You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@tvm.apache.org by GitBox <gi...@apache.org> on 2022/02/04 05:45:58 UTC
[GitHub] [tvm] AndrewZhaoLuo opened a new pull request #10162: [ONNX][QNN] Fix QLinearConvs with groups and per-channel quantization
AndrewZhaoLuo opened a new pull request #10162:
URL: https://github.com/apache/tvm/pull/10162
Solves issue https://github.com/apache/tvm/issues/10046 and https://github.com/apache/tvm/issues/10088
Note, I can not get ONNX RT tests to work unfortunately :/ any guidance here would be helpful.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: commits-unsubscribe@tvm.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [tvm] AndrewZhaoLuo edited a comment on pull request #10162: [WIP][ONNX][QNN] Fix QLinearConvs with groups and per-channel quantization
Posted by GitBox <gi...@apache.org>.
AndrewZhaoLuo edited a comment on pull request #10162:
URL: https://github.com/apache/tvm/pull/10162#issuecomment-1042101011
Yeah so this is a bit trickier than I thought. Turns out depthwise seperable convolutions have special kernel shape representation because why not. Specifically it's [out_channel / depth_multiplier, depth_multipler, ...] which destroyed some assumptions.
General grouped conv has kern shape of [out_channel, in_channels / groups, ...]
Made issue https://github.com/apache/tvm/issues/10294
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: commits-unsubscribe@tvm.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [tvm] AndrewZhaoLuo commented on pull request #10162: [ONNX][QNN] Fix QLinearConvs with groups and per-channel quantization
Posted by GitBox <gi...@apache.org>.
AndrewZhaoLuo commented on pull request #10162:
URL: https://github.com/apache/tvm/pull/10162#issuecomment-1042101011
Yeah so this is a bit trickier than I thought. Turns out depthwise seperable convolutions have special kernel shape representation because why not. Specifically it's [out_channel / depth_multiplier, depth_multipler, ...] which destroyed some assumptions.
General grouped conv has kern shape of [out_channel, in_channels / groups, ...]
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: commits-unsubscribe@tvm.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org