Edit

Two local torch training runs for attestation and dataset-filter checks.

tape_validation/torch_20260906_005141/next_token_mse vs. estimated ISO FLOPs

2 matching runs · 6ND; missing tokens estimated at 20× parameters
All 2 filtered runs
Run Scale (B) ISO FLOPs Training datatape_validation/torch_20260906_005141/next_token_mse
tape_validation/torch_20260906_005141_baking_window8
No description yet
<0.001~3.2e40.02
tape_validation/torch_20260906_005141_astronomy_window8
No description yet
<0.001~1.4e40.02
2 models