masahi commented on a change in pull request #9593:
URL: https://github.com/apache/tvm/pull/9593#discussion_r757303018
##########
File path: python/tvm/relay/op/_transform.py
##########
@@ -67,6 +67,10 @@
_reg.register_injective_schedule("adv_index")
+@_reg.register_compute("concatenate")
+def compute_concat(attrs, inputs, output_type):
+ return [topi.concatenate(inputs, attrs.axis)]
Review comment:
Right now this code doesn't work on GPU, because `topi.concatenate` is
hard coded for x86.
We should remove `register_compute("concatenate")` and instead introduce
proper op strategy for `concat`. See how compute and schedule for other ops are
registered in `relay/op/strategy`.
--
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]
