Closed
Description
Current we require two separate arguments to the device_context to specify which queue to get from dppl. Maybe we should move to a string based format e.g. gpu:0 like what other libraries like TensorFlow do.
{{ message }}
Current we require two separate arguments to the device_context to specify which queue to get from dppl. Maybe we should move to a string based format e.g. gpu:0 like what other libraries like TensorFlow do.