yzhliu commented on a change in pull request #8669:
URL: https://github.com/apache/tvm/pull/8669#discussion_r686425818
##########
File path: python/tvm/topi/x86/dense_alter_op.py
##########
@@ -39,6 +40,17 @@ def _alter_dense_layout(attrs, inputs, tinfos, out_type):
relay.op.get("nn.dense"), attrs, tinfos, out_type, target
)
workload = autotvm.task.get_workload(outs)
+
+ data_type =
relay.transform.InferType()(IRModule.from_expr(inputs[0]))["main"].body.checked_type
Review comment:
Hi @masahi sorry for late reply. Would you mind debug a little bit see
if inputs[0]'s layout is correctly inferred? My suspect is inputs[0] becomes
unknown layout during the alterlayout, and that's why `InferCorrectLayout`
doesn't work.
I implemented `AlterOpLayout` in NNVM at the first place and @merrymercy
migrated it to Relay. `InferCorrectLayout` should work with this pass or at
least it was my initial intention if I remember correctly.
--
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]
