Commit 74fd646
fix: diagnostic logging (loss scientific notation, grad_norm, advantages) (#227)
loss=0.0000 was misleading: %.4f truncation + symmetric advantages
canceling. Now logs loss in scientific notation, absolute loss per
rollout, gradient norm, and per-rollout advantages.
13 vision loss tests (was 12). New test verifies loss_abs > 0 and
advantages are symmetric with reward variance.
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>1 parent 2355d53 commit 74fd646
384 files changed
Lines changed: 3457 additions & 6 deletions
File tree
- .beads
- .claude/worktrees
- annotated_demos_core4
- annotated_demos
- demos
- custom-clear-chrome-data
- manual/manual
- synthetic_correction
- custom-notepad-hello
- manual/custom-notepad-hello/synthetic_correction
- synthetic_correction
- docs
- design
- eval_results/baseline/notepad
- flywheel_results
- chrome_executor_20260328_170219
- corrections
- phase1_screenshots
- phase3_screenshots
- notepad_20260328_155626
- corrections
- phase1_screenshots
- phase3_screenshots
- notepad_20260328_162152
- corrections
- phase1_screenshots
- phase3_screenshots
- notepad_executor_20260328_163357
- corrections
- phase1_screenshots
- notepad_executor_20260328_164252
- corrections
- phase1_screenshots
- phase3_screenshots
- run_20260327_150058
- corrections
- phase1_screenshots
- phase3_screenshots
- run_20260328_010502
- corrections
- phase1_screenshots
- phase3_screenshots
- run_20260328_100112
- corrections
- run_20260328_100310
- corrections
- phase1_screenshots
- phase3_screenshots
- run_20260328_111454
- corrections
- phase1_screenshots
- run_20260328_121531
- corrections
- phase1_screenshots
- phase3_screenshots
- run_20260328_153905
- corrections
- phase1_screenshots
- phase3_screenshots
- grpo_output
- openadapt_evals/training/standalone
- tests
- waa_recordings_archive/0bf05a7d-b28b-44d2-955a-50b41e24012a-WOS_20260304_163020
- waa_recordings_core4
- waa_recordings_smoke3/0bf05a7d-b28b-44d2-955a-50b41e24012a-WOS
- waa_recordings_smoke4/0bf05a7d-b28b-44d2-955a-50b41e24012a-WOS
- waa_recordings_smoke/0bf05a7d-b28b-44d2-955a-50b41e24012a-WOS
- waa_recordings
- 0a0faba3-5580-44df-965d-f562a99b291c-WOS
- 0bf05a7d-b28b-44d2-955a-50b41e24012a-WOS.bad
- 0bf05a7d-b28b-44d2-955a-50b41e24012a-WOS
- 0e763496-b6bb-4508-a427-fad0b6c3e195-WOS
- 70745df8-f2f5-42bd-8074-fbc10334fcc5-2-WOS
Some content is hidden
Large Commits have some content hidden by default. Use the searchbox below for content that may be hidden.
Binary file not shown.
| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
| 1 | + | |
| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
| 1 | + | |
| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
| 1 | + | |
| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
| 1 | + | |
| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
| 1 | + | |
| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
| 1 | + | |
| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
| 1 | + | |
| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
| 1 | + | |
| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
| 1 | + | |
0 commit comments