jwfromm opened a new pull request #8456:
URL: https://github.com/apache/tvm/pull/8456


   While playing with [onnxruntime's quantization 
tools](https://www.onnxruntime.ai/docs/how-to/quantization.html) I found that 
`ConvInteger` nodes were heavily used and pretty easy to map to `qnn.conv2d`. 
This PR adds support for that op. 
   
   While working with convolutions, I found some unrelated issues in how we 
were doing autopadding for `ConvTranspose`. Specifically, padding can't always 
be treated as a two step process where inputs are padded and then 
conv_transpose is performed since padding in conv_transpose actually removes 
outputs. I simplified the autopadding `ConvTranspose` and found that this new 
implementation passes more of the onnx node tests.


-- 
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: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]


Reply via email to