
562 |
{
|
563 |
+
"epoch": 0.75,
|
564 |
+
"learning_rate": 2.2066518108565137e-05,
|
565 |
+
"loss": 1.4839,
|
566 |
"step": 465
|
567 |
},
|
568 |
{
|
569 |
+
"epoch": 0.76,
|
570 |
+
"learning_rate": 2.18885551160627e-05,
|
571 |
+
"loss": 1.4942,
|
572 |
"step": 470
|
573 |
},
|
574 |
{
|
575 |
+
"epoch": 0.77,
|
576 |
+
"learning_rate": 2.1709354734798998e-05,
|
577 |
+
"loss": 1.4497,
|
578 |
"step": 475
|
579 |
},
|
580 |
{
|
581 |
+
"epoch": 0.78,
|
582 |
+
"learning_rate": 2.152894915447624e-05,
|
583 |
+
"loss": 1.4927,
|
584 |
"step": 480
|
585 |
},
|
586 |
{
|
587 |
+
"epoch": 0.78,
|
588 |
+
"learning_rate": 2.134737078128611e-05,
|
589 |
+
"loss": 1.4794,
|
590 |
"step": 485
|
591 |
},
|
592 |
{
|
593 |
+
"epoch": 0.79,
|
594 |
+
"learning_rate": 2.1164652232088674e-05,
|
595 |
+
"loss": 1.5034,
|
596 |
"step": 490
|
597 |
},
|
598 |
{
|
599 |
+
"epoch": 0.8,
|
600 |
+
"learning_rate": 2.0980826328553416e-05,
|
601 |
+
"loss": 1.5157,
|
602 |
"step": 495
|
603 |
},
|
604 |
{
|
605 |
+
"epoch": 0.81,
|
606 |
+
"learning_rate": 2.0795926091263504e-05,
|
607 |
+
"loss": 1.4681,
|
608 |
"step": 500
|
609 |
},
|
610 |
{
|
611 |
+
"epoch": 0.82,
|
612 |
+
"learning_rate": 2.0609984733784287e-05,
|
613 |
+
"loss": 1.4732,
|
614 |
"step": 505
|
615 |
},
|
616 |
{
|
617 |
+
"epoch": 0.83,
|
618 |
+
"learning_rate": 2.042303565669719e-05,
|
619 |
+
"loss": 1.5047,
|
620 |
"step": 510
|
621 |
},
|
622 |
{
|
623 |
+
"epoch": 0.83,
|
624 |
+
"learning_rate": 2.0235112441599948e-05,
|
625 |
+
"loss": 1.5093,
|
626 |
"step": 515
|
627 |
},
|
628 |
{
|
629 |
+
"epoch": 0.84,
|
630 |
+
"learning_rate": 2.0046248845074373e-05,
|
631 |
+
"loss": 1.4515,
|
632 |
"step": 520
|
633 |
},
|
634 |
{
|
635 |
+
"epoch": 0.85,
|
636 |
+
"learning_rate": 1.9856478792622666e-05,
|
637 |
+
"loss": 1.5051,
|
638 |
"step": 525
|
639 |
},
|
640 |
{
|
641 |
+
"epoch": 0.86,
|
642 |
+
"learning_rate": 1.9665836372573397e-05,
|
643 |
+
"loss": 1.5073,
|
644 |
"step": 530
|
645 |
},
|
646 |
{
|
647 |
+
"epoch": 0.87,
|
648 |
+
"learning_rate": 1.947435582995821e-05,
|
649 |
+
"loss": 1.4952,
|
650 |
"step": 535
|
651 |
},
|
652 |
{
|
653 |
+
"epoch": 0.87,
|
654 |
+
"learning_rate": 1.928207156036043e-05,
|
655 |
+
"loss": 1.4308,
|
656 |
"step": 540
|
657 |
},
|
658 |
{
|
659 |
+
"epoch": 0.88,
|
660 |
+
"learning_rate": 1.9089018103736568e-05,
|
661 |
+
"loss": 1.4588,
|
662 |
"step": 545
|
663 |
},
|
664 |
{
|
665 |
+
"epoch": 0.89,
|
666 |
+
"learning_rate": 1.8895230138211942e-05,
|
667 |
+
"loss": 1.5477,
|
668 |
"step": 550
|
669 |
},
|
670 |
{
|
671 |
+
"epoch": 0.9,
|
672 |
+
"learning_rate": 1.870074247385144e-05,
|
673 |
+
"loss": 1.4979,
|
674 |
"step": 555
|
675 |
},
|
676 |
{
|
677 |
+
"epoch": 0.91,
|
678 |
+
"learning_rate": 1.8505590046406615e-05,
|
679 |
+
"loss": 1.4487,
|
680 |
"step": 560
|
681 |
},
|
682 |
{
|
683 |
+
"epoch": 0.91,
|
684 |
+
"learning_rate": 1.8309807911040186e-05,
|
685 |
+
"loss": 1.4671,
|
686 |
"step": 565
|
687 |
},
|
688 |
{
|
689 |
+
"epoch": 0.92,
|
690 |
+
"learning_rate": 1.8113431236029078e-05,
|
691 |
+
"loss": 1.4486,
|
692 |
"step": 570
|
693 |
},
|
694 |
{
|
695 |
+
"epoch": 0.93,
|
696 |
+
"learning_rate": 1.7916495296447162e-05,
|
697 |
+
"loss": 1.483,
|
698 |
"step": 575
|
699 |
},
|
700 |
{
|
701 |
+
"epoch": 0.94,
|
702 |
+
"learning_rate": 1.771903546782883e-05,
|
703 |
+
"loss": 1.4896,
|
704 |
"step": 580
|
705 |
},
|
706 |
{
|
707 |
+
"epoch": 0.95,
|
708 |
+
"learning_rate": 1.7521087219814454e-05,
|
709 |
+
"loss": 1.5259,
|
710 |
"step": 585
|
711 |
},
|
712 |
{
|
713 |
+
"epoch": 0.95,
|
714 |
+
"learning_rate": 1.7322686109779032e-05,
|
715 |
+
"loss": 1.4845,
|
716 |
"step": 590
|
717 |
},
|
718 |
{
|
719 |
+
"epoch": 0.96,
|
720 |
+
"learning_rate": 1.7123867776445e-05,
|
721 |
+
"loss": 1.4866,
|
722 |
"step": 595
|
723 |
},
|
724 |
{
|
725 |
+
"epoch": 0.97,
|
726 |
+
"learning_rate": 1.692466793348047e-05,
|
727 |
+
"loss": 1.4968,
|
728 |
"step": 600
|
729 |
},
|
730 |
{
|
731 |
+
"epoch": 0.98,
|
732 |
+
"learning_rate": 1.6725122363084004e-05,
|
733 |
+
"loss": 1.4582,
|
734 |
"step": 605
|
735 |
},
|
736 |
{
|
737 |
+
"epoch": 0.99,
|
738 |
+
"learning_rate": 1.6525266909557046e-05,
|
739 |
+
"loss": 1.4605,
|
740 |
"step": 610
|
741 |
},
|
742 |
{
|
743 |
+
"epoch": 1.0,
|
744 |
+
"learning_rate": 1.6325137472865262e-05,
|
745 |
+
"loss": 1.4391,
|
746 |
"step": 615
|
747 |
},
|
748 |
{
|
749 |
+
"epoch": 1.0,
|
750 |
+
"eval_loss": 3.6259799003601074,
|
751 |
+
"eval_runtime": 967.3528,
|
752 |
+
"eval_samples_per_second": 8.998,
|
753 |
+
"eval_steps_per_second": 2.249,
|
754 |
+
"step": 617
|
755 |
+
},
|
756 |
+
{
|
757 |
+
"epoch": 1.0,
|
758 |
+
"learning_rate": 1.6124770002189804e-05,
|
759 |
+
"loss": 1.5992,
|
760 |
"step": 620
|
761 |
},
|
762 |
{
|
763 |
+
"epoch": 1.01,
|
764 |
+
"learning_rate": 1.5924200489469782e-05,
|
765 |
+
"loss": 1.1688,
|
766 |
"step": 625
|
767 |
},
|
768 |
{
|
769 |
+
"epoch": 1.02,
|
770 |
+
"learning_rate": 1.572346496293706e-05,
|
771 |
+
"loss": 1.1778,
|
772 |
"step": 630
|
773 |
},
|
774 |
{
|
775 |
+
"epoch": 1.03,
|
776 |
+
"learning_rate": 1.5522599480644496e-05,
|
777 |
+
"loss": 1.1652,
|
778 |
"step": 635
|
779 |
},
|
780 |
{
|
781 |
+
"epoch": 1.04,
|
782 |
+
"learning_rate": 1.532164012398886e-05,
|
783 |
+
"loss": 1.1344,
|
784 |
"step": 640
|
785 |
},
|
786 |
{
|
787 |
+
"epoch": 1.05,
|
788 |
+
"learning_rate": 1.5120622991229545e-05,
|
789 |
+
"loss": 1.1474,
|
790 |
"step": 645
|
791 |
},
|
792 |
{
|
793 |
+
"epoch": 1.05,
|
794 |
+
"learning_rate": 1.4919584191004244e-05,
|
795 |
+
"loss": 1.1457,
|
796 |
"step": 650
|
797 |
},
|
798 |
{
|
799 |
+
"epoch": 1.06,
|
800 |
+
"learning_rate": 1.471855983584276e-05,
|
801 |
+
"loss": 1.1441,
|
802 |
"step": 655
|
803 |
},
|
804 |
{
|
805 |
+
"epoch": 1.07,
|
806 |
+
"learning_rate": 1.4517586035680145e-05,
|
807 |
+
"loss": 1.1546,
|
808 |
"step": 660
|
809 |
},
|
810 |
{
|
811 |
+
"epoch": 1.08,
|
812 |
+
"learning_rate": 1.431669889137027e-05,
|
813 |
+
"loss": 1.1526,
|
814 |
"step": 665
|
815 |
},
|
816 |
{
|
817 |
+
"epoch": 1.09,
|
818 |
+
"learning_rate": 1.4115934488201047e-05,
|
819 |
+
"loss": 1.1778,
|
820 |
"step": 670
|
821 |
},
|
822 |
{
|
823 |
+
"epoch": 1.09,
|
824 |
+
"learning_rate": 1.3915328889412434e-05,
|
825 |
+
"loss": 1.1468,
|
826 |
"step": 675
|
827 |
},
|
828 |
{
|
829 |
+
"epoch": 1.1,
|
830 |
+
"learning_rate": 1.3714918129718418e-05,
|
831 |
+
"loss": 1.1367,
|
832 |
"step": 680
|
833 |
},
|
834 |
{
|
835 |
+
"epoch": 1.11,
|
836 |
+
"learning_rate": 1.3514738208834112e-05,
|
837 |
+
"loss": 1.1972,
|
838 |
"step": 685
|
839 |
},
|
840 |
{
|
841 |
+
"epoch": 1.12,
|
842 |
+
"learning_rate": 1.331482508500912e-05,
|
843 |
+
"loss": 1.1701,
|
844 |
"step": 690
|
845 |
},
|
846 |
{
|
847 |
+
"epoch": 1.13,
|
848 |
+
"learning_rate": 1.31152146685684e-05,
|
849 |
+
"loss": 1.1911,
|
850 |
"step": 695
|
851 |
},
|
852 |
{
|
853 |
+
"epoch": 1.13,
|
854 |
+
"learning_rate": 1.2915942815461677e-05,
|
855 |
+
"loss": 1.1758,
|
856 |
"step": 700
|
857 |
},
|
858 |
{
|
859 |
+
"epoch": 1.14,
|
860 |
+
"learning_rate": 1.2717045320822658e-05,
|
861 |
+
"loss": 1.1486,
|
862 |
"step": 705
|
863 |
},
|
864 |
{
|
865 |
+
"epoch": 1.15,
|
866 |
+
"learning_rate": 1.2518557912539185e-05,
|
867 |
+
"loss": 1.1502,
|
868 |
"step": 710
|
869 |
},
|
870 |
{
|
871 |
+
"epoch": 1.16,
|
872 |
+
"learning_rate": 1.232051624483541e-05,
|
873 |
+
"loss": 1.1459,
|
874 |
"step": 715
|
875 |
},
|
876 |
{
|
877 |
+
"epoch": 1.17,
|
878 |
+
"learning_rate": 1.2122955891867278e-05,
|
879 |
+
"loss": 1.1546,
|
880 |
"step": 720
|
881 |
},
|
882 |
{
|
883 |
+
"epoch": 1.17,
|
884 |
+
"learning_rate": 1.1925912341332324e-05,
|
885 |
+
"loss": 1.165,
|
886 |
"step": 725
|
887 |
},
|
888 |
{
|
889 |
+
"epoch": 1.18,
|
890 |
+
"learning_rate": 1.1729420988095042e-05,
|
891 |
+
"loss": 1.1548,
|
892 |
"step": 730
|
893 |
},
|
894 |
{
|
895 |
+
"epoch": 1.19,
|
896 |
+
"learning_rate": 1.1533517127828926e-05,
|
897 |
+
"loss": 1.1454,
|
898 |
"step": 735
|
899 |
},
|
900 |
{
|
901 |
+
"epoch": 1.2,
|
902 |
+
"learning_rate": 1.1338235950676305e-05,
|
903 |
+
"loss": 1.19,
|
904 |
"step": 740
|
905 |
},
|
906 |
{
|
907 |
+
"epoch": 1.21,
|
908 |
+
"learning_rate": 1.1143612534927153e-05,
|
909 |
+
"loss": 1.1475,
|
910 |
"step": 745
|
911 |
},
|
912 |
{
|
913 |
+
"epoch": 1.22,
|
914 |
+
"learning_rate": 1.0949681840717997e-05,
|
915 |
+
"loss": 1.1754,
|
916 |
"step": 750
|
917 |
},
|
918 |
{
|
919 |
+
"epoch": 1.22,
|
920 |
+
"learning_rate": 1.0756478703752036e-05,
|
921 |
+
"loss": 1.1041,
|
922 |
"step": 755
|
923 |
},
|
924 |
{
|
925 |
+
"epoch": 1.23,
|
926 |
+
"learning_rate": 1.0564037829041609e-05,
|
927 |
+
"loss": 1.1465,
|
928 |
"step": 760
|
929 |
},
|
930 |
{
|
931 |
+
"epoch": 1.24,
|
932 |
+
"learning_rate": 1.037239378467416e-05,
|
933 |
+
"loss": 1.1704,
|
934 |
"step": 765
|
935 |
},
|
936 |
{
|
937 |
+
"epoch": 1.25,
|
938 |
+
"learning_rate": 1.0181580995602766e-05,
|
939 |
+
"loss": 1.1716,
|
940 |
"step": 770
|
941 |
},
|
942 |
{
|
943 |
+
"epoch": 1.26,
|
944 |
+
"learning_rate": 9.991633737462405e-06,
|
945 |
+
"loss": 1.1902,
|
946 |
"step": 775
|
947 |
},
|
948 |
{
|
949 |
+
"epoch": 1.26,
|
950 |
+
"learning_rate": 9.802586130413045e-06,
|
951 |
+
"loss": 1.1342,
|
952 |
"step": 780
|
953 |
},
|
954 |
{
|
955 |
+
"epoch": 1.27,
|
956 |
+
"learning_rate": 9.614472133010623e-06,
|
957 |
+
"loss": 1.1848,
|
958 |
"step": 785
|
959 |
},
|
960 |
{
|
961 |
+
"epoch": 1.28,
|
962 |
+
"learning_rate": 9.42732553610712e-06,
|
963 |
+
"loss": 1.1652,
|
964 |
"step": 790
|
965 |
},
|
966 |
{
|
967 |
+
"epoch": 1.29,
|
968 |
+
"learning_rate": 9.241179956780689e-06,
|
969 |
+
"loss": 1.1832,
|
970 |
"step": 795
|
971 |
},
|
972 |
{
|
973 |
+
"epoch": 1.3,
|
974 |
+
"learning_rate": 9.056068832297041e-06,
|
975 |
+
"loss": 1.1935,
|
976 |
"step": 800
|
977 |
},
|
978 |
{
|
979 |
+
"epoch": 1.3,
|
980 |
+
"learning_rate": 8.872025414103135e-06,
|
981 |
+
"loss": 1.1714,
|
982 |
"step": 805
|
983 |
},
|
984 |
{
|
985 |
+
"epoch": 1.31,
|
986 |
+
"learning_rate": 8.689082761854213e-06,
|
987 |
+
"loss": 1.1525,
|
988 |
"step": 810
|
989 |
},
|
990 |
{
|
991 |
+
"epoch": 1.32,
|
992 |
+
"learning_rate": 8.507273737475307e-06,
|
993 |
+
"loss": 1.1701,
|
994 |
"step": 815
|
995 |
},
|
996 |
{
|
997 |
+
"epoch": 1.33,
|
998 |
+
"learning_rate": 8.326630999258286e-06,
|
999 |
+
"loss": 1.1761,
|
1000 |
"step": 820
|
1001 |
},
|
1002 |
{
|
1003 |
+
"epoch": 1.34,
|
1004 |
+
"learning_rate": 8.14718699599542e-06,
|
1005 |
+
"loss": 1.1323,
|
1006 |
"step": 825
|
1007 |
},
|
1008 |
{
|
1009 |
+
"epoch": 1.34,
|
1010 |
+
"learning_rate": 7.968973961150653e-06,
|
1011 |
+
"loss": 1.1768,
|
1012 |
"step": 830
|
1013 |
},
|
1014 |
{
|
1015 |
+
"epoch": 1.35,
|
1016 |
+
"learning_rate": 7.792023907069486e-06,
|
1017 |
+
"loss": 1.1487,
|
1018 |
"step": 835
|
1019 |
},
|
1020 |
{
|
1021 |
+
"epoch": 1.36,
|
1022 |
+
"learning_rate": 7.616368619228645e-06,
|
1023 |
+
"loss": 1.1445,
|
1024 |
"step": 840
|
1025 |
},
|
1026 |
{
|
1027 |
+
"epoch": 1.37,
|
1028 |
+
"learning_rate": 7.442039650526419e-06,
|
1029 |
+
"loss": 1.1348,
|
1030 |
"step": 845
|
1031 |
},
|
1032 |
{
|
1033 |
+
"epoch": 1.38,
|
1034 |
+
"learning_rate": 7.2690683156148705e-06,
|
1035 |
+
"loss": 1.1478,
|
1036 |
"step": 850
|
1037 |
},
|
1038 |
{
|
1039 |
+
"epoch": 1.39,
|
1040 |
+
"learning_rate": 7.097485685274776e-06,
|
1041 |
+
"loss": 1.1625,
|
1042 |
"step": 855
|
1043 |
},
|
1044 |
{
|
1045 |
+
"epoch": 1.39,
|
1046 |
+
"learning_rate": 6.927322580834376e-06,
|
1047 |
+
"loss": 1.1331,
|
1048 |
"step": 860
|
1049 |
},
|
1050 |
{
|
1051 |
+
"epoch": 1.4,
|
1052 |
+
"learning_rate": 6.758609568632982e-06,
|
1053 |
+
"loss": 1.1608,
|
1054 |
"step": 865
|
1055 |
},
|
1056 |
{
|
1057 |
+
"epoch": 1.41,
|
1058 |
+
"learning_rate": 6.591376954530345e-06,
|
1059 |
+
"loss": 1.1349,
|
1060 |
"step": 870
|
1061 |
},
|
1062 |
{
|
1063 |
+
"epoch": 1.42,
|
1064 |
+
"learning_rate": 6.4256547784628e-06,
|
1065 |
+
"loss": 1.1569,
|
1066 |
"step": 875
|
1067 |
},
|
1068 |
{
|
1069 |
+
"epoch": 1.43,
|
1070 |
+
"learning_rate": 6.261472809047244e-06,
|
1071 |
+
"loss": 1.152,
|
1072 |
"step": 880
|
1073 |
},
|
1074 |
{
|
1075 |
+
"epoch": 1.43,
|
1076 |
+
"learning_rate": 6.098860538233769e-06,
|
1077 |
+
"loss": 1.1498,
|
1078 |
"step": 885
|
1079 |
},
|
1080 |
{
|
1081 |
+
"epoch": 1.44,
|
1082 |
+
"learning_rate": 5.937847176008072e-06,
|
1083 |
+
"loss": 1.1802,
|
1084 |
"step": 890
|
1085 |
},
|
1086 |
{
|
1087 |
+
"epoch": 1.45,
|
1088 |
+
"learning_rate": 5.778461645144438e-06,
|
1089 |
+
"loss": 1.1717,
|
1090 |
"step": 895
|
1091 |
},
|
1092 |
{
|
1093 |
+
"epoch": 1.46,
|
1094 |
+
"learning_rate": 5.6207325760103845e-06,
|
1095 |
+
"loss": 1.152,
|
1096 |
"step": 900
|
1097 |
},
|
1098 |
{
|
1099 |
+
"epoch": 1.47,
|
1100 |
+
"learning_rate": 5.464688301423782e-06,
|
1101 |
+
"loss": 1.1414,
|
1102 |
"step": 905
|
1103 |
},
|
1104 |
{
|
1105 |
+
"epoch": 1.47,
|
1106 |
+
"learning_rate": 5.310356851563427e-06,
|
1107 |
+
"loss": 1.154,
|
1108 |
"step": 910
|
1109 |
},
|
1110 |
{
|
1111 |
+
"epoch": 1.48,
|
1112 |
+
"learning_rate": 5.1577659489340255e-06,
|
1113 |
+
"loss": 1.1353,
|
1114 |
"step": 915
|
1115 |
},
|
1116 |
{
|
1117 |
+
"epoch": 1.49,
|
1118 |
+
"learning_rate": 5.00694300338638e-06,
|
1119 |
+
"loss": 1.1783,
|
1120 |
"step": 920
|
1121 |
},
|
1122 |
{
|
1123 |
+
"epoch": 1.5,
|
1124 |
+
"learning_rate": 4.857915107193783e-06,
|
1125 |
+
"loss": 1.14,
|
1126 |
"step": 925
|
1127 |
},
|
1128 |
{
|
1129 |
+
"epoch": 1.51,
|
1130 |
+
"learning_rate": 4.710709030185422e-06,
|
1131 |
+
"loss": 1.1755,
|
1132 |
"step": 930
|
1133 |
},
|
1134 |
{
|
1135 |
+
"epoch": 1.51,
|
1136 |
+
"learning_rate": 4.565351214937748e-06,
|
1137 |
+
"loss": 1.1121,
|
1138 |
"step": 935
|
1139 |
},
|
1140 |
{
|
1141 |
+
"epoch": 1.52,
|
1142 |
+
"learning_rate": 4.421867772024601e-06,
|
1143 |
+
"loss": 1.171,
|
1144 |
"step": 940
|
1145 |
},
|
1146 |
{
|
1147 |
+
"epoch": 1.53,
|
1148 |
+
"learning_rate": 4.280284475326948e-06,
|
1149 |
+
"loss": 1.1517,
|
1150 |
"step": 945
|
1151 |
},
|
1152 |
{
|
1153 |
+
"epoch": 1.54,
|
1154 |
+
"learning_rate": 4.140626757403176e-06,
|
1155 |
+
"loss": 1.1694,
|
1156 |
"step": 950
|
1157 |
},
|
1158 |
{
|
1159 |
+
"epoch": 1.55,
|
1160 |
+
"learning_rate": 4.002919704920607e-06,
|
1161 |
+
"loss": 1.1465,
|
1162 |
"step": 955
|
1163 |
},
|
1164 |
{
|
1165 |
+
"epoch": 1.56,
|
1166 |
+
"learning_rate": 3.8671880541492236e-06,
|
1167 |
+
"loss": 1.179,
|
1168 |
"step": 960
|
1169 |
},
|
1170 |
{
|
1171 |
+
"epoch": 1.56,
|
1172 |
+
"learning_rate": 3.7334561865182694e-06,
|
1173 |
+
"loss": 1.1498,
|
1174 |
"step": 965
|
1175 |
},
|
1176 |
{
|
1177 |
+
"epoch": 1.57,
|
1178 |
+
"learning_rate": 3.6017481242366503e-06,
|
1179 |
+
"loss": 1.1438,
|
1180 |
"step": 970
|
1181 |
},
|
1182 |
{
|
1183 |
+
"epoch": 1.58,
|
1184 |
+
"learning_rate": 3.472087525977823e-06,
|
1185 |
+
"loss": 1.1647,
|
1186 |
"step": 975
|
1187 |
},
|
1188 |
{
|
1189 |
+
"epoch": 1.59,
|
1190 |
+
"learning_rate": 3.3444976826299754e-06,
|
1191 |
+
"loss": 1.1475,
|
1192 |
"step": 980
|
1193 |
},
|
1194 |
{
|
1195 |
+
"epoch": 1.6,
|
1196 |
+
"learning_rate": 3.219001513112329e-06,
|
1197 |
+
"loss": 1.15,
|
1198 |
"step": 985
|
1199 |
},
|
1200 |
{
|
1201 |
+
"epoch": 1.6,
|
1202 |
+
"learning_rate": 3.0956215602581933e-06,
|
1203 |
+
"loss": 1.1613,
|
1204 |
"step": 990
|
1205 |
},
|
1206 |
{
|
1207 |
+
"epoch": 1.61,
|
1208 |
+
"learning_rate": 2.974379986765622e-06,
|
1209 |
+
"loss": 1.1672,
|
1210 |
"step": 995
|
1211 |
},
|
1212 |
{
|
1213 |
+
"epoch": 1.62,
|
1214 |
+
"learning_rate": 2.855298571216316e-06,
|
1215 |
+
"loss": 1.1702,
|
1216 |
"step": 1000
|
1217 |
},
|
1218 |
{
|
1219 |
+
"epoch": 1.63,
|
1220 |
+
"learning_rate": 2.738398704163561e-06,
|
1221 |
+
"loss": 1.1634,
|
1222 |
"step": 1005
|
1223 |
},
|
1224 |
{
|
1225 |
+
"epoch": 1.64,
|
1226 |
+
"learning_rate": 2.6237013842898533e-06,
|
1227 |
+
"loss": 1.1756,
|
1228 |
"step": 1010
|
1229 |
},
|
1230 |
{
|
1231 |
+
"epoch": 1.64,
|
1232 |
+
"learning_rate": 2.511227214634887e-06,
|
1233 |
+
"loss": 1.1075,
|
1234 |
"step": 1015
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1235 |
}
|
1236 |
],
|
1237 |
+
"max_steps": 1234,
|
1238 |
"num_train_epochs": 2,
|
1239 |
+
"total_flos": 4.827958566725878e+17,
|
1240 |
"trial_name": null,
|
1241 |
"trial_params": null
|
1242 |
}
|
training_args.bin
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
-
size
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:41b3235764a2f119a2586ca316b39c8e5986e318c5b3665781ee71759bb93cc0
|
3 |
+
size 4271
|
vocab.json
CHANGED
The diff for this file is too large to render.
See raw diff
|
|
zero_to_fp32.py
ADDED
@@ -0,0 +1,484 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
#!/usr/bin/env python
|
2 |
+
|
3 |
+
# This script extracts fp32 consolidated weights from a zero 2 and 3 DeepSpeed checkpoints. It gets
|
4 |
+
# copied into the top level checkpoint dir, so the user can easily do the conversion at any point in
|
5 |
+
# the future. Once extracted, the weights don't require DeepSpeed and can be used in any
|
6 |
+
# application.
|
7 |
+
#
|
8 |
+
# example: python zero_to_fp32.py . pytorch_model.bin
|
9 |
+
|
10 |
+
import argparse
|
11 |
+
import torch
|
12 |
+
import glob
|
13 |
+
import math
|
14 |
+
import os
|
15 |
+
import re
|
16 |
+
from collections import OrderedDict
|
17 |
+
|
18 |
+
# while this script doesn't use deepspeed to recover data, since the checkpoints are pickled with
|
19 |
+
# DeepSpeed data structures it has to be available in the current python environment.
|
20 |
+
import deepspeed
|
21 |
+
from deepspeed.utils import logger
|
22 |
+
from deepspeed.checkpoint.constants import (DS_VERSION,
|
23 |
+
OPTIMIZER_STATE_DICT,
|
24 |
+
PARAM_SHAPES,
|
25 |
+
SINGLE_PARTITION_OF_FP32_GROUPS,
|
26 |
+
FP32_FLAT_GROUPS,
|
27 |
+
ZERO_STAGE,
|
28 |
+
PARTITION_COUNT,
|
29 |
+
PARAM_SHAPES,
|
30 |
+
BUFFER_NAMES)
|
31 |
+
|
32 |
+
debug = 0
|
33 |
+
|
34 |
+
# load to cpu
|
35 |
+
device = torch.device('cpu')
|
36 |
+
|
37 |
+
|
38 |
+
def atoi(text):
|
39 |
+
return int(text) if text.isdigit() else text
|
40 |
+
|
41 |
+
|
42 |
+
def natural_keys(text):
|
43 |
+
'''
|
44 |
+
alist.sort(key=natural_keys) sorts in human order
|
45 |
+
http://nedbatchelder.com/blog/200712/human_sorting.html
|
46 |
+
(See Toothy's implementation in the comments)
|
47 |
+
'''
|
48 |
+
return [atoi(c) for c in re.split(r'(\d+)', text)]
|
49 |
+
|
50 |
+
|
51 |
+
def get_model_state_file(checkpoint_dir, zero_stage):
|
52 |
+
if not os.path.isdir(checkpoint_dir):
|
53 |
+
raise FileNotFoundError(f"Directory '{checkpoint_dir}' doesn't exist")
|
54 |
+
|
55 |
+
# there should be only one file
|
56 |
+
if zero_stage == 2:
|
57 |
+
file = os.path.join(checkpoint_dir, "mp_rank_00_model_states.pt")
|
58 |
+
elif zero_stage == 3:
|
59 |
+
file = os.path.join(checkpoint_dir, "zero_pp_rank_0_mp_rank_00_model_states.pt")
|
60 |
+
|
61 |
+
if not os.path.exists(file):
|
62 |
+
raise FileNotFoundError(f"can't find model states file at '{file}'")
|
63 |
+
|
64 |
+
return file
|
65 |
+
|
66 |
+
|
67 |
+
def get_optim_files(checkpoint_dir):
|
68 |
+
# XXX: need to test that this simple glob rule works for multi-node setup too
|
69 |
+
optim_files = sorted(glob.glob(os.path.join(checkpoint_dir,
|
70 |
+
"*_optim_states.pt")),
|
71 |
+
key=natural_keys)
|
72 |
+
|
73 |
+
if len(optim_files) == 0:
|
74 |
+
raise FileNotFoundError(
|
75 |
+
f"can't find '*_optim_states.pt' files in directory '{checkpoint_dir}'")
|
76 |
+
|
77 |
+
return optim_files
|
78 |
+
|
79 |
+
|
80 |
+
def parse_model_state(file):
|
81 |
+
state_dict = torch.load(file, map_location=device)
|
82 |
+
|
83 |
+
if BUFFER_NAMES not in state_dict:
|
84 |
+
raise ValueError(f"{file} is not a model state checkpoint")
|
85 |
+
buffer_names = state_dict[BUFFER_NAMES]
|
86 |
+
if debug:
|
87 |
+
print("Found buffers:", buffer_names)
|
88 |
+
|
89 |
+
# recover just the buffers while restoring them to fp32 if they were saved in fp16
|
90 |
+
buffers = {
|
91 |
+
k: v.float()
|
92 |
+
for k,
|
93 |
+
v in state_dict["module"].items() if k in buffer_names
|
94 |
+
}
|
95 |
+
param_shapes = state_dict[PARAM_SHAPES]
|
96 |
+
|
97 |
+
ds_version = state_dict.get(DS_VERSION, None)
|
98 |
+
|
99 |
+
return buffers, param_shapes, ds_version
|
100 |
+
|
101 |
+
|
102 |
+
def parse_optim_states(files, ds_checkpoint_dir):
|
103 |
+
|
104 |
+
total_files = len(files)
|
105 |
+
state_dicts = []
|
106 |
+
for f in files:
|
107 |
+
state_dicts.append(torch.load(f, map_location=device))
|
108 |
+
|
109 |
+
if not ZERO_STAGE in state_dicts[0][OPTIMIZER_STATE_DICT]:
|
110 |
+
raise ValueError(f"{files[0]} is not a zero checkpoint")
|
111 |
+
zero_stage = state_dicts[0][OPTIMIZER_STATE_DICT][ZERO_STAGE]
|
112 |
+
world_size = state_dicts[0][OPTIMIZER_STATE_DICT][PARTITION_COUNT]
|
113 |
+
|
114 |
+
# For ZeRO-2 each param group can have different partition_count as data parallelism for expert
|
115 |
+
# parameters can be different from data parallelism for non-expert parameters. So we can just
|
116 |
+
# use the max of the partition_count to get the dp world_size.
|
117 |
+
|
118 |
+
if type(world_size) is list:
|
119 |
+
world_size = max(world_size)
|
120 |
+
|
121 |
+
if world_size != total_files:
|
122 |
+
raise ValueError(
|
123 |
+
f"Expected {world_size} of '*_optim_states.pt' under '{ds_checkpoint_dir}' but found {total_files} files. "
|
124 |
+
"Possibly due to an overwrite of an old checkpoint, or a checkpoint didn't get saved by one or more processes."
|
125 |
+
)
|
126 |
+
|
127 |
+
# the groups are named differently in each stage
|
128 |
+
if zero_stage == 2:
|
129 |
+
fp32_groups_key = SINGLE_PARTITION_OF_FP32_GROUPS
|
130 |
+
elif zero_stage == 3:
|
131 |
+
fp32_groups_key = FP32_FLAT_GROUPS
|
132 |
+
else:
|
133 |
+
raise ValueError(f"unknown zero stage {zero_stage}")
|
134 |
+
|
135 |
+
if zero_stage == 2:
|
136 |
+
fp32_flat_groups = [
|
137 |
+
state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key]
|
138 |
+
for i in range(len(state_dicts))
|
139 |
+
]
|
140 |
+
elif zero_stage == 3:
|
141 |
+
# if there is more than one param group, there will be multiple flattened tensors - one
|
142 |
+
# flattened tensor per group - for simplicity merge them into a single tensor
|
143 |
+
#
|
144 |
+
# XXX: could make the script more memory efficient for when there are multiple groups - it
|
145 |
+
# will require matching the sub-lists of param_shapes for each param group flattened tensor
|
146 |
+
|
147 |
+
fp32_flat_groups = [
|
148 |
+
torch.cat(state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key],
|
149 |
+
0) for i in range(len(state_dicts))
|
150 |
+
]
|
151 |
+
|
152 |
+
return zero_stage, world_size, fp32_flat_groups
|
153 |
+
|
154 |
+
|
155 |
+
def _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir):
|
156 |
+
"""
|
157 |
+
Returns fp32 state_dict reconstructed from ds checkpoint
|
158 |
+
|
159 |
+
Args:
|
160 |
+
- ``ds_checkpoint_dir``: path to the deepspeed checkpoint folder (where the optimizer files are)
|
161 |
+
|
162 |
+
"""
|
163 |
+
print(f"Processing zero checkpoint '{ds_checkpoint_dir}'")
|
164 |
+
|
165 |
+
optim_files = get_optim_files(ds_checkpoint_dir)
|
166 |
+
zero_stage, world_size, fp32_flat_groups = parse_optim_states(optim_files, ds_checkpoint_dir)
|
167 |
+
print(
|
168 |
+
f"Detected checkpoint of type zero stage {zero_stage}, world_size: {world_size}")
|
169 |
+
|
170 |
+
model_file = get_model_state_file(ds_checkpoint_dir, zero_stage)
|
171 |
+
buffers, param_shapes, ds_version = parse_model_state(model_file)
|
172 |
+
print(f'Parsing checkpoint created by deepspeed=={ds_version}')
|
173 |
+
|
174 |
+
if zero_stage == 2:
|
175 |
+
return _get_fp32_state_dict_from_zero2_checkpoint(world_size,
|
176 |
+
param_shapes,
|
177 |
+
fp32_flat_groups,
|
178 |
+
buffers)
|
179 |
+
elif zero_stage == 3:
|
180 |
+
return _get_fp32_state_dict_from_zero3_checkpoint(world_size,
|
181 |
+
param_shapes,
|
182 |
+
fp32_flat_groups,
|
183 |
+
buffers)
|
184 |
+
|
185 |
+
|
186 |
+
def _get_fp32_state_dict_from_zero2_checkpoint(world_size,
|
187 |
+
param_shapes,
|
188 |
+
fp32_flat_groups,
|
189 |
+
buffers):
|
190 |
+
|
191 |
+
# Reconstruction protocol:
|
192 |
+
#
|
193 |
+
# XXX: document this
|
194 |
+
|
195 |
+
if debug:
|
196 |
+
for i in range(world_size):
|
197 |
+
for j in range(len(fp32_flat_groups[0])):
|
198 |
+
print(
|
199 |
+
f"{FP32_FLAT_GROUPS}[{i}][{j}].shape={fp32_flat_groups[i][j].shape}")
|
200 |
+
|
201 |
+
# XXX: memory usage doubles here (zero2)
|
202 |
+
num_param_groups = len(fp32_flat_groups[0])
|
203 |
+
merged_single_partition_of_fp32_groups = []
|
204 |
+
for i in range(num_param_groups):
|
205 |
+
merged_partitions = [sd[i] for sd in fp32_flat_groups]
|
206 |
+
full_single_fp32_vector = torch.cat(merged_partitions, 0)
|
207 |
+
merged_single_partition_of_fp32_groups.append(full_single_fp32_vector)
|
208 |
+
avail_numel = sum([
|
209 |
+
full_single_fp32_vector.numel()
|
210 |
+
for full_single_fp32_vector in merged_single_partition_of_fp32_groups
|
211 |
+
])
|
212 |
+
|
213 |
+
if debug:
|
214 |
+
wanted_params = sum([len(shapes) for shapes in param_shapes])
|
215 |
+
wanted_numel = sum(
|
216 |
+
[sum(shape.numel() for shape in shapes.values()) for shapes in param_shapes])
|
217 |
+
# not asserting if there is a mismatch due to possible padding
|
218 |
+
print(f"Have {avail_numel} numels to process.")
|
219 |
+
print(f"Need {wanted_numel} numels in {wanted_params} params.")
|
220 |
+
|
221 |
+
state_dict = OrderedDict()
|
222 |
+
|
223 |
+
# buffers
|
224 |
+
state_dict.update(buffers)
|
225 |
+
if debug:
|
226 |
+
print(f"added {len(buffers)} buffers")
|
227 |
+
|
228 |
+
# params
|
229 |
+
# XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
|
230 |
+
# out-of-core computing solution
|
231 |
+
total_numel = 0
|
232 |
+
total_params = 0
|
233 |
+
for shapes, full_single_fp32_vector in zip(param_shapes, merged_single_partition_of_fp32_groups):
|
234 |
+
offset = 0
|
235 |
+
avail_numel = full_single_fp32_vector.numel()
|
236 |
+
for name, shape in shapes.items():
|
237 |
+
|
238 |
+
unpartitioned_numel = shape.numel()
|
239 |
+
total_numel += unpartitioned_numel
|
240 |
+
total_params += 1
|
241 |
+
|
242 |
+
if debug:
|
243 |
+
print(
|
244 |
+
f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} "
|
245 |
+
)
|
246 |
+
state_dict[name] = full_single_fp32_vector.narrow(
|
247 |
+
0,
|
248 |
+
offset,
|
249 |
+
unpartitioned_numel).view(shape)
|
250 |
+
offset += unpartitioned_numel
|
251 |
+
|
252 |
+
# Z2 started to align to 2*world_size to improve nccl performance. Therefore both offset and
|
253 |
+
# avail_numel can differ by anywhere between 0..2*world_size. Due to two unrelated complex
|
254 |
+
# paddings performed in the code it's almost impossible to predict the exact numbers w/o the
|
255 |
+
# live optimizer object, so we are checking that the numbers are within the right range
|
256 |
+
align_to = 2 * world_size
|
257 |
+
|
258 |
+
def zero2_align(x):
|
259 |
+
return align_to * math.ceil(x / align_to)
|
260 |
+
|
261 |
+
if debug:
|
262 |
+
print(f"original offset={offset}, avail_numel={avail_numel}")
|
263 |
+
|
264 |
+
offset = zero2_align(offset)
|
265 |
+
avail_numel = zero2_align(avail_numel)
|
266 |
+
|
267 |
+
if debug:
|
268 |
+
print(f"aligned offset={offset}, avail_numel={avail_numel}")
|
269 |
+
|
270 |
+
# Sanity check
|
271 |
+
if offset != avail_numel:
|
272 |
+
raise ValueError(
|
273 |
+
f"consumed {offset} numels out of {avail_numel} - something is wrong")
|
274 |
+
|
275 |
+
print(
|
276 |
+
f"Reconstructed fp32 state dict with {total_params} params {total_numel} elements"
|
277 |
+
)
|
278 |
+
|
279 |
+
return state_dict
|
280 |
+
|
281 |
+
|
282 |
+
def zero3_partitioned_param_info(unpartitioned_numel, world_size):
|
283 |
+
remainder = unpartitioned_numel % world_size
|
284 |
+
padding_numel = (world_size - remainder) if remainder else 0
|
285 |
+
partitioned_numel = math.ceil(unpartitioned_numel / world_size)
|
286 |
+
return partitioned_numel, padding_numel
|
287 |
+
|
288 |
+
|
289 |
+
def _get_fp32_state_dict_from_zero3_checkpoint(world_size,
|
290 |
+
param_shapes,
|
291 |
+
fp32_flat_groups,
|
292 |
+
buffers):
|
293 |
+
|
294 |
+
# Reconstruction protocol: For zero3 we need to zip the partitions together at boundary of each
|
295 |
+
# param, re-consolidating each param, while dealing with padding if any
|
296 |
+
|
297 |
+
avail_numel = fp32_flat_groups[0].numel() * world_size
|
298 |
+
# merge list of dicts, preserving order
|
299 |
+
param_shapes = {k: v for d in param_shapes for k, v in d.items()}
|
300 |
+
|
301 |
+
if debug:
|
302 |
+
for i in range(world_size):
|
303 |
+
print(f"{FP32_FLAT_GROUPS}[{i}].shape={fp32_flat_groups[i].shape}")
|
304 |
+
|
305 |
+
wanted_params = len(param_shapes)
|
306 |
+
wanted_numel = sum(shape.numel() for shape in param_shapes.values())
|
307 |
+
# not asserting if there is a mismatch due to possible padding
|
308 |
+
print(f"Have {avail_numel} numels to process.")
|
309 |
+
print(f"Need {wanted_numel} numels in {wanted_params} params.")
|
310 |
+
|
311 |
+
state_dict = OrderedDict()
|
312 |
+
|
313 |
+
# buffers
|
314 |
+
state_dict.update(buffers)
|
315 |
+
if debug:
|
316 |
+
print(f"added {len(buffers)} buffers")
|
317 |
+
|
318 |
+
# params
|
319 |
+
# XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
|
320 |
+
# out-of-core computing solution
|
321 |
+
offset = 0
|
322 |
+
total_numel = 0
|
323 |
+
total_params = 0
|
324 |
+
for name, shape in param_shapes.items():
|
325 |
+
|
326 |
+
unpartitioned_numel = shape.numel()
|
327 |
+
total_numel += unpartitioned_numel
|
328 |
+
total_params += 1
|
329 |
+
|
330 |
+
partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
|
331 |
+
|
332 |
+
if debug:
|
333 |
+
print(
|
334 |
+
f"{total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
|
335 |
+
)
|
336 |
+
|
337 |
+
# XXX: memory usage doubles here
|
338 |
+
state_dict[name] = torch.cat(
|
339 |
+
tuple(fp32_flat_groups[i].narrow(0,
|
340 |
+
offset,
|
341 |
+
partitioned_numel)
|
342 |
+
for i in range(world_size)),
|
343 |
+
0).narrow(0,
|
344 |
+
0,
|
345 |
+
unpartitioned_numel).view(shape)
|
346 |
+
offset += partitioned_numel
|
347 |
+
|
348 |
+
offset *= world_size
|
349 |
+
|
350 |
+
# Sanity check
|
351 |
+
if offset != avail_numel:
|
352 |
+
raise ValueError(
|
353 |
+
f"consumed {offset} numels out of {avail_numel} - something is wrong")
|
354 |
+
|
355 |
+
print(
|
356 |
+
f"Reconstructed fp32 state dict with {total_params} params {total_numel} elements"
|
357 |
+
)
|
358 |
+
|
359 |
+
return state_dict
|
360 |
+
|
361 |
+
|
362 |
+
def get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag=None):
|
363 |
+
"""
|
364 |
+
Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated state_dict that can be loaded with
|
365 |
+
``load_state_dict()`` and used for training without DeepSpeed or shared with others, for example
|
366 |
+
via a model hub.
|
367 |
+
|
368 |
+
Args:
|
369 |
+
- ``checkpoint_dir``: path to the desired checkpoint folder
|
370 |
+
- ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in 'latest' file. e.g., ``global_step14``
|
371 |
+
|
372 |
+
Returns:
|
373 |
+
- pytorch ``state_dict``
|
374 |
+
|
375 |
+
Note: this approach may not work if your application doesn't have sufficient free CPU memory and
|
376 |
+
you may need to use the offline approach using the ``zero_to_fp32.py`` script that is saved with
|
377 |
+
the checkpoint.
|
378 |
+
|
379 |
+
A typical usage might be ::
|
380 |
+
|
381 |
+
from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
|
382 |
+
# do the training and checkpoint saving
|
383 |
+
state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir) # already on cpu
|
384 |
+
model = model.cpu() # move to cpu
|
385 |
+
model.load_state_dict(state_dict)
|
386 |
+
# submit to model hub or save the model to share with others
|
387 |
+
|
388 |
+
In this example the ``model`` will no longer be usable in the deepspeed context of the same
|
389 |
+
application. i.e. you will need to re-initialize the deepspeed engine, since
|
390 |
+
``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
|
391 |
+
|
392 |
+
If you want it all done for you, use ``load_state_dict_from_zero_checkpoint`` instead.
|
393 |
+
|
394 |
+
"""
|
395 |
+
if tag is None:
|
396 |
+
latest_path = os.path.join(checkpoint_dir, 'latest')
|
397 |
+
if os.path.isfile(latest_path):
|
398 |
+
with open(latest_path, 'r') as fd:
|
399 |
+
tag = fd.read().strip()
|
400 |
+
else:
|
401 |
+
raise ValueError(f"Unable to find 'latest' file at {latest_path}")
|
402 |
+
|
403 |
+
ds_checkpoint_dir = os.path.join(checkpoint_dir, tag)
|
404 |
+
|
405 |
+
if not os.path.isdir(ds_checkpoint_dir):
|
406 |
+
raise FileNotFoundError(f"Directory '{ds_checkpoint_dir}' doesn't exist")
|
407 |
+
|
408 |
+
return _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir)
|
409 |
+
|
410 |
+
|
411 |
+
def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir, output_file, tag=None):
|
412 |
+
"""
|
413 |
+
Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict`` file that can be
|
414 |
+
loaded with ``torch.load(file)`` + ``load_state_dict()`` and used for training without DeepSpeed.
|
415 |
+
|
416 |
+
Args:
|
417 |
+
- ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
|
418 |
+
- ``output_file``: path to the pytorch fp32 state_dict output file (e.g. path/pytorch_model.bin)
|
419 |
+
- ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
|
420 |
+
"""
|
421 |
+
|
422 |
+
state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
|
423 |
+
print(f"Saving fp32 state dict to {output_file}")
|
424 |
+
torch.save(state_dict, output_file)
|
425 |
+
|
426 |
+
|
427 |
+
def load_state_dict_from_zero_checkpoint(model, checkpoint_dir, tag=None):
|
428 |
+
"""
|
429 |
+
1. Put the provided model to cpu
|
430 |
+
2. Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict``
|
431 |
+
3. Load it into the provided model
|
432 |
+
|
433 |
+
Args:
|
434 |
+
- ``model``: the model object to update
|
435 |
+
- ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
|
436 |
+
- ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
|
437 |
+
|
438 |
+
Returns:
|
439 |
+
- ``model`: modified model
|
440 |
+
|
441 |
+
Make sure you have plenty of CPU memory available before you call this function. If you don't
|
442 |
+
have enough use the ``zero_to_fp32.py`` utility to do the conversion. You will find it
|
443 |
+
conveniently placed for you in the checkpoint folder.
|
444 |
+
|
445 |
+
A typical usage might be ::
|
446 |
+
|
447 |
+
from deepspeed.utils.zero_to_fp32 import load_state_dict_from_zero_checkpoint
|
448 |
+
model = load_state_dict_from_zero_checkpoint(trainer.model, checkpoint_dir)
|
449 |
+
# submit to model hub or save the model to share with others
|
450 |
+
|
451 |
+
Note, that once this was run, the ``model`` will no longer be usable in the deepspeed context
|
452 |
+
of the same application. i.e. you will need to re-initialize the deepspeed engine, since
|
453 |
+
``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
|
454 |
+
|
455 |
+
"""
|
456 |
+
logger.info(f"Extracting fp32 weights")
|
457 |
+
state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
|
458 |
+
|
459 |
+
logger.info(f"Overwriting model with fp32 weights")
|
460 |
+
model = model.cpu()
|
461 |
+
model.load_state_dict(state_dict, strict=False)
|
462 |
+
|
463 |
+
return model
|
464 |
+
|
465 |
+
|
466 |
+
if __name__ == "__main__":
|
467 |
+
|
468 |
+
parser = argparse.ArgumentParser()
|
469 |
+
parser.add_argument(
|
470 |
+
"checkpoint_dir",
|
471 |
+
type=str,
|
472 |
+
help="path to the desired checkpoint folder, e.g., path/checkpoint-12")
|
473 |
+
parser.add_argument(
|
474 |
+
"output_file",
|
475 |
+
type=str,
|
476 |
+
help=
|
477 |
+
"path to the pytorch fp32 state_dict output file (e.g. path/checkpoint-12/pytorch_model.bin)"
|
478 |
+
)
|
479 |
+
parser.add_argument("-d", "--debug", action='store_true', help="enable debug")
|
480 |
+
args = parser.parse_args()
|
481 |
+
|
482 |
+
debug = args.debug
|
483 |
+
|
484 |
+
convert_zero_checkpoint_to_fp32_state_dict(args.checkpoint_dir, args.output_file)
|