Fallback CreateMutex/AtomicIter operators for mkl-dnn #11685
Conversation
|
@wesolwsk,Please review the commit. |
|
You are getting a tensor uninitialized error in data_parallel_model_test.RecurrentNetworkParallelTest.test_equiv_recurrent. Can you take a look? |
|
From log，the issue seems on device_option { device_type: 0 cuda_gpu_id: 3 }，but my patch is for ideep only,It is weird. |
|
cuda_gpu_id indicates the id of the model copy on a node within a DPM instance. It is used for both CPU and GPU runs and should be renamed to something like device_id to avoid this confusion. Here, device_type 0 indicates CPU is being used. |
facebook-github-bot
left a comment
facebook-github-bot
left a comment
There was a problem hiding this comment.
wesolwsk has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator.
|
@PenghuiCheng - could you please rebase |
PenghuiCheng
force-pushed
the
PR-180914
branch
from
c6f60ad to
478825e
Compare
|
@wesolwsk ，had rebased. |
facebook-github-bot
left a comment
facebook-github-bot
left a comment
There was a problem hiding this comment.
wesolwsk has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator.
5 participants
No description provided.