Add a context manager for activation sharding. #1127
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
What does this PR do?
This pr adds a context manager for sharding (potentially long-living) activation saved tensors across model parallel group. It can be used to wrap any non-mp regions where the activations are duplicated across all ranks in a mp group. One specific and typical use case is to wrap a series of gradient checkpoints of transformer layers, creating a layer input sharding setup similar to the one described in the GSPMD paper.
The implementation uses torch.autograd.graph.saved_tensors_hooks and share similar spirits with torch.autograd.graph.save_on_cpu.