Stephen Xie
Stephen Xie
You can replace ``` [SLoRA](https://arxiv.org/abs/2311.03285) or [CCoE](https://arxiv.org/abs/2407.11686) ``` with ``` `SLoRA `_ or `CCoE `_ ```
> Hey @StephenXie, Im interested in picking this back up. How functional is this feature? Have any of the known issues been resolved, or new ones come up? And what...
> @StephenXie I'm not sure if this is the kind of test you are looking for: > > > > I have a setting where I do GRPO on MATH,...
> @skepsun I suppose you have this issue after merging this branch with Main locally? This is because in the sharding manager, the dict holds pointers to un-sharded tensors, but...
Facing similar issues here when running Qwen3-32B with vLLM (max_model_len=32k) - would appreciate help!
Update: I resolved this issue by increasing the max_gen_tokens, as it was defaulted to 256. This is not directly related to the issue you are facing, but hopefully it helps!