v1.9.1

What's Changed

• Fix vLLM server-mode communicator initialization to use the current A… by @sywangyi in #6417
• Fix : queue wait time metric AsyncGRPOTrainer by @AmineDiro in #6489
• Hotfix CI: Temporarily pin liger-kernel < 0.8.1 by @albertvillanova in #6517
• Fix crash when training with Liger kernel on pre-Ampere GPUs by @albertvillanova in #6518
• Fix prepare_deepspeed crash with cpu offload optimizer by @roycho96 in #5916
• Add Gemma4 response template by @qgallouedec in #6535
• fix(grpo): correct DAPO/CISPO/VESPO loss normalization when steps_per_generation != gradient_accumulation_steps by @0xadvait in #6024
• Hotfix CI: Temporarily pin bitsandbytes < 0.50.0 by @albertvillanova in #6544
• Hotfix CI: Xfail NemotronH GRPO/RLOO tests against a broken transformers dev build by @albertvillanova in #6542

Full Changelog: v1.9.0...v1.9.1