What's Changed
- Fix two latent bugs: dr_grpo n=1 guard and masked_normalize broadcast by @hijkzzz in #1250
- fix: allow eval_dataset with MultiTurnAgentExecutor (#1242) by @codewithyug06 in #1251
- Fix Qwen3.5 ZeRO-3 frozen weights from set_z3_leaf_modules hybrid detection by @Functionhx in #1261
- fix(ppo): process terminal synchronous rollout batches by @ai-yang in #1300
- fix(ppo): preserve ragged batches in balance_experiences by @ai-yang in #1299
- [pre-commit.ci] pre-commit suggestions by @pre-commit-ci[bot] in #1259
- chore: upgrade vllm deepspeed and transformers by @hijkzzz in #1304
- docs: add AGENTS.md and simplicity-first skill for AI assistants by @hijkzzz in #1306
- Cleanup/simplicity tier1 2 by @hijkzzz in #1307
New Contributors
- @codewithyug06 made their first contribution in #1251
- @Functionhx made their first contribution in #1261
- @ai-yang made their first contribution in #1300
Full Changelog: v0.10.4...v0.11.0
Contributors
hijkzzz, ai-yang, and 3 other contributors