New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Remove fgrad_input from slow_conv2d #64280
Closed
Closed
Changes from 4 commits
Commits
Show all changes
7 commits
Select commit
Hold shift + click to select a range
fefa920
Remove fgrad_input from slow_conv2d
peterbell10 480cd16
Update on "Remove fgrad_input from slow_conv2d"
peterbell10 08ab268
Update on "Remove fgrad_input from slow_conv2d"
peterbell10 6dce32c
Update on "Remove fgrad_input from slow_conv2d"
peterbell10 3f7e6a7
Update on "Remove fgrad_input from slow_conv2d"
peterbell10 52f7393
Update on "Remove fgrad_input from slow_conv2d"
peterbell10 d6d2760
Update on "Remove fgrad_input from slow_conv2d"
peterbell10 File filter
Filter by extension
Conversations
Failed to load comments.
Jump to
Jump to file
Failed to load files.
Diff view
Diff view
There are no files selected for viewing
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
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
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
Oops, something went wrong.
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Outdated suggestions cannot be applied.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.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This is a much nicer way to compute
grad_bias
and seems correct AFAICT.Couple questions regarding this:
grad_bias
calculation? I briefly checked but nothing stood outThere was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
The nn tests are a bit confusing, but I think the gradcheck is done here:
pytorch/torch/testing/_internal/common_nn.py
Line 5807 in 59fcbd1
That test includes inputs and parameters, so
bias
should be covered.I've done some simple benchmarking which showed positive results. LMK if you want more.
For CUDA, I've run a convolution in a loop under nvprof with these parameters:
The result is it calls one kernel instead of two (cublas gemm here becomes
dot_kernel
+reduce_1Block_kernel
). The kernel execution time is also noticeably faster:On CPU without nvprof, the results are dwarfed by the main convolution and the same shapes showed no measurable difference. So I also tried a smaller input size with H=120, C_in = 2 and that showed some minor improvement in the forward case (presumably from vectorization).