[CMSIS-NN] Aligned scale computation with TFLM to fix numerical mismatch #10817
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
Change-Id: I89bb99f46d2f04d4cd421c80cdbea2b7e34e9309
|
cc: @manupa-arm @Mousius @grant-arm for code review. |
Mousius
requested changes
Mar 30, 2022
| @@ -304,6 +304,94 @@ def test_conv2d_asymmetric_padding_int8( | |||
| ) | |||
|
|
|||
|
|
|||
| # This test expects assertion as the output should mismatch between TVM and CMSIS-NN | |||
There was a problem hiding this comment.
Could this assert that it matches the expected output? Similar to how the NPU tests check that they are consistent with TFLite?
There was a problem hiding this comment.
Yes, we could compare against TFLite output. So far all the CMSIS-NN tests compare against TVM output.
Change-Id: I3b58d74a11e6c0649fe9454a85efd91cfbc52ccd
Change-Id: I5b788da3c34d9ed480d6f9197182e074df35e14e
Change-Id: I822ba3b99b47c3becdd925e1068b894ca07c84bb
…checks Change-Id: I103b3b89692ac24249b619f0806e9f269f64dad3
Change-Id: Ibac370437e7961c1a65845d4861c3db2613f35ae
Change-Id: Icdcdc3bcc2f3330fc78410eede66c8c4ea273d43
Change-Id: I6f87d09b67c33b0cb37fef5d4d9fce72bd385cf8
|
@Mousius I have added a TFLite test. PTAL when you have time. As discussed offline, we can move this test to tvm/testing area in future. |
Mousius
approved these changes
Apr 6, 2022
There was a problem hiding this comment.
LGTM! Thanks for the testing updates @ashutosh-arm 😸 looking forward to the generalisation later on!
pfk-beta
pushed a commit
to pfk-beta/tvm
that referenced
this pull request
Apr 11, 2022
…tch (apache#10817) Fixes numerical mismatch in Conv2D layers by aligning order of output scale computation with TFLM. Correct output scale is needed to calculate quantization parameters needed by CMSIS-NN.
mehrdadh
pushed a commit
to mehrdadh/tvm
that referenced
this pull request
Apr 11, 2022
…tch (apache#10817) Fixes numerical mismatch in Conv2D layers by aligning order of output scale computation with TFLM. Correct output scale is needed to calculate quantization parameters needed by CMSIS-NN.
Lucien0
pushed a commit
to Lucien0/tvm
that referenced
this pull request
Apr 19, 2022
…tch (apache#10817) Fixes numerical mismatch in Conv2D layers by aligning order of output scale computation with TFLM. Correct output scale is needed to calculate quantization parameters needed by CMSIS-NN.
altanh
pushed a commit
to altanh/tvm
that referenced
this pull request
Apr 28, 2022
…tch (apache#10817) Fixes numerical mismatch in Conv2D layers by aligning order of output scale computation with TFLM. Correct output scale is needed to calculate quantization parameters needed by CMSIS-NN.
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.
This PR fixes numerical mismatch in Conv2D layers by
aligning order of output scale computation
with TFLM. Correct output scale is needed
to calculate quantization parameters needed
by CMSIS-NN.