Act offload lora fix (#2928) [skip ci]
* fix activation offloading with lora * update w e2e test * add docs for error
This commit is contained in:
@@ -136,3 +136,7 @@ description: Frequently asked questions
|
||||
> dynamic: false
|
||||
> mode: max-autotune-no-cudagraphs
|
||||
> ```
|
||||
|
||||
**Q: `ValueError("Backward pass should have cleared tracker of all tensors")`
|
||||
|
||||
> A: This may happen due to edge cases in using the modern OffloadActivations context manager for CUDA streams. If you encounter this error, you may have success using the naive implementation with `offload_activations: legacy` in your YAML.
|
||||
|
||||
Reference in New Issue
Block a user