-
Notifications
You must be signed in to change notification settings - Fork 6.8k
Fused RNN Operators have nonsupport of add
grad_req with mkl-dnn
#16578
Comments
@mxnet-label-bot add [MKLDNN] |
Thanks for your suggestion. MKL-DNN RNN operators do have this issue, and it will terminate the program when meets |
@zixuanweeei does the issue resolve by MKLDNN upgrade PR? |
This issue exists in the upgrade PR. It will give an error message when uses |
OK, is there a workaround for the user and what's the plan for the next step? |
It will be appreciated that users could report their usage with |
Should be fixed via #17075. Feel free to re-open if the problem is still there. @zixuanweeei |
Currently, we have not integrated the
add
grad_req routine inFusedRNNCell
with mkl-dnn fusion. It will be highly appreciated if anyone could tell us the application scenario usingadd
during the training process.FYI, #568 and #725. When weights are updated every batch,
add
andwrite
will produce the same results.The text was updated successfully, but these errors were encountered: