Xtensa LSTM: #2150
Merged
Xtensa LSTM: #2150
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
Enabled LSTM kernel support for XTENSA target.
|
#2171 updates the test case to have zero_point=0. Will the tests now pass for the Xtensa kernels? |
Yes, with this update, tests should pass for the Xtensa kernels. |
…rget. This is due to API changes during integration code optimization.
rascani
previously approved these changes
Sep 20, 2023
…fi nn hlibraries and updated transpose_conv to use the latest nn library.
|
"This PR is being marked as stale due to inactivity. Remove label or comment to prevent closure in 5 days." |
|
@cad-audio - bump. I'd like to land this, but the tests are still failing. |
to use number of samples, instead of bytes.
rascani
previously approved these changes
Dec 5, 2023
TFLM has historically used a HIFI4 define for both HiFi3/3z and HiFi4 cores. That was recently split to use a separate define for HiFi3/3z to include the different NDSP libraries. This commit updates the PR to use this new convention within the xtensa kernels that were changed.
The xa_nnlib_hifi5 library has a modified copy of the NDSP library contained within it. The TFLM repository has recently switched to using the standalone copy of NDSP library for all HiFis. There are some mismatches between these two versions which caused compilation failures in TFLM for HiFi5. In particular, the DISCARD_FUN_FOR_NONVOID_RETURN macro was missing from the NDSP HiFi5 library, which is necessary for compiling xa_nn_matumul_f16.c. This commit adds that macro to the local TFLM patch that we maintain.
rascani
approved these changes
Jan 2, 2024
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Outdated suggestions cannot be applied.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.
Enabled LSTM kernel support for XTENSA target.
Updated xtensa_downloads script to use the latest HiFi NN Libraries.
The 8x16 unit test cases has non-zero zero_point for 16 bit output.
https://github.com/tensorflow/tflite-micro/blob/main/tensorflow/lite/micro/kernels/testdata/lstm_test_data.cc#L255C1-L258C61
Default run for all the 8x16 unit test cases result: FAIL. This is due to non-zero output offset value.
BUG=#1867