levmckinney

Levmckinney's group workspace

Group: pythia-12b-deduped

1-1

of 1

Timestamps visible

2023-05-05 15:19:47

Using 'GPTNeoXLayer' for transformer_auto_wrap_policy.

2023-05-05 15:20:29

Gradient accumulation steps: 32

2023-05-05 15:20:29

Using 262_144 tokens per training step.

2023-05-05 15:20:29

All processes have completed setup. Starting training.

2023-05-05 20:33:32

Training: 100%|██████████| 8000/8000 [5:13:03<00:00, 2.35s/it]

2023-05-05 20:33:32

Saving lens to /output/EleutherAI/pythia-12b-deduped-1683299166