[GitHub] [tvm] areusch edited a comment on pull request #7250: Remove check_correctness in AutoTVM, which is busted

[GitBox <gi...@apache.org>]

areusch edited a comment on pull request #7250:
URL: https://github.com/apache/tvm/pull/7250#issuecomment-758292075


   I would be okay with that except that this was the network I tried to tune:
   
   ```
   model = keras.models.Sequential()
   model.add(keras.layers.Conv2D(2,  3, input_shape=(64, 64, 3)))
   model.build()
   ```
   
   I could be wrong but I think this is a pretty common, basic use case. with conv2d_NCHWc schedule on by default, it doesn't seem like the warning case is an edge case (other than that perhaps we consider `check_correctness` to be an edge case--but...I would push back pretty strongly that a new developer is going to think: "why would I not want this?" and turn it on).
   
   is there an easy way we can detect which operator a layout transformation belongs to in task extraction maybe?


----------------------------------------------------------------
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:
users@infra.apache.org