Skip to content

feat: TorchArc YAML network architectures#532

Merged
kengz merged 1 commit intomasterfrom
feat/torcharc-yaml
Feb 19, 2026
Merged

feat: TorchArc YAML network architectures#532
kengz merged 1 commit intomasterfrom
feat/torcharc-yaml

Conversation

@kengz
Copy link
Owner

@kengz kengz commented Feb 19, 2026

Summary

  • Replace hardcoded PyTorch network modules with TorchArc YAML-defined architectures across all algorithms (REINFORCE, SARSA, DQN, DDQN+PER, A2C, PPO, SAC)
  • Add YAML spec support (spec_util, torcharc_net.py) and benchmark specs under benchmark_arc/
  • Full benchmark validation: Classic Control, Box2D, MuJoCo (11 envs), Atari (54 games) — results match or exceed original scores
  • New SAC Atari benchmarks (48 games) with discrete action support
  • Pre-commit hook for conventional commit validation (.githooks/commit-msg)
  • Version bump 5.0.2 → 5.1.0, changelog moved to docs/CHANGELOG.md

Test plan

  • uv run pytest test/agent/net/test_torcharc_net.py — TorchArc net unit tests
  • uv run pytest test/spec/test_yaml_spec.py — YAML spec parsing tests
  • uv run slm-lab slm_lab/spec/benchmark/ppo/ppo_cartpole.yaml ppo_cartpole dev — quick smoke test
  • Verify benchmark plots render in docs/BENCHMARKS.md

🤖 Generated with Claude Code

@kengz kengz force-pushed the feat/torcharc-yaml branch from 57a83de to 59f1b1d Compare February 19, 2026 03:05
Replace hardcoded PyTorch network modules with TorchArc YAML-defined
architectures across all algorithms. Add YAML spec support, benchmark
specs, and validate on Classic Control, Box2D, MuJoCo (11 envs), and
Atari (54 games). Includes SAC Atari benchmarks (48 games), pre-commit
hooks, and version bump to 5.1.0.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
@kengz kengz force-pushed the feat/torcharc-yaml branch from 59f1b1d to e56a414 Compare February 19, 2026 03:36
@kengz kengz merged commit d3128a8 into master Feb 19, 2026
3 checks passed
@kengz kengz deleted the feat/torcharc-yaml branch February 19, 2026 03:46
@github-actions
Copy link

🎉 This PR is included in version 5.1.0 🎉

The release is available on GitHub release

Your semantic-release bot 📦🚀

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant