GoodStartLabs
/

gin-rummy-mdp

Model card Files Files and versions

YAML Metadata Warning:empty or missing yaml metadata in repo card

Check out the documentation for more information.

Gin Rummy MDP - RL Agent Checkpoints

JAX-native Gin Rummy RL agent trained with PPO self-play.

Checkpoints

r24_stage1_final.pkl - Run 24: Surgical fix (reward clipping, tuned hyperparams). ~49-50% win rate vs heuristic.
r24_stage2_final.pkl - Run 24: Stage 2 continued training.
r25_stage1_final.pkl - Run 25: Observation upgrade (gin rating, opponent DW estimate, unseen pool). >52% win rate vs heuristic. Best model.

Architecture

2x1024 MLP with shared trunk
OBS_DIM=284 (Run 25) / OBS_DIM=230 (Run 24)
16 discrete actions
Pure JAX (Flax linen)

Project

GitHub: GoodStartLabs/GinRummyMdp
W&B: good-start-labs/gsl-gin-rummy-mdp

Downloads last month: -; Downloads are not tracked for this model. How to track

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support