lhutton1 commented on a change in pull request #10763:
URL: https://github.com/apache/tvm/pull/10763#discussion_r835207931
##########
File path: python/tvm/relay/backend/contrib/ethosu/te/convolution.py
##########
@@ -187,7 +187,7 @@ def conv2d_compute(
[1, 0, 0, 0, 0, 0],
[0, 1, 0, 0, 0, 0],
[0, 0, 0, 1, 0, 0],
- [0, 0, 16, 0, 1, -16],
+ [0, 0, 0, 0, 0, ofm_channels],
Review comment:
Just curious: Any reason for not making this change in
`binary_elementwise` and `unary_elementwise`? Additionally it looks like the
`nhcwb16_to_nhwc` and `nhwc_to_nhcwb16` matrices are the same across the
operators, should we move it to a common place like util to reduce duplication?
##########
File path: python/tvm/relay/backend/contrib/ethosu/te/pooling.py
##########
@@ -169,7 +169,7 @@ def pooling_compute(
[1, 0, 0, 0, 0, 0],
[0, 1, 0, 0, 0, 0],
[0, 0, 0, 1, 0, 0],
- [0, 0, 16, 0, 1, -16],
+ [0, 0, 0, 0, 0, int(ofm_channels)],
Review comment:
Is `int(...)` necessary here?
--
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]
