Refactor(examples/mnist): Clarify loss reduction and correct dataset length #1396
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
…length This commit addresses two potential points of confusion and error in the MNIST example, as detailed in issue pytorch#623. 1. Explicit Loss Reduction: The train() function's use of F.nll_loss implicitly defaults to reduction='mean', whereas the test() function uses reduction='sum'. This change makes the `reduction='mean'` explicit in the train() function. This improves code clarity. 2. Correct Dataset Size with Samplers: Using `len(loader.dataset)` to get the number of samples is incorrect when a Sampler (e.g., SubsetRandomSampler for a validation split) is used. It incorrectly reports the full dataset size, not the subset size. The logic is updated to first check `len(loader.sampler)`. If a sampler exists, its length is used. Otherwise, it falls back to `len(loader.dataset)`. This ensures the correct number of samples is used for logging and calculations. Fixes pytorch#623
✅ Deploy Preview for pytorch-examples-preview canceled.
|
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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Outdated suggestions cannot be applied.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 commit addresses two potential points of confusion and error in the MNIST example, as detailed in issue #623.
1.Explicit Loss Reduction: The train() function's use of F.nll_loss implicitly defaults to reduction='mean', whereas the test() function uses reduction='sum'.
This change makes the
reduction='mean'explicit in the train() function. This improves code clarity.len(loader.dataset)to get the number of samples is incorrect when a Sampler (e.g., SubsetRandomSampler for a validation split) is used. It incorrectly reports the full dataset size, not the subset size.The logic is updated to first check
len(loader.sampler). If a sampler exists, its length is used. Otherwise, it falls back tolen(loader.dataset). This ensures the correct number of samples is used for logging and calculations.Fixes #623