New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
QSeparableConv1D and 2D #50
Conversation
@lishanok could you check this PR and internal CL? thanks. cc @danielemoro |
happy holidays! courtesy ping @lishanok and @danielemoro |
PTAL @lishanok and @danielemoro |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I believe this pull request is important since it properly implements QSeparableConv2D as a class and quantized layer instead of a function that simply creates other layers. However, we need to make sure that all parts of the QSeparableConv2D can be quantized.
-- 1847f67 by Vladimir Loncar <vloncar@users.noreply.github.com>: Separable CNN implementation -- bb717bc by Vladimir Loncar <vloncar@users.noreply.github.com>: Add support for SeparableConv in utils -- cc23ec4 by Vladimir Loncar <vloncar@users.noreply.github.com>: Test case for separable conv -- bb2e6c7 by Vladimir Loncar <vloncar@users.noreply.github.com>: Support SeparableConv in AutoQ -- 72e2bd0 by Vladimir Loncar <vloncar@users.noreply.github.com>: Update readme with new layers -- 5ab8ec0 by Vladimir Loncar <vloncar@users.noreply.github.com>: Add a note about quantizing depthwise activations COPYBARA_INTEGRATE_REVIEW=#50 from vloncar:sep_conv 5ab8ec0 PiperOrigin-RevId: 355256946 Change-Id: I106d43da286e519899425a83544ffa2e4afedb35
The test is actually not failed since we have it tested successfully from internal system. It was just due to travis-ci.org service failure. |
This PR adds
QSeparableConv1D
andQSeparableConv2D
. The existingQSeparableConv2D
(which expands toQDepthwiseConv2D
and 1x1QConv2D
) that is based on MobileNet is retained and renamedQMobileNetSeparableConv2D