Update latest best checkpoint pointer for main5_bot_mode_until_perfect_resume_20260524
Browse files
README.md
CHANGED
|
@@ -5,3 +5,23 @@ Accepted fixed-bot initialization checkpoint for the selected-five MuZero/Effici
|
|
| 5 |
- Source run: `main5_hf_bot_mode_recovery_50k_20260522`
|
| 6 |
- Checkpoint: `checkpoints/envstep_150000.pth.tar`
|
| 7 |
- Uploaded: `2026-05-23T11:46:20+00:00`
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 5 |
- Source run: `main5_hf_bot_mode_recovery_50k_20260522`
|
| 6 |
- Checkpoint: `checkpoints/envstep_150000.pth.tar`
|
| 7 |
- Uploaded: `2026-05-23T11:46:20+00:00`
|
| 8 |
+
|
| 9 |
+
<!-- CRPT_LATEST_BEST_START -->
|
| 10 |
+
## Latest Best Checkpoint
|
| 11 |
+
|
| 12 |
+
The current latest best checkpoint pointer for this repository is:
|
| 13 |
+
|
| 14 |
+
`checkpoints/main5_bot_mode_until_perfect_resume_20260524/ckpt_best.pth.tar`
|
| 15 |
+
|
| 16 |
+
Metadata:
|
| 17 |
+
- Game: `simplified__first_attack`
|
| 18 |
+
- Source run: `main5_bot_mode_until_perfect_resume_20260524`
|
| 19 |
+
- Checkpoint role: current training-selected `ckpt_best.pth.tar` from the active bot-mode resume
|
| 20 |
+
- Local source at upload: `/home/molfetta/combinatorial_reasoning_post_training/models/main5_bot_mode_until_perfect_resume_20260524/simplified__first_attack/attempt-01_260524_122727/ckpt/ckpt_best.pth.tar`
|
| 21 |
+
- Checkpoint SHA256 at upload: `d6408844d344bb249d1d8719ddc994600422ef29b7afafc14434be02546818a4`
|
| 22 |
+
- Source note: resume ckpt_best loaded from the best available bot-lineage First Attack checkpoint
|
| 23 |
+
- Evaluation note: Active bot-mode resume pointer. Recent collector summaries were still weak or negative, so this records the current training-selected ckpt_best rather than a solved checkpoint.
|
| 24 |
+
- Upload manifest: `metadata/main5_bot_mode_until_perfect_resume_20260524_latest_best_manifest.json`
|
| 25 |
+
|
| 26 |
+
Older checkpoint files in this repository are preserved; this section is the canonical pointer to use when a consumer needs the latest best checkpoint.
|
| 27 |
+
<!-- CRPT_LATEST_BEST_END -->
|