Reizxn commited on
Commit
b5de494
·
verified ·
1 Parent(s): 803b5e8

Upload folder using huggingface_hub

Browse files
Files changed (3) hide show
  1. logs/dedup_run.log +19 -0
  2. logs/sft_log.json +487 -0
  3. logs/training_log.json +930 -0
logs/dedup_run.log ADDED
@@ -0,0 +1,19 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ Loading documents from C:\Users\karsl\Documents\holy devin\retriever500m\data\raw_large\documents.jsonl...
2
+ Loaded 26,362 documents
3
+ Total size: 325.2 MB
4
+
5
+ ============================================================
6
+ STAGE 1: Exact document-level dedup (SHA-256 of normalized content)
7
+ ============================================================
8
+ Processed 5000/26362 | kept 5000 | dropped 0
9
+ Processed 10000/26362 | kept 9894 | dropped 106
10
+ Processed 15000/26362 | kept 14876 | dropped 124
11
+ Processed 20000/26362 | kept 19874 | dropped 126
12
+ Processed 25000/26362 | kept 24423 | dropped 577
13
+ Input: 26362 | Kept: 25785 | Dropped: 577 (2.2%)
14
+
15
+ ============================================================
16
+ STAGE 2: Near-dedup (MinHash LSH, threshold=0.5, perm=128, bands=32)
17
+ ============================================================
18
+ Processed 5000/25785 | clusters so far: 3188
19
+ Processed 10000/25785 | clusters so far: 5795
logs/sft_log.json ADDED
@@ -0,0 +1,487 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "config": {
3
+ "vocab_size": 32009,
4
+ "d_model": 1280,
5
+ "n_layers": 23,
6
+ "n_heads": 20,
7
+ "d_ff": 3456,
8
+ "max_seq_len": 768,
9
+ "rope_theta": 10000.0,
10
+ "rope_pct": 0.25,
11
+ "dropout": 0.0,
12
+ "tie_embeddings": true
13
+ },
14
+ "train_args": {
15
+ "steps": 500,
16
+ "batch_size": 4,
17
+ "grad_accum": 8,
18
+ "seq_len": 768,
19
+ "lr": 5e-05,
20
+ "warmup": 20,
21
+ "save_every": 100,
22
+ "log_every": 10,
23
+ "use_8bit_adam": true
24
+ },
25
+ "total_params": 496998400,
26
+ "steps": [
27
+ {
28
+ "step": 0,
29
+ "loss": 6.937812924385071,
30
+ "ema_loss": 0.3468906462192536,
31
+ "lr": 2.5e-06,
32
+ "elapsed_s": 4.15720272064209,
33
+ "steps_per_sec": 0.2405463642738951,
34
+ "vram_gb": 9.164018176
35
+ },
36
+ {
37
+ "step": 10,
38
+ "loss": 2.739780366420746,
39
+ "ema_loss": 1.9638136853068686,
40
+ "lr": 2.75e-05,
41
+ "elapsed_s": 40.6308798789978,
42
+ "steps_per_sec": 0.2707300465251781,
43
+ "vram_gb": 10.17390848
44
+ },
45
+ {
46
+ "step": 20,
47
+ "loss": 1.2084012106060982,
48
+ "ema_loss": 1.882433597453757,
49
+ "lr": 5e-05,
50
+ "elapsed_s": 77.03947591781616,
51
+ "steps_per_sec": 0.27258752412078047,
52
+ "vram_gb": 10.17390848
53
+ },
54
+ {
55
+ "step": 30,
56
+ "loss": 1.0680485740303993,
57
+ "ema_loss": 1.5722938780650435,
58
+ "lr": 4.995182577286858e-05,
59
+ "elapsed_s": 113.55863189697266,
60
+ "steps_per_sec": 0.27298673365601217,
61
+ "vram_gb": 10.17390848
62
+ },
63
+ {
64
+ "step": 40,
65
+ "loss": 0.7858129814267159,
66
+ "ema_loss": 1.2813095330935025,
67
+ "lr": 4.980750938091074e-05,
68
+ "elapsed_s": 149.9709255695343,
69
+ "steps_per_sec": 0.27338632367772026,
70
+ "vram_gb": 10.17390848
71
+ },
72
+ {
73
+ "step": 50,
74
+ "loss": 0.6811780035495758,
75
+ "ema_loss": 1.0847747264114023,
76
+ "lr": 4.956766880907269e-05,
77
+ "elapsed_s": 186.4907727241516,
78
+ "steps_per_sec": 0.2734719753423769,
79
+ "vram_gb": 10.17390848
80
+ },
81
+ {
82
+ "step": 60,
83
+ "loss": 0.6801792122423649,
84
+ "ema_loss": 0.9220971225814172,
85
+ "lr": 4.9233331091504034e-05,
86
+ "elapsed_s": 222.90015697479248,
87
+ "steps_per_sec": 0.27366512804608933,
88
+ "vram_gb": 10.17390848
89
+ },
90
+ {
91
+ "step": 70,
92
+ "loss": 0.4824926480650902,
93
+ "ema_loss": 0.8074420473552227,
94
+ "lr": 4.8805927913639885e-05,
95
+ "elapsed_s": 259.41771626472473,
96
+ "steps_per_sec": 0.2736898659902916,
97
+ "vram_gb": 10.17390848
98
+ },
99
+ {
100
+ "step": 80,
101
+ "loss": 0.4920233264565468,
102
+ "ema_loss": 0.7012675245586312,
103
+ "lr": 4.8287289481503954e-05,
104
+ "elapsed_s": 295.8301751613617,
105
+ "steps_per_sec": 0.2738057399175667,
106
+ "vram_gb": 10.17390848
107
+ },
108
+ {
109
+ "step": 90,
110
+ "loss": 0.6347984820604324,
111
+ "ema_loss": 0.6271864858820199,
112
+ "lr": 4.767963668448549e-05,
113
+ "elapsed_s": 332.34874868392944,
114
+ "steps_per_sec": 0.2738087637168837,
115
+ "vram_gb": 10.17390848
116
+ },
117
+ {
118
+ "step": 100,
119
+ "loss": 0.6809600777924061,
120
+ "ema_loss": 0.5743832670414681,
121
+ "lr": 4.6985571585149876e-05,
122
+ "elapsed_s": 414.7752454280853,
123
+ "steps_per_sec": 0.24350537095278896,
124
+ "vram_gb": 10.17390848
125
+ },
126
+ {
127
+ "step": 110,
128
+ "loss": 0.4763348139822483,
129
+ "ema_loss": 0.521244228570747,
130
+ "lr": 4.620806627680727e-05,
131
+ "elapsed_s": 451.26919865608215,
132
+ "steps_per_sec": 0.24597291446118502,
133
+ "vram_gb": 10.17390848
134
+ },
135
+ {
136
+ "step": 120,
137
+ "loss": 0.5240486040711403,
138
+ "ema_loss": 0.494888873638692,
139
+ "lr": 4.535045015655279e-05,
140
+ "elapsed_s": 487.7379126548767,
141
+ "steps_per_sec": 0.24808405674549147,
142
+ "vram_gb": 10.17390848
143
+ },
144
+ {
145
+ "step": 130,
146
+ "loss": 0.286451730877161,
147
+ "ema_loss": 0.46101876181189894,
148
+ "lr": 4.4416395668276995e-05,
149
+ "elapsed_s": 524.1727938652039,
150
+ "steps_per_sec": 0.24991758735514977,
151
+ "vram_gb": 10.17390848
152
+ },
153
+ {
154
+ "step": 140,
155
+ "loss": 0.35978504456579685,
156
+ "ema_loss": 0.4399799286456892,
157
+ "lr": 4.340990257669732e-05,
158
+ "elapsed_s": 560.6376507282257,
159
+ "steps_per_sec": 0.25149934153878484,
160
+ "vram_gb": 10.17390848
161
+ },
162
+ {
163
+ "step": 150,
164
+ "loss": 0.3526405543088913,
165
+ "ema_loss": 0.40866163158804786,
166
+ "lr": 4.2335280839751554e-05,
167
+ "elapsed_s": 597.072151184082,
168
+ "steps_per_sec": 0.25290075864457046,
169
+ "vram_gb": 10.17390848
170
+ },
171
+ {
172
+ "step": 160,
173
+ "loss": 0.3449416421353817,
174
+ "ema_loss": 0.36948892724765275,
175
+ "lr": 4.1197132152696215e-05,
176
+ "elapsed_s": 633.5287528038025,
177
+ "steps_per_sec": 0.2541321120587878,
178
+ "vram_gb": 10.17390848
179
+ },
180
+ {
181
+ "step": 170,
182
+ "loss": 0.32148140482604504,
183
+ "ema_loss": 0.3509791943321394,
184
+ "lr": 4.000033024294105e-05,
185
+ "elapsed_s": 669.9639327526093,
186
+ "steps_per_sec": 0.2552376204752255,
187
+ "vram_gb": 10.17390848
188
+ },
189
+ {
190
+ "step": 180,
191
+ "loss": 0.40944845974445343,
192
+ "ema_loss": 0.3403033324976888,
193
+ "lr": 3.875e-05,
194
+ "elapsed_s": 706.4267771244049,
195
+ "steps_per_sec": 0.2562190532142372,
196
+ "vram_gb": 10.17390848
197
+ },
198
+ {
199
+ "step": 190,
200
+ "loss": 0.2508099125698209,
201
+ "ema_loss": 0.31061072625747166,
202
+ "lr": 3.7451495529927534e-05,
203
+ "elapsed_s": 742.8610553741455,
204
+ "steps_per_sec": 0.25711403043440195,
205
+ "vram_gb": 10.17390848
206
+ },
207
+ {
208
+ "step": 200,
209
+ "loss": 0.2846630644053221,
210
+ "ema_loss": 0.29097493865487023,
211
+ "lr": 3.6110377228214525e-05,
212
+ "elapsed_s": 823.1199431419373,
213
+ "steps_per_sec": 0.2441928441592138,
214
+ "vram_gb": 10.17390848
215
+ },
216
+ {
217
+ "step": 210,
218
+ "loss": 0.1874328227713704,
219
+ "ema_loss": 0.2691413417018003,
220
+ "lr": 3.473238796932114e-05,
221
+ "elapsed_s": 859.5507295131683,
222
+ "steps_per_sec": 0.24547707628554513,
223
+ "vram_gb": 10.17390848
224
+ },
225
+ {
226
+ "step": 220,
227
+ "loss": 0.18770729657262564,
228
+ "ema_loss": 0.2469535500748924,
229
+ "lr": 3.332342851480672e-05,
230
+ "elapsed_s": 896.0201919078827,
231
+ "steps_per_sec": 0.24664622739073316,
232
+ "vram_gb": 10.17390848
233
+ },
234
+ {
235
+ "step": 230,
236
+ "loss": 0.20278130657970905,
237
+ "ema_loss": 0.2411670786782403,
238
+ "lr": 3.1889532245362895e-05,
239
+ "elapsed_s": 932.4535627365112,
240
+ "steps_per_sec": 0.2477335164252839,
241
+ "vram_gb": 10.17390848
242
+ },
243
+ {
244
+ "step": 240,
245
+ "loss": 0.24111063033342361,
246
+ "ema_loss": 0.2326045524210834,
247
+ "lr": 3.0436839324951166e-05,
248
+ "elapsed_s": 968.919579744339,
249
+ "steps_per_sec": 0.24873065323294502,
250
+ "vram_gb": 10.17390848
251
+ },
252
+ {
253
+ "step": 250,
254
+ "loss": 0.2577468641102314,
255
+ "ema_loss": 0.22216086817509004,
256
+ "lr": 2.897157040767822e-05,
257
+ "elapsed_s": 1005.3552136421204,
258
+ "steps_per_sec": 0.24966300128956143,
259
+ "vram_gb": 10.17390848
260
+ },
261
+ {
262
+ "step": 260,
263
+ "loss": 0.21209231484681368,
264
+ "ema_loss": 0.2089893973874345,
265
+ "lr": 2.75e-05,
266
+ "elapsed_s": 1041.8542084693909,
267
+ "steps_per_sec": 0.25051489726517534,
268
+ "vram_gb": 10.17390848
269
+ },
270
+ {
271
+ "step": 270,
272
+ "loss": 0.16555726807564497,
273
+ "ema_loss": 0.20300413108044887,
274
+ "lr": 2.602842959232178e-05,
275
+ "elapsed_s": 1078.29061460495,
276
+ "steps_per_sec": 0.25132371211381216,
277
+ "vram_gb": 10.17390848
278
+ },
279
+ {
280
+ "step": 280,
281
+ "loss": 0.20196319418027997,
282
+ "ema_loss": 0.1959523809991305,
283
+ "lr": 2.4563160675048846e-05,
284
+ "elapsed_s": 1114.7595565319061,
285
+ "steps_per_sec": 0.25207229519001423,
286
+ "vram_gb": 10.17390848
287
+ },
288
+ {
289
+ "step": 290,
290
+ "loss": 0.1890458231791854,
291
+ "ema_loss": 0.19173463181336906,
292
+ "lr": 2.3110467754637118e-05,
293
+ "elapsed_s": 1151.1741054058075,
294
+ "steps_per_sec": 0.2527853941758165,
295
+ "vram_gb": 10.17390848
296
+ },
297
+ {
298
+ "step": 300,
299
+ "loss": 0.14224259229376912,
300
+ "ema_loss": 0.170971237515532,
301
+ "lr": 2.1676571485193282e-05,
302
+ "elapsed_s": 1231.5103888511658,
303
+ "steps_per_sec": 0.2444153153111381,
304
+ "vram_gb": 10.17390848
305
+ },
306
+ {
307
+ "step": 310,
308
+ "loss": 0.1485865106806159,
309
+ "ema_loss": 0.17176215697934846,
310
+ "lr": 2.026761203067887e-05,
311
+ "elapsed_s": 1267.9178903102875,
312
+ "steps_per_sec": 0.24528402223577067,
313
+ "vram_gb": 10.17390848
314
+ },
315
+ {
316
+ "step": 320,
317
+ "loss": 0.17956801364198327,
318
+ "ema_loss": 0.1700756775799304,
319
+ "lr": 1.888962277178548e-05,
320
+ "elapsed_s": 1304.4403862953186,
321
+ "steps_per_sec": 0.2460825372876237,
322
+ "vram_gb": 10.17390848
323
+ },
324
+ {
325
+ "step": 330,
326
+ "loss": 0.12682513892650604,
327
+ "ema_loss": 0.15713850400483031,
328
+ "lr": 1.7548504470072478e-05,
329
+ "elapsed_s": 1340.8541731834412,
330
+ "steps_per_sec": 0.24685756782495105,
331
+ "vram_gb": 10.17390848
332
+ },
333
+ {
334
+ "step": 340,
335
+ "loss": 0.12957867234945297,
336
+ "ema_loss": 0.1510191265222257,
337
+ "lr": 1.6250000000000005e-05,
338
+ "elapsed_s": 1377.3766162395477,
339
+ "steps_per_sec": 0.24757208448258908,
340
+ "vram_gb": 10.17390848
341
+ },
342
+ {
343
+ "step": 350,
344
+ "loss": 0.12622385285794735,
345
+ "ema_loss": 0.14607439549355963,
346
+ "lr": 1.4999669757058956e-05,
347
+ "elapsed_s": 1413.787192106247,
348
+ "steps_per_sec": 0.24826933074495,
349
+ "vram_gb": 10.17390848
350
+ },
351
+ {
352
+ "step": 360,
353
+ "loss": 0.1389128677546978,
354
+ "ema_loss": 0.1373236412001944,
355
+ "lr": 1.3802867847303787e-05,
356
+ "elapsed_s": 1450.3096079826355,
357
+ "steps_per_sec": 0.24891236878872158,
358
+ "vram_gb": 10.17390848
359
+ },
360
+ {
361
+ "step": 370,
362
+ "loss": 0.11274528503417969,
363
+ "ema_loss": 0.1358663910950191,
364
+ "lr": 1.2664719160248452e-05,
365
+ "elapsed_s": 1486.721604347229,
366
+ "steps_per_sec": 0.24954234801941552,
367
+ "vram_gb": 10.17390848
368
+ },
369
+ {
370
+ "step": 380,
371
+ "loss": 0.12164316419512033,
372
+ "ema_loss": 0.12515423124366815,
373
+ "lr": 1.1590097423302684e-05,
374
+ "elapsed_s": 1523.1735746860504,
375
+ "steps_per_sec": 0.2501356420121259,
376
+ "vram_gb": 10.17390848
377
+ },
378
+ {
379
+ "step": 390,
380
+ "loss": 0.10157549497671425,
381
+ "ema_loss": 0.1216169996836354,
382
+ "lr": 1.0583604331723012e-05,
383
+ "elapsed_s": 1559.6568088531494,
384
+ "steps_per_sec": 0.25069617737732386,
385
+ "vram_gb": 10.17390848
386
+ },
387
+ {
388
+ "step": 400,
389
+ "loss": 0.10004387237131596,
390
+ "ema_loss": 0.12196290672243944,
391
+ "lr": 9.649549843447213e-06,
392
+ "elapsed_s": 1641.5576264858246,
393
+ "steps_per_sec": 0.24428018458203227,
394
+ "vram_gb": 10.17390848
395
+ },
396
+ {
397
+ "step": 410,
398
+ "loss": 0.15670533804222941,
399
+ "ema_loss": 0.11768572540664866,
400
+ "lr": 8.79193372319273e-06,
401
+ "elapsed_s": 1678.030652999878,
402
+ "steps_per_sec": 0.24492997149083062,
403
+ "vram_gb": 10.17390848
404
+ },
405
+ {
406
+ "step": 420,
407
+ "loss": 0.0967125273309648,
408
+ "ema_loss": 0.12051557885116983,
409
+ "lr": 8.01442841485013e-06,
410
+ "elapsed_s": 1714.4823727607727,
411
+ "steps_per_sec": 0.24555516387262588,
412
+ "vram_gb": 10.17390848
413
+ },
414
+ {
415
+ "step": 430,
416
+ "loss": 0.08109346427954733,
417
+ "ema_loss": 0.11255813170642848,
418
+ "lr": 7.320363315514515e-06,
419
+ "elapsed_s": 1750.9651441574097,
420
+ "steps_per_sec": 0.24614995988820987,
421
+ "vram_gb": 10.17390848
422
+ },
423
+ {
424
+ "step": 440,
425
+ "loss": 0.06976099917665124,
426
+ "ema_loss": 0.10574265881842762,
427
+ "lr": 6.712710518496049e-06,
428
+ "elapsed_s": 1787.420247077942,
429
+ "steps_per_sec": 0.24672429481591848,
430
+ "vram_gb": 10.17390848
431
+ },
432
+ {
433
+ "step": 450,
434
+ "loss": 0.08222721493802965,
435
+ "ema_loss": 0.09805770252970188,
436
+ "lr": 6.194072086360125e-06,
437
+ "elapsed_s": 1823.9047615528107,
438
+ "steps_per_sec": 0.24727168298855356,
439
+ "vram_gb": 10.17390848
440
+ },
441
+ {
442
+ "step": 460,
443
+ "loss": 0.09003248857334256,
444
+ "ema_loss": 0.09465487723876762,
445
+ "lr": 5.766668908495966e-06,
446
+ "elapsed_s": 1860.354943037033,
447
+ "steps_per_sec": 0.24780217437830257,
448
+ "vram_gb": 10.17390848
449
+ },
450
+ {
451
+ "step": 470,
452
+ "loss": 0.050968108931556344,
453
+ "ema_loss": 0.0877422503218975,
454
+ "lr": 5.432331190927316e-06,
455
+ "elapsed_s": 1896.8366348743439,
456
+ "steps_per_sec": 0.24830815228914083,
457
+ "vram_gb": 10.17390848
458
+ },
459
+ {
460
+ "step": 480,
461
+ "loss": 0.09658047696575522,
462
+ "ema_loss": 0.08559591419795948,
463
+ "lr": 5.192490619089267e-06,
464
+ "elapsed_s": 1933.280499458313,
465
+ "steps_per_sec": 0.24879990261877244,
466
+ "vram_gb": 10.17390848
467
+ },
468
+ {
469
+ "step": 490,
470
+ "loss": 0.057105707470327616,
471
+ "ema_loss": 0.08398783932396543,
472
+ "lr": 5.048174227131422e-06,
473
+ "elapsed_s": 1969.7602517604828,
474
+ "steps_per_sec": 0.24926891461088543,
475
+ "vram_gb": 10.17390848
476
+ },
477
+ {
478
+ "step": 499,
479
+ "loss": 0.0780253498815,
480
+ "ema_loss": 0.08407452203461949,
481
+ "lr": 5.000481912557093e-06,
482
+ "elapsed_s": 2002.5768718719482,
483
+ "steps_per_sec": 0.24967830549876227,
484
+ "vram_gb": 10.17390848
485
+ }
486
+ ]
487
+ }
logs/training_log.json ADDED
@@ -0,0 +1,930 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "config": {
3
+ "vocab_size": 32000,
4
+ "d_model": 1280,
5
+ "n_layers": 23,
6
+ "n_heads": 20,
7
+ "d_ff": 3456,
8
+ "max_seq_len": 512,
9
+ "rope_theta": 10000.0,
10
+ "rope_pct": 0.25,
11
+ "dropout": 0.0,
12
+ "tie_embeddings": true
13
+ },
14
+ "train_args": {
15
+ "steps": 1000,
16
+ "batch_size": 8,
17
+ "grad_accum": 4,
18
+ "seq_len": 512,
19
+ "lr": 0.0003,
20
+ "warmup": 100,
21
+ "save_every": 200,
22
+ "log_every": 10,
23
+ "corpus": "curated",
24
+ "resume": true,
25
+ "use_8bit_adam": true
26
+ },
27
+ "total_params": 496986880,
28
+ "steps": [
29
+ {
30
+ "step": 720,
31
+ "loss": 4.28348183631897,
32
+ "ema_loss": 4.180650027796879,
33
+ "lr": 0.00021277697160092742,
34
+ "elapsed_s": 17.697318077087402,
35
+ "steps_per_sec": 0.5650573695088259,
36
+ "vram_gb": 15.430634496
37
+ },
38
+ {
39
+ "step": 730,
40
+ "loss": 4.118941903114319,
41
+ "ema_loss": 4.135945131418912,
42
+ "lr": 0.00021030580199599434,
43
+ "elapsed_s": 35.07432723045349,
44
+ "steps_per_sec": 0.5702176372077319,
45
+ "vram_gb": 15.430634496
46
+ },
47
+ {
48
+ "step": 740,
49
+ "loss": 3.6725871562957764,
50
+ "ema_loss": 4.0054830158378945,
51
+ "lr": 0.00020781740384119815,
52
+ "elapsed_s": 52.46801733970642,
53
+ "steps_per_sec": 0.5717768942127871,
54
+ "vram_gb": 15.430634496
55
+ },
56
+ {
57
+ "step": 750,
58
+ "loss": 3.6983203291893005,
59
+ "ema_loss": 3.8964221631363376,
60
+ "lr": 0.00020531272340588638,
61
+ "elapsed_s": 69.84280610084534,
62
+ "steps_per_sec": 0.5727146750410399,
63
+ "vram_gb": 15.430634496
64
+ },
65
+ {
66
+ "step": 760,
67
+ "loss": 3.6233853101730347,
68
+ "ema_loss": 3.7636052763477563,
69
+ "lr": 0.00020279271315110968,
70
+ "elapsed_s": 87.27109622955322,
71
+ "steps_per_sec": 0.5729273741271987,
72
+ "vram_gb": 15.428668416
73
+ },
74
+ {
75
+ "step": 770,
76
+ "loss": 3.3926801681518555,
77
+ "ema_loss": 3.6054279373031792,
78
+ "lr": 0.00020025833136742718,
79
+ "elapsed_s": 104.76516604423523,
80
+ "steps_per_sec": 0.5727094440404559,
81
+ "vram_gb": 15.428668416
82
+ },
83
+ {
84
+ "step": 780,
85
+ "loss": 3.1609978079795837,
86
+ "ema_loss": 3.4584633810890293,
87
+ "lr": 0.00019771054181049524,
88
+ "elapsed_s": 122.30735993385315,
89
+ "steps_per_sec": 0.5723285993406916,
90
+ "vram_gb": 15.428668416
91
+ },
92
+ {
93
+ "step": 790,
94
+ "loss": 3.1603373289108276,
95
+ "ema_loss": 3.309057160849107,
96
+ "lr": 0.00019515031333457712,
97
+ "elapsed_s": 139.78667426109314,
98
+ "steps_per_sec": 0.5723006175150589,
99
+ "vram_gb": 15.428668416
100
+ },
101
+ {
102
+ "step": 800,
103
+ "loss": 2.7271658778190613,
104
+ "ema_loss": 3.1769729487969567,
105
+ "lr": 0.00019257861952411492,
106
+ "elapsed_s": 200.5202157497406,
107
+ "steps_per_sec": 0.44883255118937515,
108
+ "vram_gb": 15.428668416
109
+ },
110
+ {
111
+ "step": 810,
112
+ "loss": 3.0822166204452515,
113
+ "ema_loss": 3.088718658479337,
114
+ "lr": 0.00018999643832350203,
115
+ "elapsed_s": 217.9351396560669,
116
+ "steps_per_sec": 0.4588521160828604,
117
+ "vram_gb": 15.428668416
118
+ },
119
+ {
120
+ "step": 820,
121
+ "loss": 2.85100781917572,
122
+ "ema_loss": 3.003455477485217,
123
+ "lr": 0.00018740475166519754,
124
+ "elapsed_s": 235.42823815345764,
125
+ "steps_per_sec": 0.4672336711295415,
126
+ "vram_gb": 15.428668416
127
+ },
128
+ {
129
+ "step": 830,
130
+ "loss": 2.7287580966949463,
131
+ "ema_loss": 2.9638366754161387,
132
+ "lr": 0.0001848045450963247,
133
+ "elapsed_s": 252.95333313941956,
134
+ "steps_per_sec": 0.4743958045963361,
135
+ "vram_gb": 15.428668416
136
+ },
137
+ {
138
+ "step": 840,
139
+ "loss": 2.7534950971603394,
140
+ "ema_loss": 2.85498482454621,
141
+ "lr": 0.00018219680740389431,
142
+ "elapsed_s": 270.4455711841583,
143
+ "steps_per_sec": 0.48068821918875965,
144
+ "vram_gb": 15.428668416
145
+ },
146
+ {
147
+ "step": 850,
148
+ "loss": 2.6869542002677917,
149
+ "ema_loss": 2.7344452315263985,
150
+ "lr": 0.00017958253023879622,
151
+ "elapsed_s": 287.96280574798584,
152
+ "steps_per_sec": 0.48617389886985163,
153
+ "vram_gb": 15.428668416
154
+ },
155
+ {
156
+ "step": 860,
157
+ "loss": 2.446716547012329,
158
+ "ema_loss": 2.6342544765284797,
159
+ "lr": 0.00017696270773870167,
160
+ "elapsed_s": 305.5090491771698,
161
+ "steps_per_sec": 0.4909838199686599,
162
+ "vram_gb": 15.428668416
163
+ },
164
+ {
165
+ "step": 870,
166
+ "loss": 2.5103383660316467,
167
+ "ema_loss": 2.5549315583614782,
168
+ "lr": 0.0001743383361500204,
169
+ "elapsed_s": 323.075519323349,
170
+ "steps_per_sec": 0.49524024703296865,
171
+ "vram_gb": 15.428668416
172
+ },
173
+ {
174
+ "step": 880,
175
+ "loss": 2.2758803963661194,
176
+ "ema_loss": 2.480248449238682,
177
+ "lr": 0.00017171041344905526,
178
+ "elapsed_s": 340.6705768108368,
179
+ "steps_per_sec": 0.4990157987562144,
180
+ "vram_gb": 15.428668416
181
+ },
182
+ {
183
+ "step": 890,
184
+ "loss": 2.280333399772644,
185
+ "ema_loss": 2.428152495952108,
186
+ "lr": 0.00016907993896249943,
187
+ "elapsed_s": 358.2743320465088,
188
+ "steps_per_sec": 0.5024083053112317,
189
+ "vram_gb": 15.428668416
190
+ },
191
+ {
192
+ "step": 900,
193
+ "loss": 2.089028149843216,
194
+ "ema_loss": 2.309313885635373,
195
+ "lr": 0.00016644791298741992,
196
+ "elapsed_s": 375.8712692260742,
197
+ "steps_per_sec": 0.5054922138401625,
198
+ "vram_gb": 15.428668416
199
+ },
200
+ {
201
+ "step": 910,
202
+ "loss": 2.392301082611084,
203
+ "ema_loss": 2.231137390297292,
204
+ "lr": 0.00016381533641087205,
205
+ "elapsed_s": 393.4869828224182,
206
+ "steps_per_sec": 0.5082760262243811,
207
+ "vram_gb": 15.428668416
208
+ },
209
+ {
210
+ "step": 920,
211
+ "loss": 2.169742077589035,
212
+ "ema_loss": 2.1655222685266406,
213
+ "lr": 0.00016118321032928992,
214
+ "elapsed_s": 411.1121575832367,
215
+ "steps_per_sec": 0.5108095105591274,
216
+ "vram_gb": 15.428668416
217
+ },
218
+ {
219
+ "step": 930,
220
+ "loss": 1.7299185693264008,
221
+ "ema_loss": 2.0710785211805653,
222
+ "lr": 0.00015855253566779667,
223
+ "elapsed_s": 428.7321665287018,
224
+ "steps_per_sec": 0.5131408771617605,
225
+ "vram_gb": 15.428668416
226
+ },
227
+ {
228
+ "step": 940,
229
+ "loss": 2.0483551025390625,
230
+ "ema_loss": 1.9999614091930757,
231
+ "lr": 0.00015592431279958026,
232
+ "elapsed_s": 446.3725197315216,
233
+ "steps_per_sec": 0.5152646944716431,
234
+ "vram_gb": 15.428668416
235
+ },
236
+ {
237
+ "step": 950,
238
+ "loss": 1.770757019519806,
239
+ "ema_loss": 1.942065600070645,
240
+ "lr": 0.00015329954116547926,
241
+ "elapsed_s": 464.01725935935974,
242
+ "steps_per_sec": 0.5172221402526133,
243
+ "vram_gb": 15.428668416
244
+ },
245
+ {
246
+ "step": 960,
247
+ "loss": 1.6447540819644928,
248
+ "ema_loss": 1.9088268255451808,
249
+ "lr": 0.00015067921889392276,
250
+ "elapsed_s": 481.6594491004944,
251
+ "steps_per_sec": 0.5190389194416895,
252
+ "vram_gb": 15.428668416
253
+ },
254
+ {
255
+ "step": 970,
256
+ "loss": 1.758211225271225,
257
+ "ema_loss": 1.837405344969316,
258
+ "lr": 0.00014806434242136985,
259
+ "elapsed_s": 499.30819845199585,
260
+ "steps_per_sec": 0.5207204704550765,
261
+ "vram_gb": 15.428668416
262
+ },
263
+ {
264
+ "step": 980,
265
+ "loss": 1.5929311215877533,
266
+ "ema_loss": 1.7585266107130897,
267
+ "lr": 0.0001454559061133919,
268
+ "elapsed_s": 516.9588961601257,
269
+ "steps_per_sec": 0.522285237773273,
270
+ "vram_gb": 15.428668416
271
+ },
272
+ {
273
+ "step": 990,
274
+ "loss": 1.3716070353984833,
275
+ "ema_loss": 1.664784491957035,
276
+ "lr": 0.00014285490188654308,
277
+ "elapsed_s": 534.6144878864288,
278
+ "steps_per_sec": 0.5237418856847029,
279
+ "vram_gb": 15.428668416
280
+ },
281
+ {
282
+ "step": 1000,
283
+ "loss": 1.5660839676856995,
284
+ "ema_loss": 1.6258334135726564,
285
+ "lr": 0.00014026231883116147,
286
+ "elapsed_s": 592.9495124816895,
287
+ "steps_per_sec": 0.48908042572841365,
288
+ "vram_gb": 15.428668416
289
+ },
290
+ {
291
+ "step": 1010,
292
+ "loss": 1.3805969655513763,
293
+ "ema_loss": 1.5515371576395063,
294
+ "lr": 0.00013767914283524546,
295
+ "elapsed_s": 610.4368507862091,
296
+ "steps_per_sec": 0.4914513263961972,
297
+ "vram_gb": 15.428668416
298
+ },
299
+ {
300
+ "step": 1020,
301
+ "loss": 1.6006771624088287,
302
+ "ema_loss": 1.4895273508195535,
303
+ "lr": 0.0001351063562095476,
304
+ "elapsed_s": 627.9561829566956,
305
+ "steps_per_sec": 0.4936650174226852,
306
+ "vram_gb": 15.428668416
307
+ },
308
+ {
309
+ "step": 1030,
310
+ "loss": 1.2836307883262634,
311
+ "ema_loss": 1.4065686388898713,
312
+ "lr": 0.00013254493731402865,
313
+ "elapsed_s": 645.4572558403015,
314
+ "steps_per_sec": 0.4957725660445192,
315
+ "vram_gb": 15.428668416
316
+ },
317
+ {
318
+ "step": 1040,
319
+ "loss": 1.3605864644050598,
320
+ "ema_loss": 1.3870377512296692,
321
+ "lr": 0.0001299958601858145,
322
+ "elapsed_s": 662.9888000488281,
323
+ "steps_per_sec": 0.49774596490271933,
324
+ "vram_gb": 15.428668416
325
+ },
326
+ {
327
+ "step": 1050,
328
+ "loss": 1.2070442289113998,
329
+ "ema_loss": 1.3173434739374783,
330
+ "lr": 0.00012746009416879654,
331
+ "elapsed_s": 680.5442576408386,
332
+ "steps_per_sec": 0.4996001306052267,
333
+ "vram_gb": 15.428668416
334
+ },
335
+ {
336
+ "step": 1060,
337
+ "loss": 1.0203208923339844,
338
+ "ema_loss": 1.2372483979334374,
339
+ "lr": 0.0001249386035450169,
340
+ "elapsed_s": 698.1243305206299,
341
+ "steps_per_sec": 0.5013433634936998,
342
+ "vram_gb": 15.428668416
343
+ },
344
+ {
345
+ "step": 1070,
346
+ "loss": 1.2167232036590576,
347
+ "ema_loss": 1.1832412849618354,
348
+ "lr": 0.00012243234716797824,
349
+ "elapsed_s": 715.7207224369049,
350
+ "steps_per_sec": 0.5029894883779004,
351
+ "vram_gb": 15.428668416
352
+ },
353
+ {
354
+ "step": 1080,
355
+ "loss": 0.9296802133321762,
356
+ "ema_loss": 1.13472966931652,
357
+ "lr": 0.00011994227809801857,
358
+ "elapsed_s": 733.3160738945007,
359
+ "steps_per_sec": 0.504557329604138,
360
+ "vram_gb": 15.428668416
361
+ },
362
+ {
363
+ "step": 1090,
364
+ "loss": 1.1032112538814545,
365
+ "ema_loss": 1.0848564241444187,
366
+ "lr": 0.00011746934323988837,
367
+ "elapsed_s": 750.9138088226318,
368
+ "steps_per_sec": 0.506050089284957,
369
+ "vram_gb": 15.428668416
370
+ },
371
+ {
372
+ "step": 1100,
373
+ "loss": 0.7624894231557846,
374
+ "ema_loss": 1.035134231257411,
375
+ "lr": 0.0001150144829826688,
376
+ "elapsed_s": 768.5393669605255,
377
+ "steps_per_sec": 0.5074561132013314,
378
+ "vram_gb": 15.428668416
379
+ },
380
+ {
381
+ "step": 1110,
382
+ "loss": 1.0394368767738342,
383
+ "ema_loss": 0.9800935883989342,
384
+ "lr": 0.00011257863084216771,
385
+ "elapsed_s": 786.1641900539398,
386
+ "steps_per_sec": 0.5087995676482738,
387
+ "vram_gb": 15.428668416
388
+ },
389
+ {
390
+ "step": 1120,
391
+ "loss": 0.8956251740455627,
392
+ "ema_loss": 0.9325636114854202,
393
+ "lr": 0.0001101627131059291,
394
+ "elapsed_s": 803.8073284626007,
395
+ "steps_per_sec": 0.5100724831461603,
396
+ "vram_gb": 15.428668416
397
+ },
398
+ {
399
+ "step": 1130,
400
+ "loss": 0.7532132491469383,
401
+ "ema_loss": 0.8859757228483139,
402
+ "lr": 0.00010776764848099159,
403
+ "elapsed_s": 821.4637365341187,
404
+ "steps_per_sec": 0.5112824599805761,
405
+ "vram_gb": 15.428668416
406
+ },
407
+ {
408
+ "step": 1140,
409
+ "loss": 0.7787892669439316,
410
+ "ema_loss": 0.858024045693639,
411
+ "lr": 0.00010539434774452933,
412
+ "elapsed_s": 839.1095054149628,
413
+ "steps_per_sec": 0.5124480145024136,
414
+ "vram_gb": 15.428668416
415
+ },
416
+ {
417
+ "step": 1150,
418
+ "loss": 0.7684804052114487,
419
+ "ema_loss": 0.8112544161741339,
420
+ "lr": 0.0001030437133975086,
421
+ "elapsed_s": 856.7590296268463,
422
+ "steps_per_sec": 0.5135633063495555,
423
+ "vram_gb": 15.428668416
424
+ },
425
+ {
426
+ "step": 1160,
427
+ "loss": 0.6312309950590134,
428
+ "ema_loss": 0.7749217716574573,
429
+ "lr": 0.00010071663932149149,
430
+ "elapsed_s": 874.4188885688782,
431
+ "steps_per_sec": 0.5146274924784557,
432
+ "vram_gb": 15.428668416
433
+ },
434
+ {
435
+ "step": 1170,
436
+ "loss": 0.6993876099586487,
437
+ "ema_loss": 0.7453027083515941,
438
+ "lr": 9.841401043871715e-05,
439
+ "elapsed_s": 892.0863609313965,
440
+ "steps_per_sec": 0.5156451439519039,
441
+ "vram_gb": 15.428668416
442
+ },
443
+ {
444
+ "step": 1180,
445
+ "loss": 0.5929347425699234,
446
+ "ema_loss": 0.7042416589248688,
447
+ "lr": 9.613670237559053e-05,
448
+ "elapsed_s": 909.741069316864,
449
+ "steps_per_sec": 0.5166305181241613,
450
+ "vram_gb": 15.428668416
451
+ },
452
+ {
453
+ "step": 1190,
454
+ "loss": 0.6512271761894226,
455
+ "ema_loss": 0.6882631183908091,
456
+ "lr": 9.388558112970555e-05,
457
+ "elapsed_s": 927.3969824314117,
458
+ "steps_per_sec": 0.5175777030690304,
459
+ "vram_gb": 15.428668416
460
+ },
461
+ {
462
+ "step": 1200,
463
+ "loss": 0.6192804276943207,
464
+ "ema_loss": 0.6477409202862138,
465
+ "lr": 9.166150274053016e-05,
466
+ "elapsed_s": 986.2773616313934,
467
+ "steps_per_sec": 0.4968176489314273,
468
+ "vram_gb": 15.428668416
469
+ },
470
+ {
471
+ "step": 1210,
472
+ "loss": 0.5546711385250092,
473
+ "ema_loss": 0.6117017021277563,
474
+ "lr": 8.946531296387777e-05,
475
+ "elapsed_s": 1003.7613523006439,
476
+ "steps_per_sec": 0.49812637122757175,
477
+ "vram_gb": 15.428668416
478
+ },
479
+ {
480
+ "step": 1220,
481
+ "loss": 0.481452077627182,
482
+ "ema_loss": 0.5780685955709823,
483
+ "lr": 8.729784695029022e-05,
484
+ "elapsed_s": 1021.3000912666321,
485
+ "steps_per_sec": 0.4993635116271165,
486
+ "vram_gb": 15.428668416
487
+ },
488
+ {
489
+ "step": 1230,
490
+ "loss": 0.45283564925193787,
491
+ "ema_loss": 0.5569614956021417,
492
+ "lr": 8.515992892745223e-05,
493
+ "elapsed_s": 1038.7958240509033,
494
+ "steps_per_sec": 0.5005796018434118,
495
+ "vram_gb": 15.428668416
496
+ },
497
+ {
498
+ "step": 1240,
499
+ "loss": 0.47686392813920975,
500
+ "ema_loss": 0.5290220321572775,
501
+ "lr": 8.305237188676093e-05,
502
+ "elapsed_s": 1056.3251626491547,
503
+ "steps_per_sec": 0.5017394441980485,
504
+ "vram_gb": 15.428668416
505
+ },
506
+ {
507
+ "step": 1250,
508
+ "loss": 0.3735135644674301,
509
+ "ema_loss": 0.49133690958435916,
510
+ "lr": 8.097597727416727e-05,
511
+ "elapsed_s": 1073.8807201385498,
512
+ "steps_per_sec": 0.5028491431807532,
513
+ "vram_gb": 15.428668416
514
+ },
515
+ {
516
+ "step": 1260,
517
+ "loss": 0.45909684151411057,
518
+ "ema_loss": 0.48467543677400826,
519
+ "lr": 7.893153468540834e-05,
520
+ "elapsed_s": 1091.462209224701,
521
+ "steps_per_sec": 0.5039111710433675,
522
+ "vram_gb": 15.428668416
523
+ },
524
+ {
525
+ "step": 1270,
526
+ "loss": 0.3922751322388649,
527
+ "ema_loss": 0.45441256367994354,
528
+ "lr": 7.691982156574558e-05,
529
+ "elapsed_s": 1109.0576066970825,
530
+ "steps_per_sec": 0.5049331942889357,
531
+ "vram_gb": 15.428668416
532
+ },
533
+ {
534
+ "step": 1280,
535
+ "loss": 0.37874237447977066,
536
+ "ema_loss": 0.4349958523394238,
537
+ "lr": 7.494160291432407e-05,
538
+ "elapsed_s": 1126.652606010437,
539
+ "steps_per_sec": 0.5059234736237052,
540
+ "vram_gb": 15.428668416
541
+ },
542
+ {
543
+ "step": 1290,
544
+ "loss": 0.35153336822986603,
545
+ "ema_loss": 0.4098208888180574,
546
+ "lr": 7.299763099326428e-05,
547
+ "elapsed_s": 1144.2501864433289,
548
+ "steps_per_sec": 0.5068821546822843,
549
+ "vram_gb": 15.428668416
550
+ },
551
+ {
552
+ "step": 1300,
553
+ "loss": 0.4052524045109749,
554
+ "ema_loss": 0.3939671492847197,
555
+ "lr": 7.108864504159727e-05,
556
+ "elapsed_s": 1161.8791389465332,
557
+ "steps_per_sec": 0.5077980834865048,
558
+ "vram_gb": 15.428668416
559
+ },
560
+ {
561
+ "step": 1310,
562
+ "loss": 0.32005714252591133,
563
+ "ema_loss": 0.3771534058817519,
564
+ "lr": 6.921537099415279e-05,
565
+ "elapsed_s": 1179.5162498950958,
566
+ "steps_per_sec": 0.5086831148391241,
567
+ "vram_gb": 15.428668416
568
+ },
569
+ {
570
+ "step": 1320,
571
+ "loss": 0.2465466745197773,
572
+ "ema_loss": 0.35445263362599905,
573
+ "lr": 6.737852120550606e-05,
574
+ "elapsed_s": 1197.1513810157776,
575
+ "steps_per_sec": 0.5095429113421043,
576
+ "vram_gb": 15.428668416
577
+ },
578
+ {
579
+ "step": 1330,
580
+ "loss": 0.3049493618309498,
581
+ "ema_loss": 0.3454900266396032,
582
+ "lr": 6.55787941790888e-05,
583
+ "elapsed_s": 1214.7954704761505,
584
+ "steps_per_sec": 0.510373980697331,
585
+ "vram_gb": 15.428668416
586
+ },
587
+ {
588
+ "step": 1340,
589
+ "loss": 0.3119439110159874,
590
+ "ema_loss": 0.3349345600271566,
591
+ "lr": 6.381687430156783e-05,
592
+ "elapsed_s": 1232.4373080730438,
593
+ "steps_per_sec": 0.511182188232378,
594
+ "vram_gb": 15.428668416
595
+ },
596
+ {
597
+ "step": 1350,
598
+ "loss": 0.2921401411294937,
599
+ "ema_loss": 0.31941364171661696,
600
+ "lr": 6.209343158259153e-05,
601
+ "elapsed_s": 1250.084626674652,
602
+ "steps_per_sec": 0.511965339260641,
603
+ "vram_gb": 15.428668416
604
+ },
605
+ {
606
+ "step": 1360,
607
+ "loss": 0.24287717416882515,
608
+ "ema_loss": 0.3099492300999808,
609
+ "lr": 6.0409121400003627e-05,
610
+ "elapsed_s": 1267.7480945587158,
611
+ "steps_per_sec": 0.5127201553604033,
612
+ "vram_gb": 15.428668416
613
+ },
614
+ {
615
+ "step": 1370,
616
+ "loss": 0.22612635791301727,
617
+ "ema_loss": 0.2947590519243712,
618
+ "lr": 5.876458425062101e-05,
619
+ "elapsed_s": 1285.4061183929443,
620
+ "steps_per_sec": 0.513456401487456,
621
+ "vram_gb": 15.428668416
622
+ },
623
+ {
624
+ "step": 1380,
625
+ "loss": 0.22580422461032867,
626
+ "ema_loss": 0.2781091778342552,
627
+ "lr": 5.7160445506670415e-05,
628
+ "elapsed_s": 1303.058067560196,
629
+ "steps_per_sec": 0.514175090642343,
630
+ "vram_gb": 15.428668416
631
+ },
632
+ {
633
+ "step": 1390,
634
+ "loss": 0.2514393739402294,
635
+ "ema_loss": 0.26574903508551445,
636
+ "lr": 5.559731517797666e-05,
637
+ "elapsed_s": 1320.7140231132507,
638
+ "steps_per_sec": 0.5148730066461105,
639
+ "vram_gb": 15.428668416
640
+ },
641
+ {
642
+ "step": 1400,
643
+ "loss": 0.24877634644508362,
644
+ "ema_loss": 0.2662288530131991,
645
+ "lr": 5.4075787679992615e-05,
646
+ "elapsed_s": 1380.0000896453857,
647
+ "steps_per_sec": 0.49999996751978987,
648
+ "vram_gb": 15.428668416
649
+ },
650
+ {
651
+ "step": 1410,
652
+ "loss": 0.29013000428676605,
653
+ "ema_loss": 0.2518670179663263,
654
+ "lr": 5.259644160775935e-05,
655
+ "elapsed_s": 1397.4809682369232,
656
+ "steps_per_sec": 0.5009012758743523,
657
+ "vram_gb": 15.428668416
658
+ },
659
+ {
660
+ "step": 1420,
661
+ "loss": 0.21814599260687828,
662
+ "ema_loss": 0.23880236520769688,
663
+ "lr": 5.1159839515882365e-05,
664
+ "elapsed_s": 1415.0194308757782,
665
+ "steps_per_sec": 0.5017598942514659,
666
+ "vram_gb": 15.428668416
667
+ },
668
+ {
669
+ "step": 1430,
670
+ "loss": 0.20425041019916534,
671
+ "ema_loss": 0.2289297893300412,
672
+ "lr": 4.9766527704607636e-05,
673
+ "elapsed_s": 1432.520180940628,
674
+ "steps_per_sec": 0.5026107203091759,
675
+ "vram_gb": 15.428668416
676
+ },
677
+ {
678
+ "step": 1440,
679
+ "loss": 0.21563759073615074,
680
+ "ema_loss": 0.22410501963253138,
681
+ "lr": 4.8417036012078574e-05,
682
+ "elapsed_s": 1450.0505871772766,
683
+ "steps_per_sec": 0.5034307123181445,
684
+ "vram_gb": 15.428668416
685
+ },
686
+ {
687
+ "step": 1450,
688
+ "loss": 0.21186616644263268,
689
+ "ema_loss": 0.2213783268398314,
690
+ "lr": 4.711187761285319e-05,
691
+ "elapsed_s": 1467.6139550209045,
692
+ "steps_per_sec": 0.5042197898625592,
693
+ "vram_gb": 15.428668416
694
+ },
695
+ {
696
+ "step": 1460,
697
+ "loss": 0.17390306666493416,
698
+ "ema_loss": 0.20964586315877706,
699
+ "lr": 4.5851548822757895e-05,
700
+ "elapsed_s": 1485.1958084106445,
701
+ "steps_per_sec": 0.5049839191255185,
702
+ "vram_gb": 15.428668416
703
+ },
704
+ {
705
+ "step": 1470,
706
+ "loss": 0.19438401982188225,
707
+ "ema_loss": 0.20803681597852408,
708
+ "lr": 4.463652891015232e-05,
709
+ "elapsed_s": 1502.7898688316345,
710
+ "steps_per_sec": 0.5057260604177967,
711
+ "vram_gb": 15.428668416
712
+ },
713
+ {
714
+ "step": 1480,
715
+ "loss": 0.18505235388875008,
716
+ "ema_loss": 0.20459605664240096,
717
+ "lr": 4.346727991367676e-05,
718
+ "elapsed_s": 1520.390971660614,
719
+ "steps_per_sec": 0.5064486795518025,
720
+ "vram_gb": 15.428668416
721
+ },
722
+ {
723
+ "step": 1490,
724
+ "loss": 0.1997264064848423,
725
+ "ema_loss": 0.19819135992439754,
726
+ "lr": 4.234424646655138e-05,
727
+ "elapsed_s": 1538.0068547725677,
728
+ "steps_per_sec": 0.5071498853074632,
729
+ "vram_gb": 15.428668416
730
+ },
731
+ {
732
+ "step": 1500,
733
+ "loss": 0.16755592823028564,
734
+ "ema_loss": 0.1918936216767847,
735
+ "lr": 4.1267855627494735e-05,
736
+ "elapsed_s": 1555.638013601303,
737
+ "steps_per_sec": 0.5078302234149894,
738
+ "vram_gb": 15.428668416
739
+ },
740
+ {
741
+ "step": 1510,
742
+ "loss": 0.17541416734457016,
743
+ "ema_loss": 0.18443936061329902,
744
+ "lr": 4.023851671832496e-05,
745
+ "elapsed_s": 1573.2746274471283,
746
+ "steps_per_sec": 0.5084935497231776,
747
+ "vram_gb": 15.428668416
748
+ },
749
+ {
750
+ "step": 1520,
751
+ "loss": 0.14539901167154312,
752
+ "ema_loss": 0.17578462170763828,
753
+ "lr": 3.925662116830581e-05,
754
+ "elapsed_s": 1590.9256908893585,
755
+ "steps_per_sec": 0.5091375446625632,
756
+ "vram_gb": 15.428668416
757
+ },
758
+ {
759
+ "step": 1530,
760
+ "loss": 0.1841065064072609,
761
+ "ema_loss": 0.17096551335138493,
762
+ "lr": 3.8322542365297113e-05,
763
+ "elapsed_s": 1608.5824146270752,
764
+ "steps_per_sec": 0.5097656125938094,
765
+ "vram_gb": 15.428668416
766
+ },
767
+ {
768
+ "step": 1540,
769
+ "loss": 0.15982552990317345,
770
+ "ema_loss": 0.16729554632520202,
771
+ "lr": 3.743663551376553e-05,
772
+ "elapsed_s": 1626.2370433807373,
773
+ "steps_per_sec": 0.5103806996515938,
774
+ "vram_gb": 15.428668416
775
+ },
776
+ {
777
+ "step": 1550,
778
+ "loss": 0.17187346518039703,
779
+ "ema_loss": 0.1643358000447934,
780
+ "lr": 3.6599237499710016e-05,
781
+ "elapsed_s": 1643.8927867412567,
782
+ "steps_per_sec": 0.5109822287529833,
783
+ "vram_gb": 15.428668416
784
+ },
785
+ {
786
+ "step": 1560,
787
+ "loss": 0.1427095215767622,
788
+ "ema_loss": 0.15744941248457628,
789
+ "lr": 3.581066676255348e-05,
790
+ "elapsed_s": 1661.5445902347565,
791
+ "steps_per_sec": 0.5115721871056769,
792
+ "vram_gb": 15.428668416
793
+ },
794
+ {
795
+ "step": 1570,
796
+ "loss": 0.12492887675762177,
797
+ "ema_loss": 0.15700561553855466,
798
+ "lr": 3.5071223174048926e-05,
799
+ "elapsed_s": 1679.1984930038452,
800
+ "steps_per_sec": 0.5121491018382129,
801
+ "vram_gb": 15.428668416
802
+ },
803
+ {
804
+ "step": 1580,
805
+ "loss": 0.14507505670189857,
806
+ "ema_loss": 0.15353575625923596,
807
+ "lr": 3.438118792424645e-05,
808
+ "elapsed_s": 1696.8557302951813,
809
+ "steps_per_sec": 0.5127130046870023,
810
+ "vram_gb": 15.428668416
811
+ },
812
+ {
813
+ "step": 1590,
814
+ "loss": 0.14994086883962154,
815
+ "ema_loss": 0.15015913800263153,
816
+ "lr": 3.374082341456457e-05,
817
+ "elapsed_s": 1714.52734375,
818
+ "steps_per_sec": 0.5132609889296202,
819
+ "vram_gb": 15.428668416
820
+ },
821
+ {
822
+ "step": 1600,
823
+ "loss": 0.14491559006273746,
824
+ "ema_loss": 0.1493680333619406,
825
+ "lr": 3.315037315800602e-05,
826
+ "elapsed_s": 1776.030238866806,
827
+ "steps_per_sec": 0.501117593903054,
828
+ "vram_gb": 15.428668416
829
+ },
830
+ {
831
+ "step": 1610,
832
+ "loss": 0.13732506521046162,
833
+ "ema_loss": 0.14832111416366284,
834
+ "lr": 3.2610061686556506e-05,
835
+ "elapsed_s": 1793.507203578949,
836
+ "steps_per_sec": 0.5018100837309419,
837
+ "vram_gb": 15.428668416
838
+ },
839
+ {
840
+ "step": 1620,
841
+ "loss": 0.1437094621360302,
842
+ "ema_loss": 0.14366319953712975,
843
+ "lr": 3.2120094465801376e-05,
844
+ "elapsed_s": 1811.0531499385834,
845
+ "steps_per_sec": 0.5024700683306065,
846
+ "vram_gb": 15.428668416
847
+ },
848
+ {
849
+ "step": 1630,
850
+ "loss": 0.1201076190918684,
851
+ "ema_loss": 0.1377247703954839,
852
+ "lr": 3.168065781679277e-05,
853
+ "elapsed_s": 1828.5478429794312,
854
+ "steps_per_sec": 0.503131489576425,
855
+ "vram_gb": 15.428668416
856
+ },
857
+ {
858
+ "step": 1640,
859
+ "loss": 0.13061938621103764,
860
+ "ema_loss": 0.13602811819866328,
861
+ "lr": 3.129191884519671e-05,
862
+ "elapsed_s": 1846.0855515003204,
863
+ "steps_per_sec": 0.5037686358815742,
864
+ "vram_gb": 15.428668416
865
+ },
866
+ {
867
+ "step": 1650,
868
+ "loss": 0.12817695923149586,
869
+ "ema_loss": 0.1344849350545901,
870
+ "lr": 3.095402537774757e-05,
871
+ "elapsed_s": 1863.6434466838837,
872
+ "steps_per_sec": 0.5043883268940796,
873
+ "vram_gb": 15.428668416
874
+ },
875
+ {
876
+ "step": 1660,
877
+ "loss": 0.13230768591165543,
878
+ "ema_loss": 0.13204115287728813,
879
+ "lr": 3.066710590603351e-05,
880
+ "elapsed_s": 1881.2273790836334,
881
+ "steps_per_sec": 0.5049894609033149,
882
+ "vram_gb": 15.428668416
883
+ },
884
+ {
885
+ "step": 1670,
886
+ "loss": 0.11046527698636055,
887
+ "ema_loss": 0.1296894053225092,
888
+ "lr": 3.0431269537634746e-05,
889
+ "elapsed_s": 1898.828557729721,
890
+ "steps_per_sec": 0.50557486935408,
891
+ "vram_gb": 15.428668416
892
+ },
893
+ {
894
+ "step": 1680,
895
+ "loss": 0.13638370111584663,
896
+ "ema_loss": 0.12728797289501487,
897
+ "lr": 3.0246605954632904e-05,
898
+ "elapsed_s": 1916.423189163208,
899
+ "steps_per_sec": 0.5061512537966854,
900
+ "vram_gb": 15.428668416
901
+ },
902
+ {
903
+ "step": 1690,
904
+ "loss": 0.12420566566288471,
905
+ "ema_loss": 0.12502314340227097,
906
+ "lr": 3.0113185379507383e-05,
907
+ "elapsed_s": 1934.0418214797974,
908
+ "steps_per_sec": 0.506710862772435,
909
+ "vram_gb": 15.428668416
910
+ },
911
+ {
912
+ "step": 1700,
913
+ "loss": 0.10359455086290836,
914
+ "ema_loss": 0.12350706497894226,
915
+ "lr": 3.0031058548431695e-05,
916
+ "elapsed_s": 1951.6798777580261,
917
+ "steps_per_sec": 0.5072553195236368,
918
+ "vram_gb": 15.428668416
919
+ },
920
+ {
921
+ "step": 1710,
922
+ "loss": 0.11198867857456207,
923
+ "ema_loss": 0.12169409054894889,
924
+ "lr": 3.0000256691979898e-05,
925
+ "elapsed_s": 1969.3243565559387,
926
+ "steps_per_sec": 0.5077883674525075,
927
+ "vram_gb": 15.428668416
928
+ }
929
+ ]
930
+ }