Fix, pipeline model with moe cause error when send grad #7055
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Outdated suggestions cannot be applied.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.
Fix, pipeline model with moe, send and recv grad use same buffer will cause error when moe loss not in position 1. Solution, activation and grad use diff buffer. BUG DETAIL: in forward, the grad_buf will allocate hidden_states,some
tensors(required_grad is false), moe_loss;in backward,only send act grad(floating point),so only hidden_states and moe loss's grads will be send, and forward and backward use same grad_buf. This will cause send and recv grad use diff dtype, and cause numerical error。