Lunderberg commented on PR #16356:
URL: https://github.com/apache/tvm/pull/16356#issuecomment-1881630188

   @tqchen Thank you on the details, and agreed that this is primarily for ease 
of readability.  Until and unless the hidden struct info can round-trip in all 
cases, it shouldn't be used for saving results.
   
   For the opaque functions, I *think* the current implementation will avoid 
the issue you described.  In order to hide the struct info, three conditions 
need to hold.  (1) The `show_all_struct_info` config is `False`, (2) the struct 
info of the expression can be inferred, and (3) the inferred struct info is 
structurally equal to the actual struct info.  I think the example you gave 
would fail the third condition, with `R.Tensor[s]` and `R.Tensor[s1]` being 
recognized as different.
   
   I don't see any unit tests that currently use opaque functions.  Are there 
examples somewhere so I could add an appropriate unit test?


-- 
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]

Reply via email to