Fix senet pretrained weights issue #1689
Merged
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Outdated suggestions cannot be applied.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.
Signed-off-by: Yiheng Wang vennw@nvidia.com
Fixes #1688 .
Description
Except for the re-implementation of load pre-trained weights function, a new test case is added. For the same input tensor, the output between the torchvision's densenet and our version should be close (since we use nn.Linear for the fc layer, and Cadene's version uses conv layer with kernel size =1, it brings a few difference).
Status
Ready
Types of changes
./runtests.sh --codeformat --coverage
../runtests.sh --quick
.make html
command in thedocs/
folder.