Fix(informer): Correct tensor shape for input_size=1 #38780
Closed
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
When InformerConfig's input_size is set to 1, the loc and scale tensors retain an extra dimension, causing a RuntimeError. This commit fixes the issue by overriding and applying .squeeze(1) to both tensors to ensure correct dimensionality for all input_size cases. Fixes huggingface#38745
|
Hi @Flink-ddd, thanks for the PR! However, it creates quite a large change in the Can you try either:
|
|
Hi @Rocketknight1, thank you so much for taking the time to review and for the great suggestion! I completely agree that fixing the bug in the parent 'TimeSeriesTransformerModel' is a much more elegant solution than overriding the entire function. It's cleaner and avoids code duplication. I will go ahead and close this PR, create a new branch to apply the fix directly to the parent class, and then open a new PR for review. Thanks again for your guidance! |
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Outdated suggestions cannot be applied.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.
What does this PR do?
This PR fixes a RuntimeError that occurs in InformerModel when config.input_size is set to 1.
When input_size=1, the loc and scale tensors calculated by the scaler retained an extra dimension (e.g., shape [B, 1, 1] instead of [B, 1]). This incorrect shape caused a dimension mismatch error during the expand() operation later when creating expanded_static_feat.
This fix applies .squeeze(-1) to both the loc and scale tensors to ensure they have the correct dimensionality before being used. This resolves the crash and allows the model to run correctly with univariate time series data.
Fixes #38745