Skip to content

Conversation

gujinghui
Copy link
Collaborator

Implement operators for DNNLOWP, including int8_conv, int8_FC, int8_pooling, int8_relu, int8_sum, quantize/dequantize, and order_swtich operators.

@yinghai yinghai self-requested a review March 30, 2019 18:30
@gujinghui gujinghui force-pushed the dnnlowp_ops branch 2 times, most recently from e2c5a95 to 9ab664b Compare April 4, 2019 01:35
@gujinghui
Copy link
Collaborator Author

gujinghui commented Apr 4, 2019

@yinghai
rebased. pls help review.

@Jiaming-Liu
@jgong5
@uyongw

Copy link
Contributor

@facebook-github-bot facebook-github-bot left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@yinghai has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator.

@yinghai
Copy link
Contributor

yinghai commented Apr 4, 2019

There are no tests for most of the op change?

@gujinghui
Copy link
Collaborator Author

There are no tests for most of the op change?

@yinghai
I cannot find any test cases for DNNLOWP based-on FBGEMM.
So, I have no references to impl cases for DNNLOWP...
If needed, I will try to add cases soon.

@gujinghui
Copy link
Collaborator Author

@yinghai
test cases for dnnlowp have been added.
pls help review.

Copy link
Contributor

@facebook-github-bot facebook-github-bot left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@yinghai has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator.

Copy link
Contributor

@yinghai yinghai left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Please address the nits.

Signed-off-by: Gu, Jinghui <jinghui.gu@intel.com>
Signed-off-by: Gu, Jinghui <jinghui.gu@intel.com>
Signed-off-by: Gu, Jinghui <jinghui.gu@intel.com>
Signed-off-by: Gu, Jinghui <jinghui.gu@intel.com>
Signed-off-by: Gu, Jinghui <jinghui.gu@intel.com>
Signed-off-by: Gu, Jinghui <jinghui.gu@intel.com>
Signed-off-by: Gu, Jinghui <jinghui.gu@intel.com>
Signed-off-by: Gu, Jinghui <jinghui.gu@intel.com>
The weights format of wino is changed if input shape was changed.
The cached weights has to be reordered according to current input shape
if algorithm is winograd.

Signed-off-by: Gu, Jinghui <jinghui.gu@intel.com>
Signed-off-by: Gu, Jinghui <jinghui.gu@intel.com>
Signed-off-by: Gu, Jinghui <jinghui.gu@intel.com>
@gujinghui
Copy link
Collaborator Author

@yinghai
Fixed the the namespace issue.
Pls help review.

Copy link
Contributor

@facebook-github-bot facebook-github-bot left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@yinghai has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator.

@facebook-github-bot
Copy link
Contributor

@yinghai merged this pull request in 575aebc.

zhangguanheng66 pushed a commit to zhangguanheng66/pytorch that referenced this pull request May 6, 2019
Summary:
Implement operators for DNNLOWP, including int8_conv, int8_FC, int8_pooling, int8_relu, int8_sum, quantize/dequantize, and order_swtich operators.
Pull Request resolved: pytorch#18656

Differential Revision: D14767092

Pulled By: yinghai

fbshipit-source-id: 1f3e24929a358a42214da333bd304c593ea4468f
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

5 participants