Unable to start GPU-assisted computing

[RobinChen3]

We are testing the TinyImageNet example, with a GPU card (Nvidia 1050ti) installed on the host, and a value set in cuda_devices:[0] in envoy_config.yaml. However, still can't start GPU assisted computing, any suggestions?

[mansishr]

Hi @RobinChen3 , could you try nvidia-smi to check the list of available GPUs? Also, if you are trying out this example, the device is explicitly set to 'cpu' in train and validate tasks so you must comment those lines.