Added adaptive_max_poolNd batch rule #263
Merged
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
|
@vfdev-5 Could you use one of the boxed fallbacks to handle this? Or modify one of the boxed fallbacks? |
|
@Chillee which boxed fallbacks you think of ? |
|
Looks similar at first glance to |
|
@Chillee I updated the PR according to your suggestion. Single cuda ci job is failing maybe due to a flaky test ... |
Chillee
reviewed
Nov 18, 2021
functorch/csrc/BatchRulesPooling.cpp
Outdated
| @@ -82,6 +82,46 @@ max_pool2d_with_indices_batch_rule( | |||
| reshape_dim_outof(0, bdim_size, std::get<1>(result)), 0); | |||
| } | |||
|
|
|||
| // We can't use ALL_TENSORS_HAVE_OPTIONAL_BDIM_BOXED because the CUDA | |||
There was a problem hiding this comment.
Perhaps it's better to just augment the boxed rule with the option to make some args contiguous?
There was a problem hiding this comment.
@Chillee good idea! I did an implementation with one tensor that could be contiguous (later we can make it more generic if needed). What do you think ?
Description: - Added adaptive_max_poolNd fw/bw batch rules - Updated tests Related to pytorch#240 Notes: I created two additional macros to handle adaptive_max_pool2d and adaptive_max_pool3d_backward. Not sure if we could make a generic rule to handle max_pool2d_with_indices_backward_batch_rule and adaptive_max_pool3d_backward, as max_pool2d_with_indices_backward_batch_rule requires some args in the middle between gradOutput, input and indices.
- max_pool2d_with_indices_backward - adaptive_max_pool2d_backward - adaptive_max_pool3d_backward and added ALL_TENSORS_HAVE_OPTIONAL_BDIM_BOXED_CONTIG1 to handle that
vfdev-5
force-pushed
the
add-adaptive_max_pool
branch
from
952a75b to
f5777a6
Compare
zou3519
pushed a commit
to zou3519/pytorch
that referenced
this pull request
Jul 20, 2022
* Added adaptive_max_poolNd fw/bw batch rules Description: - Added adaptive_max_poolNd fw/bw batch rules - Updated tests Related to pytorch/functorch#240 Notes: I created two additional macros to handle adaptive_max_pool2d and adaptive_max_pool3d_backward. Not sure if we could make a generic rule to handle max_pool2d_with_indices_backward_batch_rule and adaptive_max_pool3d_backward, as max_pool2d_with_indices_backward_batch_rule requires some args in the middle between gradOutput, input and indices. * Replaced EXISTING_BDIM_MULTIOUT by EXISTING_BDIM_ALL_BOXED * Removed specific implementations with indices.contiguous() for - max_pool2d_with_indices_backward - adaptive_max_pool2d_backward - adaptive_max_pool3d_backward and added ALL_TENSORS_HAVE_OPTIONAL_BDIM_BOXED_CONTIG1 to handle that
bigfootjon
pushed a commit
to pytorch/pytorch
that referenced
this pull request
Jul 21, 2022
* Added adaptive_max_poolNd fw/bw batch rules Description: - Added adaptive_max_poolNd fw/bw batch rules - Updated tests Related to pytorch/functorch#240 Notes: I created two additional macros to handle adaptive_max_pool2d and adaptive_max_pool3d_backward. Not sure if we could make a generic rule to handle max_pool2d_with_indices_backward_batch_rule and adaptive_max_pool3d_backward, as max_pool2d_with_indices_backward_batch_rule requires some args in the middle between gradOutput, input and indices. * Replaced EXISTING_BDIM_MULTIOUT by EXISTING_BDIM_ALL_BOXED * Removed specific implementations with indices.contiguous() for - max_pool2d_with_indices_backward - adaptive_max_pool2d_backward - adaptive_max_pool3d_backward and added ALL_TENSORS_HAVE_OPTIONAL_BDIM_BOXED_CONTIG1 to handle that
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.
Description:
Related to #240
Notes:
I created two additional macros to handle adaptive_max_pool2d and adaptive_max_pool3d_backward.
Not sure if we could make a generic rule to handle max_pool2d_with_indices_backward_batch_rule and adaptive_max_pool3d_backward, as max_pool2d_with_indices_backward_batch_rule requires some args in the middle between gradOutput, input and indices.