Can't Stop project models
Public trained models from the CS 441 Can't Stop final project. These are project-specific policies for the repository's two-player simulator, not generic Transformers models.
Which file should I use?
| Model | File | Use when |
|---|---|---|
| Double DQN | best.pt |
You want the retained DDQN reference |
| PPO | ppo/best.pt |
You want the retained PPO-451 policy |
| Ridge | ridge/ridge_domain.joblib |
You want the lightweight classical-ML baseline |
The project repository contains the loading and evaluation code:
https://github.com/BrianHuang2003/CS441_AML.
The local-vs-download instructions are in docs/MODEL_GUIDE.md.
Provenance
- DDQN source:
ddqn_mc_counterfactual_seed447/best.pt, seed 447 - PPO source:
ppo_mc_focused_seed451/top_1.pt, seed 451 - Ridge source:
ridge_multi_policy_seed642/ridge_domain.joblib, seed 642 - Fixed dice pairing for the retained learned-policy comparisons: Glenn--Aloi
- Complete per-file provenance and checksums:
MODEL_FILES.jsonandSHA256SUMS
Use scripts/evaluate_ml.py for .pt checkpoints and the project's classical
ML loader for the .joblib model. Preserve the model name, seed, source commit,
and evaluation configuration when reporting results.
- Downloads last month
- 17