AMD GPU device not found, but GPU test passed

[nolanlad]

System Info:

OS: Ubuntu 16.04 CUDA version: None GPU: Advanced Micro Devices, Inc. [AMD/ATI] Richland [Radeon HD 8650G] TensorFlow Build: Source

I ran the bazel gpu test with the command:

bazel test -c opt --verbose_failures --test_timeout 3600 //tensorflow/python/kernel_tests:basic_gpu_test

which it passed.

However, I am trying to run someone else's code found here:

https://github.com/cysmith/neural-style-tf

command I run is:

bash stylize_image.sh ./image_input/lion.jpg ./styles/kandinsky.jpg

which gives me the error:

tensorflow.python.framework.errors_impl.InvalidArgumentError: Cannot assign a device for operation 'Variable': Operation was explicitly assigned to /device:GPU:0 but available devices are [ /job:localhost/replica:0/task:0/device:CPU:0 ]. Make sure the device specification refers to a valid device. [[Node: Variable = VariableV2container="", dtype=DT_FLOAT, shape=[1,512,512,3], shared_name="", _device="/device:GPU:0"]]

How should I change the code to get it to work with my AMD GPU?

[rodburns]

The latest OpenCL for TensorFlow work now lives here https://github.com/codeplaysoftware/tensorflow You can follow the guide here https://developer.codeplay.com/products/computecpp/ce/guides/tensorflow-guide/tensorflow-generic-setup