cfRod opened a new pull request #20482:
URL: https://github.com/apache/incubator-mxnet/pull/20482


   ## Description ##
   
   This fix is a workaround for the accuracy issue in the grouped convolution 
unit test observed when MXNet is built with Compute Library for Arm 
Architecture (ACL).
   Fixes https://github.com/apache/incubator-mxnet/issues/20265
   This change includes:
   * Updating MXNet's existing AddSign function to generate unique hashes for 
MKLDNN-ACL backend.
   * Adds a new AddSign overloaded function to generate hashes based on 
uint64_t type.
   * Adding DNNL_AARCH64_USE_ACL definition to CMakeLists.txt
   * Adding Crefeda Rodrigues to the contributors list
   
   ## Checklist ##
   ### Essentials ###
   - [X] PR's title starts with a category (e.g. [BUGFIX], [MODEL], [TUTORIAL], 
[FEATURE], [DOC], etc)
   - [X] Changes are complete (i.e. I finished coding on this PR)
   - [X] All changes have test coverage
   - [X] Code is well-documented
   
   ### Changes ###
   - [X] Updating MXNet's existing AddSign function to generate unique hashes 
for MKLDNN-ACL backend.
   - [X] Adds a new AddSign overloaded function to generate hashes based on 
uint64_t type.
   - [X] Adding DNNL_AARCH64_USE_ACL definition to CMakeLists.txt
   - [X] Adding Crefeda Rodrigues to the contributors list
   
   ## Comments ##
   - This change is a temporary workaround to resolve the accuracy issue seen 
test_operator.test_convolution_grouping until future releases of Compute 
library is compatible with reuse of MKLDNN/oneDNN primitives at the 
framework-level.


-- 
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