Round inputs for dense unrolled RNN tests to make pytests more stable #1284
+1
−0
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
calad0i
reviewed
Apr 29, 2025
| @@ -107,6 +107,7 @@ def test_resource_unrolled_rnn(rnn_layer, backend, io_type, static, reuse_factor | |||
| # Subtract 0.5 to include negative values | |||
| input_shape = (12, 8) | |||
| X = np.random.rand(50, *input_shape) - 0.5 | |||
| X = np.round(X * 2**16) * 2**-16 # make it exact ap_fixed<32,16> | |||
There was a problem hiding this comment.
This is indeed ap_fixed<17, 0>. Though, as you mentioned that the issue is more prevalent with unrolled dense, is there anything compromising bit-exactness between implementations?
There was a problem hiding this comment.
I think it might be the RNN, or maybe just LSTM, in general. Jovan did this change to get the tests to behave better in the pytorch parser case. It seemed to have worked there, so I'd be in favor of merging this now so get more meaningful test results, and look into why this is such an issue later.
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Outdated suggestions cannot be applied.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.
We are still having problems with numerical stability of RNNs in the pytests for dense unrolled. This PR aims to fix that by implementing the rounding @jmitrevs added in #1215 for these tests as well. As I haven't been able to reproduce the failure locally, I couldn't check if this actually fixes it
Type of change
Tests
Checklist
pre-commiton the files I edited or added.