Closed Whytehorse closed 7 years ago
Yes, if you do for example with tf.device('/gpu:0'):
, then it forces everything under that to be on gpu0, or else crash. You can also turn on device placement logging with with tf.Session(config=tf.ConfigProto(log_device_placement=False)) as sess:
Is there a quick way to test my installation to ensure that tensorflow is using the GPU and OpenCL as desired?