Add mkldnn support for adaptive_avg_pool2d #19818
Conversation
pytorchbot
added
module: mkldnn
bddppq
force-pushed
the
mkldnn_adaptive_avg_pool2d
branch
3 times, most recently
from
3d5e91a to
8911a99
Compare
facebook-github-bot
left a comment
facebook-github-bot
left a comment
There was a problem hiding this comment.
@bddppq has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator.
facebook-github-bot
left a comment
facebook-github-bot
left a comment
There was a problem hiding this comment.
@bddppq has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator.
bddppq
force-pushed
the
mkldnn_adaptive_avg_pool2d
branch
from
8911a99 to
b6f75bc
Compare
wanchaol
left a comment
wanchaol
left a comment
There was a problem hiding this comment.
lgtm. minor questions.
| - func: mkldnn_adaptive_avg_pool2d(Tensor self, int[2] output_size) -> Tensor | ||
| dispatch: | ||
| MkldnnCPU: mkldnn_adaptive_avg_pool2d | ||
| requires_tensor: True |
There was a problem hiding this comment.
I might miss some context here. What's the meaning of this option?
There was a problem hiding this comment.
This makes the codegen script to insert code to unwrap a variable to a tensor before passing the input to the kernel:
https://github.com/pytorch/pytorch/blob/de19eee/tools/autograd/gen_variable_type.py#L869-L870
https://github.com/pytorch/pytorch/blob/de19eee/tools/autograd/gen_variable_type.py#L954
|
|
||
| self.assertEqual( | ||
| adaptive_avg_pool2d(x), | ||
| adaptive_avg_pool2d(x.to_mkldnn()).to_dense()) |
There was a problem hiding this comment.
Does it mean that everytime we call mkldnn ops we are required to call to_dense ?
There was a problem hiding this comment.
not every op, but every chunk of ops (in this case the unittest is only testing one op so yeah it's calling to_dense after each op to convert the mkldnn tensor to dense tensor so that we can compare with the regular dense kernel's output).
basically one needs to do:
to_mkldnn -> mkldnn_op1 -> mkldnn_op2 ... -> mkldnn_opn -> to_dense
facebook-github-bot
left a comment
facebook-github-bot
left a comment
There was a problem hiding this comment.
@bddppq has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator.
Summary: AdaptiveAvgPool2d is used in torchvision resnet models https://github.com/pytorch/vision/blob/9a481d0/torchvision/models/resnet.py#L145 Fixes #19797 Pull Request resolved: pytorch/pytorch#19818 Differential Revision: D15112777 Pulled By: bddppq fbshipit-source-id: 6c9b29c805d28356cda49c10c2cd3ce9d7a8b3f5
Summary: AdaptiveAvgPool2d is used in torchvision resnet models https://github.com/pytorch/vision/blob/9a481d0/torchvision/models/resnet.py#L145 Fixes pytorch#19797 Pull Request resolved: pytorch#19818 Differential Revision: D15112777 Pulled By: bddppq fbshipit-source-id: 6c9b29c805d28356cda49c10c2cd3ce9d7a8b3f5
4 participants
AdaptiveAvgPool2d is used in torchvision resnet models https://github.com/pytorch/vision/blob/9a481d0/torchvision/models/resnet.py#L145
Fixes #19797