bgawrych opened a new pull request, #21106:
URL: https://github.com/apache/incubator-mxnet/pull/21106

   ## Description ##
   Some oneDNN operators are not optimal choice when working on dynamic shaped 
input like in GPT-2. This change add size threshold to determine whether to 
generate JIT kernel or run with native one
   
   Benchmarks on model from gluon-nlp (AWS C6i.x12large) with following OMP 
settings (OMP_PROC_BIND=true OMP_PLACES=sockets OMP_NUM_THREADS=24)
   
   
![image](https://user-images.githubusercontent.com/59644968/181220832-eccbf40b-a012-49d7-b873-dae0c943023d.png)
   
   


-- 
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