[LoweringContext] Support an optimized parameter mapping for SPMD#8460
Merged
tengyifei merged 1 commit intopytorch:masterfrom Dec 7, 2024
Merged
[LoweringContext] Support an optimized parameter mapping for SPMD#8460tengyifei merged 1 commit intopytorch:masterfrom
tengyifei merged 1 commit intopytorch:masterfrom
Conversation
Collaborator
Author
|
Re-opened from #8453, cleaned up the merge commit. |
tengyifei
approved these changes
Dec 5, 2024
8fd7ac7 to
9858577
Compare
tengyifei
added a commit
that referenced
this pull request
Jan 2, 2025
Previously scan uses `parameter_id_tensor_mapping` to fetch tensors hoisted to HLO parameters e.g. the fn being scanned may create additional tensors while its running. `parameter_id_tensor_mapping` will fetch those tensors back to host as XLA literals and create new tensors wrapphing those, resulting in additional host RAM usage. PR #8460 added `device_parameter_id_tensor_mapping` that returns the actual device backed tensors instead of another copy. So we'll use that and test that this avoids host transfers.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Currently, the existing parameter mapping for the lowering context is not well suited for SPMD. In case of large models, it will cause a large synchronous bottleneck when transferring all device data to the host. This is caused by each ReplicateShardedData computation that gathers and reassembles each sharded data across multiple devices. This is by design, since it is expected to collect all parameters regardless of their allocation.
In this PR, we introduce a new mapping that does not invoke the sharded replication, but instead uses references to the device data. This is generally sufficient and preferred in most cases, where the user only wants to access the validate parameters (those that are not returned as
-1fromtensor_parameter_id, as 'fake' parameters).