[microNPU][ETHOSU] Add restrictions to convert to NHCWB16 layout in LayoutOptimization pass #14464
Conversation
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
…ayoutOptimization pass Before format NHCWB16 was used for operations performed by NPU, but there are limitations, for example ReduceSum requires NHWC input in case input tensor has type int32 or accelerator is Ethos_U65_512 https://review.mlplatform.org/plugins/gitiles/ml/ethos-u/ethos-u-vela/+/refs/tags/3.7.0/ethosu/vela/graph_optimiser_util.py#126
Collaborator
ilyag-grovety
suggested changes
Apr 3, 2023
ilyag-grovety
approved these changes
Apr 7, 2023
lhutton1
approved these changes
Apr 11, 2023
Contributor
lhutton1
left a comment
lhutton1
left a comment
There was a problem hiding this comment.
Thanks @Aleksei-grovety, LGTM!
Contributor
|
Thanks @Aleksei-grovety, @ilyag-grovety! |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 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Outdated suggestions cannot be applied.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.
Before format NHCWB16 was used for operations performed by NPU, but there are limitations, for example ReduceSum requires NHWC input in case input tensor has type int32 or accelerator is Ethos_U65_512 https://review.mlplatform.org/plugins/gitiles/ml/ethos-u/ethos-u-vela/+/refs/tags/3.7.0/ethosu/vela/graph_optimiser_util.py#126
cc @neildhickey, @lhutton1, @ekalda, @leandron, @ilyag-grovety