tvm-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From GitBox <...@apache.org>
Subject [GitHub] [incubator-tvm] trevor-m opened a new pull request #5731: [TensorFlow] Don't add cast for batch norm when type isn't changing
Date Thu, 04 Jun 2020 20:47:22 GMT

trevor-m opened a new pull request #5731:
URL: https://github.com/apache/incubator-tvm/pull/5731


   TensorFlow batch norm op has two type attributes:
   `U` which is the type of the parameters scale, offset, mean, and variance.
   `T` which is the type of the input and output.
   
   The TF importer currently adds casts the input from `T` type to `U` type so that the computation
is done in `U` type, and then casts the output back to `T`. However, it adds these casts even
when `U` and `T` are the same. There ends up being many casts from float to float throughout
the model which introduces issues for BYOCG.
   


----------------------------------------------------------------
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



Mime
View raw message