diff options
| author | YurenHao0426 <blackhao0426@gmail.com> | 2026-06-13 12:35:36 -0500 |
|---|---|---|
| committer | YurenHao0426 <blackhao0426@gmail.com> | 2026-06-13 12:35:36 -0500 |
| commit | 66e0d8b9fd4d0f7a2231d689c055e26fdf1cf04a (patch) | |
| tree | c29cba61124018755a19b02c9d33e3ad5f2e05cc /research/flossing/report_bundle_20260603/tables/hrm_honly_step26040_eval.csv | |
Curated export for clone-and-run Maze training (2x A6000) + diagnostics.
trm/hrm pretrain.py carry trajectory-augmentation code (backward-compatible).
Heavy artifacts (checkpoints/wandb/npz) gitignored; see PROVENANCE.md.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Diffstat (limited to 'research/flossing/report_bundle_20260603/tables/hrm_honly_step26040_eval.csv')
| -rw-r--r-- | research/flossing/report_bundle_20260603/tables/hrm_honly_step26040_eval.csv | 2 |
1 files changed, 2 insertions, 0 deletions
diff --git a/research/flossing/report_bundle_20260603/tables/hrm_honly_step26040_eval.csv b/research/flossing/report_bundle_20260603/tables/hrm_honly_step26040_eval.csv new file mode 100644 index 0000000..d3e3ab9 --- /dev/null +++ b/research/flossing/report_bundle_20260603/tables/hrm_honly_step26040_eval.csv @@ -0,0 +1,2 @@ +step,accuracy,exact_accuracy,lm_loss,q_halt_accuracy,q_halt_loss,steps +26040,0.85720146,0.62264127,0.35935268,0.99625105,0.024241636,16.0 |