[GitHub] lihaofd opened a new pull request #11333: [WIP] fix issue of test_gru_bidirectional #11219 and add robust code

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

lihaofd opened a new pull request #11333: [WIP] fix issue of test_gru_bidirectional #11219 and add robust code
URL: https://github.com/apache/incubator-mxnet/pull/11333
 
 
   ## Description ##
   In this PR, it fixed issue of test_gru_bidirectional #11219 and add robust code. @pengzhao-intel, @TaoLv 
   
   ## Feature changes ##
   ### New features ###
   - Fixed issue of test_gru_bidirectional #11219. The reason is when creating temporary tensor for reshape, the code utilize the unallocated memory because of the incorrect start and end point. Because the data in unallocated memory didn't be used in linalg_gemm. Sometimes the code can pass, sometimes it will crash. We added 100 times checking for test_gru_bidirectional in tests/python/unittests/test_operator.py. And will keep 1 time checking after testing is passed
   - Add robust code like param initialization. It can also make GRU/LSTM Convergence Curve (example/rnn/bucketing/cudnn_rnn_bucketing.py) be stable. 
   - Replace memcpy/memset by using omp for better performance.
   
   ### Unit-test changes ###
   - Tested 100 times checking for test_gru_bidirectional in tests/python/unittests/test_operator.py. And will keep 1 time checking after testing is passed.
   
   ## Checklist ##
    - [X] Passed code style checking (make lint).
    - [X] All changes have test coverage.
    - [ ] Code is well-documented.

----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on 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


With regards,
Apache Git Services