ibsidorenko commented on code in PR #13854:
URL: https://github.com/apache/tvm/pull/13854#discussion_r1093561085
##########
tests/python/relay/test_pass_qnn_legalize.py:
##########
@@ -296,7 +297,98 @@ def _get_mod(data_dtype, kernel_dtype):
assert "cast" in legalized_mod.astext() and "qnn" in
legalized_mod.astext()
+def test_qnn_legalize_qnn_conv2d_non_scalar_qnn_params():
+ """
+ Test QNN legalization for qnn.dense op for Hexagon target when kernel zero
point and kernel
+ scale are not scalars.
Review Comment:
So, by "not scalar" I mean constant vector of scalars. For example:
`relay.const([val1, val2 ... valN])`
This is more interesting case for testing because on legalization we do
padding of weights. And dimension of weights, vector of scale/zero point and
'channels' attribute should be aligned.
--
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]
