You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@tvm.apache.org by GitBox <gi...@apache.org> on 2022/09/15 01:21:02 UTC

[GitHub] [tvm] yangulei opened a new pull request, #12787: [AMP] refine AMP and the corresponding tests for bfloat16

yangulei opened a new pull request, #12787:
URL: https://github.com/apache/tvm/pull/12787

   This PR fixes issue https://github.com/apache/tvm/issues/12763, where some OP are marked to keep the original dtype but some of its input is `bfloat16` while a `Cast` is missing.
   The AMP tests have also been refined to cover `bfloat16` **without** accuracy checking.


-- 
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

[GitHub] [tvm] masahi merged pull request #12787: [AMP] refine AMP and the corresponding tests for bfloat16

Posted by GitBox <gi...@apache.org>.
masahi merged PR #12787:
URL: https://github.com/apache/tvm/pull/12787


-- 
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

[GitHub] [tvm] billishyahao commented on pull request #12787: [AMP] refine AMP and the corresponding tests for bfloat16

Posted by GitBox <gi...@apache.org>.
billishyahao commented on PR #12787:
URL: https://github.com/apache/tvm/pull/12787#issuecomment-1247713331

   Thanks for the patch, Youlei! I found a bunch of statement like "op->dtype.is_float() || op->dtype.is_bfloat16()" in tvm folder
   
   Shall we simply add new float type definition in tvm/include/tvm/runtime/data_type.h to eliminate those statements?
   
   ```
     /*! \return whether type is a general float type, including float/float16/bfloat16. */
     bool is_general_float() const { return is_float() || is_bfloat16(); }


-- 
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

[GitHub] [tvm] yangulei commented on pull request #12787: [AMP] refine AMP and the corresponding tests for bfloat16

Posted by GitBox <gi...@apache.org>.
yangulei commented on PR #12787:
URL: https://github.com/apache/tvm/pull/12787#issuecomment-1293170957

   @masahi Could you help to review this? Thanks.


-- 
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

[GitHub] [tvm] yangulei commented on pull request #12787: [AMP] refine AMP and the corresponding tests for bfloat16

Posted by GitBox <gi...@apache.org>.
yangulei commented on PR #12787:
URL: https://github.com/apache/tvm/pull/12787#issuecomment-1248794416

   @billishyahao 
   I agree that `is_general_float()` could make the code **cleaner**, but **not clearer**. `general float` is a broader concept rather than IEEE float point plus bfloat16, for example, `TensorFloat-32` is also a `general float`. I prefer to keep the expr like `op->dtype.is_float() || op->dtype.is_bfloat16()` as its more clear and specific.


-- 
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

[GitHub] [tvm] billishyahao commented on pull request #12787: [AMP] refine AMP and the corresponding tests for bfloat16

Posted by GitBox <gi...@apache.org>.
billishyahao commented on PR #12787:
URL: https://github.com/apache/tvm/pull/12787#issuecomment-1247673401

   @tvm-bot rerun


-- 
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