-
v19b training state (update 1260/4000)
-
v19b best model — stage 11 (Hard, 15 opponents), 90% win rate, update 1260
-
v19b training state
-
v19b best model (update 580, reward 1.29, best sweep params + potential shaping alpha=1)
-
Update best_model.pt to v18b (obs_dim=96, 512-512-256, reward=0.584, loss penalty)
-
Update best_model.pt to v17 (obs_dim=80, best_reward=0.535)
-
Update default best_model.pt to v16
-
Upload v16 best model (wilderness + boat attack fix, reward 0.516)
-
Upload v15 best model (wilderness + optimized BFS, reward 0.555)
-
Update README for v13b
-
v13b (update 1550): normalized elim + winner bonus, vf=0.5, best generalization
-
v12a: 100% win rate on Easy/2, normalized elimination reward
-
Upload README.md with huggingface_hub
-
Upload training_log.json with huggingface_hub
-
Upload best_model.pt with huggingface_hub
-
Upload README.md with huggingface_hub
-
Upload training_log.json with huggingface_hub
-
Upload best_model.pt with huggingface_hub
-
Upload README.md with huggingface_hub
-
Upload training_log.json with huggingface_hub
-
Upload best_model.pt with huggingface_hub
-
Upload README.md with huggingface_hub
-
Upload training_log.json with huggingface_hub
-
Upload README.md with huggingface_hub
-
Upload training_log.json with huggingface_hub
-
Upload best_model.pt with huggingface_hub
-
Upload README.md with huggingface_hub
-
Upload training_log.json with huggingface_hub
-
Upload best_model.pt with huggingface_hub
-
initial commit