TicTacChess-CK / README.md
LorMolf's picture
Update tic_tac_chess latest-best pointer
4beeb15 verified
|
Raw
History Blame Contribute Delete
1.83 kB
---
license: apache-2.0
tags:
- muzero
- efficientzero
- board-games
- checkpoint
---
# TicTacChess EfficientZero Checkpoint
This repository stores retained CRPT EfficientZero checkpoints for `tic_tac_chess`.
## Latest Upload
- Checkpoint: `checkpoints/envstep_100000.pth.tar`
- Source path: `models/simplified5_constant_lr_fixed_20260519/round-01/tic_tac_chess/attempt-01_260519_093506/ckpt/envstep_100000.pth.tar`
- Uploaded at: `2026-05-21T09:56:43Z`
- Selection: latest retained env-step checkpoint from the previous simplified5 training run per user instruction.
Metadata and resolved runtime config snapshots are under `metadata/`.
<!-- CRPT_LATEST_BEST_START -->
## Latest Best Checkpoint
The current latest best checkpoint pointer for this repository is:
`checkpoints/main5_bot_only_stabilize_overnight_20260524/ckpt_last.pth.tar`
Metadata:
- Game: `tic_tac_chess`
- Source run: `main5_bot_only_stabilize_overnight_20260524`
- Checkpoint role: final bot-only stabilization `ckpt_last.pth.tar`
- Local source at upload: `/mnt/nvme/home/molfetta/molfetta-reasoning/models/main5_bot_only_stabilize_overnight_20260524/tic_tac_chess/attempt-01_260524_201957/ckpt/ckpt_last.pth.tar`
- Checkpoint SHA256 at upload: `53046fdb3c54669f8926e426f4e7005d380ffb5dbb6d195a1bb0fd63158476f2`
- Final env steps: `200050`
- Validation note: final watchdog gate on 2026-05-24/2026-05-25 reported the last three non-initial fixed-bot evals at `reward_mean=1.0`; one-sided fixed-bot gate by experiment design; seat_swap=false and seat1_reward_mean=1.0.
- Upload manifest: `metadata/main5_bot_only_stabilize_overnight_20260524_latest_best_manifest.json`
Older checkpoint files in this repository are preserved; this section is the canonical pointer to use when a consumer needs the latest best checkpoint.
<!-- CRPT_LATEST_BEST_END -->