[topi][CuDNN] Removed requirement for GPU from topi conv2d_cudnn.cuda and conv3d_cudnn.cuda - #8276
Merged
Merged
Conversation
Contributor
Author
|
Related PR #8275 is for the same goal of allowing CuDNN modules to be build on a local non-GPU machine for use on a remote GPU machine. The two implementations are independent, and are separate PRs for reviewing purposes. Potential reviewer: @mdw-octoml |
… and conv3d_cudnn.cuda Previously, `conv2d_cudnn.cuda` would use cudnn's benchmarking function to select a forward convolution when `cfg.is_fallback`, and `conv3d_cudnn.cuda` would use cudnn's benchmarking at all times. After this commit, both expose the cudnn algorithm choice as an option. If `cfg.is_fallback`, the local device will be benchmarked if present, otherwise will select a default cudnn implementation. In the future, to better support RPC use-cases, the fallback config should be based on cudnn-specific parameters saved in the Target object.
Lunderberg
force-pushed
the
cudnn_conv_find_algo
branch
from
June 17, 2021 18:13
0cdeef6 to
f7fa507
Compare
jwfromm
approved these changes
Jun 17, 2021
jwfromm
left a comment
jwfromm
left a comment
Contributor
There was a problem hiding this comment.
Really nice change, thanks Eric!
ylc
pushed a commit
to ylc/tvm
that referenced
this pull request
Sep 29, 2021
… and conv3d_cudnn.cuda (apache#8276) Previously, `conv2d_cudnn.cuda` would use cudnn's benchmarking function to select a forward convolution when `cfg.is_fallback`, and `conv3d_cudnn.cuda` would use cudnn's benchmarking at all times. After this commit, both expose the cudnn algorithm choice as an option. If `cfg.is_fallback`, the local device will be benchmarked if present, otherwise will select a default cudnn implementation. In the future, to better support RPC use-cases, the fallback config should be based on cudnn-specific parameters saved in the Target object. Co-authored-by: Eric Lunderberg <elunderberg@octoml.ai>
zxy844288792
pushed a commit
to zxy844288792/tvm
that referenced
this pull request
Mar 4, 2022
… and conv3d_cudnn.cuda (apache#8276) Previously, `conv2d_cudnn.cuda` would use cudnn's benchmarking function to select a forward convolution when `cfg.is_fallback`, and `conv3d_cudnn.cuda` would use cudnn's benchmarking at all times. After this commit, both expose the cudnn algorithm choice as an option. If `cfg.is_fallback`, the local device will be benchmarked if present, otherwise will select a default cudnn implementation. In the future, to better support RPC use-cases, the fallback config should be based on cudnn-specific parameters saved in the Target object. Co-authored-by: Eric Lunderberg <elunderberg@octoml.ai>
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
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.
Previously,
conv2d_cudnn.cudawould use cudnn's benchmarking function to select a forward convolution whencfg.is_fallback, andconv3d_cudnn.cudawould use cudnn's benchmarking at all times. After this commit, both expose the cudnn algorithm choice as an option. Ifcfg.is_fallback, the local device will be benchmarked if present, otherwise will select a default cudnn implementation.In the future, to better support RPC use-cases, the fallback config should be based on cudnn-specific parameters saved in the Target object.