| # Selected scientific lines; this is a derived compact log, not the full raw worker log. |
| 2026-10-19 03:21:53 [cfg] unit_bf16_d910_c1_lr0.001_acc1_s7_st152580_ev5000_corpc420b_perm_cos_mb8_w8 |
| 2026-10-19 03:21:53 [env] {'torch': '2.11.0+cu128', 'gpu': 'NVIDIA B200', 'tf32_matmul': False, 'f32_mode': 'highest', 'python': '3.12.3'} |
| 2026-10-19 03:22:00 [model] 967.0M parameters |
| 2026-10-19 03:22:00 [data] corpus c420b (c4_20b_tokens.i32) |
| 2026-10-19 03:22:00 [data] 19999.70M training tokens; this cell processes 19999.0M, 1.00 of the corpus |
| 2026-10-19 03:22:00 [ood] 256 sequences from wikitext_train_tokens.pt |
| 2026-10-19 03:22:01 [radius] max over 49 normalizers 30.150; u = R/sqrt(c) = 30.150 |
| 2026-10-19 03:22:01 [sampler] permutation of 9765477 windows; this cell uses 9765120 of them (1.000 passes) |
| 2026-10-19 03:22:01 [sched] warmup 3051 steps to 0.001, cosine to 0.0001 at step 152580 |
| 2026-10-19 03:41:41 [health] steps 1-1000: loss 6.7814 grad-norm mean 1.542 max 3.774 clipped 82% |
| 2026-10-19 03:41:52 [resume] saved at step 1000 |
| 2026-10-19 04:01:50 [health] steps 1001-2000: loss 4.8932 grad-norm mean 0.999 max 2.113 clipped 44% |
| 2026-10-19 04:02:03 [resume] saved at step 2000 |
| 2026-10-19 04:21:58 [health] steps 2001-3000: loss 4.3014 grad-norm mean 0.660 max 3.503 clipped 2% |
| 2026-10-19 04:22:11 [resume] saved at step 3000 |
| 2026-10-19 04:42:06 [health] steps 3001-4000: loss 4.0418 grad-norm mean 0.462 max 1.165 clipped 0% |
| 2026-10-19 04:42:19 [resume] saved at step 4000 |
| 2026-10-19 05:02:36 [eval] first evaluation did not change model state (deb65a149f253fc2 -> deb65a149f253fc2) |
| 2026-10-19 05:02:50 [unit_bf16_d910_c1_lr0.001_acc1_s7_st152580_ev5000_corpc420b_perm_cos_mb8_w8] step 5000 ppl 46.7 ood 106.6 (100m train) |
| 2026-10-19 05:02:50 [health] steps 4001-5000: loss 3.8959 grad-norm mean 0.418 max 1.101 clipped 0% |
| 2026-10-19 05:03:03 [resume] saved at step 5000 |
| 2026-10-19 05:23:02 [health] steps 5001-6000: loss 3.8097 grad-norm mean 0.382 max 0.894 clipped 0% |
| 2026-10-19 05:23:16 [resume] saved at step 6000 |
| 2026-10-19 05:43:20 [health] steps 6001-7000: loss 3.7306 grad-norm mean 0.362 max 0.811 clipped 0% |
| 2026-10-19 05:43:35 [resume] saved at step 7000 |
| 2026-10-19 06:03:37 [health] steps 7001-8000: loss 3.6725 grad-norm mean 0.354 max 1.470 clipped 0% |
| 2026-10-19 06:03:51 [resume] saved at step 8000 |
| 2026-10-19 06:23:45 [health] steps 8001-9000: loss 3.6368 grad-norm mean 0.361 max 25.631 clipped 0% |
| 2026-10-19 06:23:59 [resume] saved at step 9000 |
| 2026-10-19 06:44:19 [unit_bf16_d910_c1_lr0.001_acc1_s7_st152580_ev5000_corpc420b_perm_cos_mb8_w8] step 10000 ppl 36.3 ood 84.5 (201m train) |
| 2026-10-19 06:44:19 [health] steps 9001-10000: loss 3.6035 grad-norm mean 0.340 max 1.862 clipped 0% |
| 2026-10-19 06:44:33 [resume] saved at step 10000 |
| 2026-10-19 06:44:33 [stop] step limit at step 10000; evaluation and resume saved |
| 2026-10-19 06:44:33 [done] unit_bf16_d910_c1_lr0.001_acc1_s7_st152580_ev5000_corpc420b_perm_cos_mb8_w8: STEP-LIMITED at step 10000; evaluation and resume retained; planned horizon remains 152580 |
|
|
