electriclilies commented on a change in pull request #6260:
URL: https://github.com/apache/incubator-tvm/pull/6260#discussion_r469550172
##########
File path: python/tvm/relay/op/_tensor.py
##########
@@ -201,11 +201,11 @@ def elemwise_shape_func(attrs, inputs, _):
return [topi.math.identity(inputs[0])]
register_shape_func("cast", False, elemwise_shape_func)
-register_shape_func("zeros", True, no_data_full_shape_func)
+register_shape_func("zeros", False, full_shape_func)
register_shape_func("zeros_like", False, elemwise_shape_func)
-register_shape_func("ones", True, no_data_full_shape_func)
+register_shape_func("ones", False, full_shape_func)
Review comment:
No, I was going through my code and I forgot to change this for the
zeros and ones PR -- for the static shape ones and zeros, it is not data
dependent since the shape of the input is in attributes, and uses the normal
full_shape_func. It passes the tests for static ones and zeros..
----------------------------------------------------------------
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.
For queries about this service, please contact Infrastructure at:
[email protected]
