verl-project/verl

[RFC][rollout] Should `actor_rollout_ref.rollout.mode` be removed since it should always be `"async"` now? 12 days ago

good first issue call for contribution code quality

verl-project/verl

22.3K

verl/HybridFlow: A Flexible and Efficient RL Post-Training Framework

Python

[FR] Better metric reduction utility 14 days ago

enhancement good first issue feature request

verl-project/verl

22.3K

verl/HybridFlow: A Flexible and Efficient RL Post-Training Framework

Python

[Bug][CI] FDSP2 test in `model_rmpad` job seems unstable about 1 month ago

bug good first issue call for contribution

verl-project/verl

22.3K

verl/HybridFlow: A Flexible and Efficient RL Post-Training Framework

Python

`process_validation_metrics()` crashes on None-filled sparse `reward_extra_info` keys (metrics emitted for only some samples) about 1 month ago

bug help wanted

verl-project/verl

22.3K

verl/HybridFlow: A Flexible and Efficient RL Post-Training Framework

Python

Additional memory optimization features about 2 months ago

enhancement good first issue call for contribution

verl-project/verl

22.3K

verl/HybridFlow: A Flexible and Efficient RL Post-Training Framework

Python

Fused CE loss integration about 2 months ago

help wanted call for contribution

verl-project/verl

22.3K

verl/HybridFlow: A Flexible and Efficient RL Post-Training Framework

Python

[FR][Dev] Unify (& Complete) the docstring format about 2 months ago

documentation enhancement good first issue call for contribution

verl-project/verl

22.3K

verl/HybridFlow: A Flexible and Efficient RL Post-Training Framework

Python

maybe_filter_out_long_prompts过滤prompts的不支持图片路径，他读取的是bytes图像数据，或者Image.Image 2 months ago

help wanted

verl-project/verl

22.3K

verl/HybridFlow: A Flexible and Efficient RL Post-Training Framework

Python

希望可以增加单独检查权重转换逻辑 3 months ago

help wanted

verl-project/verl

22.3K

verl/HybridFlow: A Flexible and Efficient RL Post-Training Framework

Python

Migrate volcengine/verl references to verl-project/verl 4 months ago

good first issue call for contribution feature request

verl-project/verl

22.3K

verl/HybridFlow: A Flexible and Efficient RL Post-Training Framework

Python

Support Generative Reward Model (GenRM) 5 months ago

enhancement good first issue

verl-project/verl

22.3K

verl/HybridFlow: A Flexible and Efficient RL Post-Training Framework

Python

Open Issues Need Help