[ETHOSN] Get buffer sizes from the compiled network #12160
Merged
Conversation
This file contains 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
lhutton1
force-pushed
the
fix-buffer-size
branch
from
76d6411
to
7af2390
Compare
The NPU support library compiler sometimes adds padding to input tensors which means the buffer sizes calculated at runtime can sometimes be smaller than necessary. Instead, buffer sizes are now collected at compile time and passed to the runtime so that they match the sizes expected by the compiled network. This was seen when running a fully connected operation with an input that is not a multiple of 1024, so testing has been added to cover this case. Additionally changed the fully connected test case to use pytest parameterization as part of a general cleanup, and fixed an issue with specifying a different output shape and weights with more than 1 output channel. Change-Id: Iad319d75326b9ac41950de982603660a084dc27b
lhutton1
force-pushed
the
fix-buffer-size
branch
from
7af2390
to
6ccae1d
Compare
|
friendly ping for review |
NicolaLancellotti
approved these changes
Aug 4, 2022
|
Thanks @lhutton1 @NicolaLancellotti! |
xinetzone
pushed a commit
to daobook/tvm
that referenced
this pull request
Nov 25, 2022
The NPU support library compiler sometimes adds padding to input tensors which means the buffer sizes calculated at runtime can sometimes be smaller than necessary. Instead, buffer sizes are now collected at compile time and passed to the runtime so that they match the sizes expected by the compiled network. This was seen when running a fully connected operation with an input that is not a multiple of 1024, so testing has been added to cover this case. Additionally changed the fully connected test case to use pytest parameterization as part of a general cleanup, and fixed an issue with specifying a different output shape and weights with more than 1 output channel. Change-Id: Iad319d75326b9ac41950de982603660a084dc27b
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
The NPU support library compiler sometimes adds padding to input tensors which means the buffer sizes calculated at runtime can sometimes be smaller than necessary. Instead, buffer sizes are now collected at compile time and passed to the runtime so that they match the sizes expected by the compiled network. This was seen when running a fully connected operation with an input that is not a multiple of 1024, so testing has been added to cover this case.
Additionally changed the fully connected test case to use pytest parameterization as part of a general cleanup, and fixed the fully connected testing to support output channels > 1.
cc @Leo-arm @manupa-arm @leandron