[GitHub] [tvm] manupa-arm commented on pull request #11228: [QNN] Enable constant folding for QNN operations.

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

manupa-arm commented on PR #11228:
URL: https://github.com/apache/tvm/pull/11228#issuecomment-1124990231

   Thanks @ibsidorenko for tests.
   
   Thanks @apeskov for the detailed explaination. 
   
   IIUC, the requirement is to find Expr s that would be folded down to a relay.Constant (going through qnn folding in the process) and keep the rest of qnn ops non-legalized. I recognize this feature to be important and happy to see this feature being completed. Sorry, I did not fully understand before what you were trying to achieve here.
   
   However, looking at the test cases, I could not recognize the above mentioned feature is achieved because all the test cases started with const arguments for qnn ops. Therefore, left me wondering why we cant achieve running a sequential of Legalize, FoldConstant.
   
   Since this change is being done to IRModule --> IRModule pass, I think it would be good to have a test case where we could observe that such constant subgraphs are folded while rest of the qnn ops (that does not have all arguments as constants) are not legalized.
   
   
   
   
   
   
   


-- 
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: commits-unsubscribe@tvm.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org