[mkldnn][pooling] support ceil mode by padding changes #21310
Conversation
pytorchbot
added
module: mkldnn
facebook-github-bot
left a comment
facebook-github-bot
left a comment
There was a problem hiding this comment.
@akyrola has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator.
|
Ok, how about this: I think the math is too complex (with strides and dilations), so better not spread it around., so let's follow the Caffe-approach and just keep adjusting pad_r until the output shapes match. |
facebook-github-bot
left a comment
facebook-github-bot
left a comment
There was a problem hiding this comment.
@akyrola 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.
@akyrola 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.
@akyrola has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator.
|
@pytorchbot retest this please |
facebook-github-bot
left a comment
facebook-github-bot
left a comment
There was a problem hiding this comment.
@akyrola has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator.
Summary: Modify MKLDNN pooling operation to support ceil mode by adjusting the right/bottom padding accordingly. This is done similarly as in Caffe (see discussion pytorch/pytorch#19205 (comment)). To make this possible, I split the padding to left and right (top / bottom). This naming is confusing but actually follows mkldnn's own naming for pooling::compute(). We increase the r paddings so that it matches the ceiling mode expected output size. Strengthened the test case. Pull Request resolved: pytorch/pytorch#21310 Reviewed By: bddppq Differential Revision: D15611664 Pulled By: akyrola fbshipit-source-id: 46b40015dafef69a8fd5e7b2c261d8dbf448cd20
7 participants
Modify MKLDNN pooling operation to support ceil mode by adjusting the right/bottom padding accordingly. This is done similarly as in Caffe (see discussion #19205 (comment)).
To make this possible, I split the padding to left and right (top / bottom). This naming is confusing but actually follows mkldnn's own naming for pooling::compute(). We increase the r paddings so that it matches the ceiling mode expected output size.
Strengthened the test case.