checking "transposed convolution" at beginning #3631
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
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/3631
Note: Links to docs will display an error until the docs builds have been completed. ✅ You can merge normally! (1 Unrelated Failure)As of commit 3e3a6b8 with merge base ad419a5 ( FLAKY - The following job failed but was likely due to flakiness present on trunk:
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
facebook-github-bot
added
the
CLA Signed
|
This pull request was exported from Phabricator. Differential Revision: D57424572 |
Summary: Current `check_convolution_args` checks shapes of parameters before checking `transposed`. For a model using `ConvTranspose2d`, this may incur error like ``` executorch:tensor_util.h:655] Check failed (a.size(dim_a) == b.size(dim_b)): Tensors do not match: a.size(0) = 512 does not match b.size(0) = 1024 ``` instead of showing `"transposed convolution not supported yet."` (see "ET CPU" in the test plan of D57258810). The reason for getting the error above is due to the difference of shape of `weight` between `ConvTranspose2d` and `Conv2d`, specifically, - [shape of `weight` in `ConvTranspose2d` ](https://pytorch.org/docs/stable/generated/torch.nn.ConvTranspose2d.html) is (note that the first item is `in_channels`) ``` (in_channels, out_channels/groups, kernel_size[0], kernel_size[1]) ``` - instead, [shape of `weight` in `Conv2d`](https://pytorch.org/docs/stable/generated/torch.nn.Conv2d.html) is (note that the first item is `out_channels`) ``` (out_channels, in_channels/groups, kernel_size[0], kernel_size[1]) ``` so [`tensors_have_same_size_at_dims(bias.value(), 0, weight, 0));`](https://www.internalfb.com/code/fbsource/[d34494a8af88b8c9742d00d62046d7884da47d7e]/fbcode/executorch/kernels/portable/cpu/util/kernel_ops_util.cpp?lines=314) may fail for `ConvTranspose2d`. The error message would be more informative if we check `transposed` at the beginning. Differential Revision: D57424572
copyrightly
force-pushed
the
export-D57424572
branch
from
326ba2f
to
3e3a6b8
Compare
|
This pull request was exported from Phabricator. Differential Revision: D57424572 |
manuelcandales
approved these changes
May 16, 2024
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.
Summary:
Current
check_convolution_argschecks shapes of parameters before checkingtransposed. For a model usingConvTranspose2d, this may incur error likeinstead of showing
"transposed convolution not supported yet."(see "ET CPU" in the test plan of D57258810).The reason for getting the error above is due to the difference of shape of
weightbetweenConvTranspose2dandConv2d, specifically,weightinConvTranspose2dis (note that the first item isin_channels)weightinConv2dis (note that the first item isout_channels)so
tensors_have_same_size_at_dims(bias.value(), 0, weight, 0));may fail forConvTranspose2d.The error message would be more informative if we check
transposedat the beginning.Differential Revision: D57424572