Qwen3.5-9B General Game Stage2 — checkpoint 4000
Stage2 optimizer step 4000, seed 1234. Source run: gg-q35-stage2-256g-s1234-resume2200-20260926a.
Use revision="checkpoint-4000" for both model and processor/tokenizer.
Decode with skip_special_tokens=False to retain action/thought delimiters.
This inference export omits optimizer, RNG and dataloader resume state.
Tensor integrity and image/text smoke checks are not game-performance evaluation.
main contains checkpoint-4000.
| Step | Branch | Commit |
|---|---|---|
| 100 | checkpoint-100 | 0848a3c3ee47b7dc375a6b4904966a23e83c1a24 |
| 200 | checkpoint-200 | 755b7543fcda8267a25d8dd31cbb65c04d6dec4d |
| 300 | checkpoint-300 | 14e21a8b927cce3175f7ecfb1b5f36787fa39453 |
| 400 | checkpoint-400 | cd11326d00fbc5ea09e633baee6ca1f7adba1a38 |
| 500 | checkpoint-500 | 4dbe59ef45723c425f76206214a82800cb2e32b2 |
| 600 | checkpoint-600 | 14ea9b3d5f712a94077ad751b12fec5498ad9134 |
| 700 | checkpoint-700 | 6af4e0e6dbfa3cda004a8acb00cd45739fe53a83 |
| 800 | checkpoint-800 | 2b6c4b2170109f6e9a68c5bb44e1e65364b013ab |
| 900 | checkpoint-900 | 1c87ffdf8e0af8d94c7c8c995eafc80cf76e9628 |
| 1000 | checkpoint-1000 | 844824999733d869abf848dc27150fef67444267 |
| 1100 | checkpoint-1100 | a7dae158349318f0dc21535de8ad6cdc9bd2afac |
| 1200 | checkpoint-1200 | 821b8a6263db157e7d5cf8ec9f111163ad76605a |
| 1300 | checkpoint-1300 | bc21a1417bfe68dbbe32f05d90acea89efb030bc |
| 1400 | checkpoint-1400 | 8008df7b85a8dd8e9d35124ed6df557272972739 |
| 1500 | checkpoint-1500 | 9cb9ce62448de1ab5b21e57e292e5437735d0373 |
| 1600 | checkpoint-1600 | 86305c71b96d12575b5151bf579e462883869d28 |
| 1700 | checkpoint-1700 | 83201e4e1bc8aae8af8589e5d625e390616eb590 |
| 1800 | checkpoint-1800 | aa01b20c82c8736d4833f228d6e11aab60773b68 |
| 1900 | checkpoint-1900 | 6e8635ddd2f7eaa02c4ce9218c83460f4393c076 |
| 2000 | checkpoint-2000 | 5d1290e12e5bce01e03dcacc5b5b03ce9f25ab48 |
| 2100 | checkpoint-2100 | acfec3f7781bfc29d5b0b66facdcfda2e16f37b6 |
| 2200 | checkpoint-2200 | 0b41a8b451893a990d313574f29ce0a88b7a784d |
| 2300 | checkpoint-2300 | 9e022e65983311fb6b1e5b16df9ed4b94dee85d5 |
| 2400 | checkpoint-2400 | 72123e3d4a64485e58ae4bc7842a9fd1e1d9b1e4 |
| 2500 | checkpoint-2500 | c0625102a4829f3703423794ff93fedbb999b6d7 |
| 2600 | checkpoint-2600 | 05529872e200673af26e8091a9c76bfdc7e17ffc |
| 2700 | checkpoint-2700 | 4be32575bc8838f8cb6fdfa36261f8aed56133a7 |
| 2800 | checkpoint-2800 | 58e70f0fd8f5a1347107b1c41b150717dada6dec |
| 2900 | checkpoint-2900 | 277d1d5a4c401fa136e3f5fea6d9c175ba2863ec |
| 3000 | checkpoint-3000 | ac5022d18f41d57176b543eec22f1c895b585852 |
| 3100 | checkpoint-3100 | 69ca51c0d96d5b9012886395514454139c5aeb3b |
| 3200 | checkpoint-3200 | 331971835a5d69233cdb8f2a860775bb58524b8f |
| 3300 | checkpoint-3300 | 31a652fa8b96fd6475cfe92f0bb52e377d35f39a |
| 3400 | checkpoint-3400 | a9160cf401035ed1ce54d0dd4cccf455aeb13376 |
| 3500 | checkpoint-3500 | d2f8d4fc3955a95857cf44aee2ca5de2377b1ce8 |
| 3600 | checkpoint-3600 | c2b4ed46a144f686c55bc51de375c6c85f05cb5e |
| 3700 | checkpoint-3700 | a1209f214984a6cf765417f203568c10e0ade2b0 |
| 3800 | checkpoint-3800 | 41bfed8a950f3b9f6d09f441120b2cce5adee853 |
| 3900 | checkpoint-3900 | 37f1044b9bff79531cf5c2a87e5098c0ab53fde6 |
| 4000 | checkpoint-4000 | 9aa40a5bb589261354af07c6be1ae410e404e07b |