Skip to content

Remove envtorch dir nesting and add base env code#2

Merged
Darktex merged 1 commit into
mainfrom
env_code
Oct 6, 2025
Merged

Remove envtorch dir nesting and add base env code#2
Darktex merged 1 commit into
mainfrom
env_code

Conversation

@pankit-eng
Copy link
Copy Markdown
Contributor

No description provided.

@pankit-eng pankit-eng requested a review from Darktex October 6, 2025 18:54
@meta-cla meta-cla Bot added the CLA Signed This label is managed by the Meta Open Source bot. label Oct 6, 2025
@Darktex Darktex merged commit a43a5cc into main Oct 6, 2025
1 check passed
burtenshaw added a commit that referenced this pull request Nov 1, 2025
pankit-eng pushed a commit that referenced this pull request Nov 3, 2025
FIX: Create robust client parser for reset/step inconsistency
rycerzes referenced this pull request in rycerzes/OpenEnv Nov 19, 2025
Remove envtorch dir nesting and add base env code
rycerzes referenced this pull request in rycerzes/OpenEnv Nov 19, 2025
rycerzes referenced this pull request in rycerzes/OpenEnv Nov 19, 2025
FIX: Create robust client parser for reset/step inconsistency
EchoRaven pushed a commit to BillChan226/openenv-gen that referenced this pull request Jan 5, 2026
akashkathole7 added a commit to akashkathole7/OpenEnv that referenced this pull request Apr 20, 2026
Previous Colab run failed Cell 9 (plot) with FileNotFoundError because
Cell 7 crashed during model load or LM-sampling eval before any curve was
written.

- Write an empty curves JSON stub to output_dir the moment it's created,
  before model load. Plot cell now always has a file to open.
- Swap scaffold eval from LM-based (`_stub_eval_policy_factory` → 100
  slow generate() calls per eval) to heuristic (`evaluate_heuristic` uses
  prompted_policy — no LM calls, deterministic, completes in <1s for
  20 seeds). Scaffold curves are flat by design.
- Wrap model load in best-effort try/except. If it fails, continue in
  heuristic-only mode: curves still produced, checkpoints skipped with a
  clear warning. Done-gate huggingface#2 (5 curves × ≥4 eval points) satisfied
  regardless of GPU availability.
- Keep `_stub_eval_policy_factory` and `evaluate_checkpoint` for the
  post-swap real training loop.

Local test (no transformers installed):
  scaffold ran 20 steps, wrote 5 eval points, exited 0 in 0.1s.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
@Darktex Darktex mentioned this pull request Apr 25, 2026
12 tasks
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

CLA Signed This label is managed by the Meta Open Source bot.

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants