awslabs / keras-apache-mxnet

[DEPRECATED] Amazon Deep Learning's Keras with Apache MXNet support
https://github.com/awslabs/keras-apache-mxnet/wiki
Other
289 stars 65 forks source link

fix gpu detection #255

Closed roywei closed 4 years ago

roywei commented 4 years ago

Summary

From MXNet 1.5, the detection of number of GPUs changed from using command linenvidia-smi to MXNet C API call using CUDA.

As a result, keras-mxnet should catch both CalledProcessError and MXNetError

Related Issues

PR Overview

sandeep-krishnamurthy commented 4 years ago

When will this error be thrown?

roywei commented 4 years ago

When will this error be thrown?

@sandeep-krishnamurthy when GPU driver not installed or driver version and cuda version mismatch.

We used CalledProcessError to catch nvidia-smi failure before, but now the error type changed to MXNetError in MXNet 1.5