GPU Version not working with CUDA 10

[mdinoulis]

I'm getting the following error loading TensorFlow with the GPU version, but it works ok with the CPU version. I'm running with CUDA 10.0 and cuDNN v7.6.5 on Windows 10 with an RTX Titan GPU. TensorFlow is running fine elsewhere on my machine under Anaconda

Is it the CUDA 10.0 that is the problem? I note that your spec suggests CUDA 9.0. If so, will there be a GPU version available that runs on CUDA 10.x ?

import tensorflow
Traceback (most recent call last):
File "C:\Users\mdinoulis\AppData\Roaming\Python\Python36\site-packages\tensorflow\python\pywrap_tensorflow.py", line 58, in
from tensorflow.python.pywrap_tensorflow_internal import *
File "C:\Users\mdinoulis\AppData\Roaming\Python\Python36\site-packages\tensorflow\python\pywrap_tensorflow_internal.py", line 28, in
_pywrap_tensorflow_internal = swig_import_helper()
File "C:\Users\mdinoulis\AppData\Roaming\Python\Python36\site-packages\tensorflow\python\pywrap_tensorflow_internal.py", line 24, in swig_import_helper
_mod = imp.load_module('_pywrap_tensorflow_internal', fp, pathname, description)
File "imp.py", line 243, in load_module
File "imp.py", line 343, in load_dynamic
ImportError: DLL load failed: The specified module could not be found.
Failed to load the native TensorFlow runtime.
See https://www.tensorflow.org/install/errors
for some common reasons and solutions. Include the entire stack trace
above this error message when asking for help.
Traceback (most recent call last):
File "C:\Users\mdinoulis\AppData\Roaming\Python\Python36\site-packages\tensorflow\python\pywrap_tensorflow.py", line 58, in
from tensorflow.python.pywrap_tensorflow_internal import *
File "C:\Users\mdinoulis\AppData\Roaming\Python\Python36\site-packages\tensorflow\python\pywrap_tensorflow_internal.py", line 28, in
_pywrap_tensorflow_internal = swig_import_helper()
File "C:\Users\mdinoulis\AppData\Roaming\Python\Python36\site-packages\tensorflow\python\pywrap_tensorflow_internal.py", line 24, in swig_import_helper
mod = imp.load_module('pywrap_tensorflow_internal', fp, pathname, description)
File "imp.py", line 243, in load_module
File "imp.py", line 343, in load_dynamic
ImportError: DLL load failed: The specified module could not be found.
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "", line 1, in
File "C:\Users\mdinoulis\AppData\Roaming\Python\Python36\site-packages\tensorflow_init.py", line 24, in
from tensorflow.python import pywrap_tensorflow # pylint: disable=unused-import
File "C:\Users\mdinoulis\AppData\Roaming\Python\Python36\site-packages\tensorflow\python_init.py", line 49, in
from tensorflow.python import pywrap_tensorflow
File "C:\Users\mdinoulis\AppData\Roaming\Python\Python36\site-packages\tensorflow\python\pywrap_tensorflow.py", line 74, in
raise ImportError(msg)
ImportError: Traceback (most recent call last):
File "C:\Users\mdinoulis\AppData\Roaming\Python\Python36\site-packages\tensorflow\python\pywrap_tensorflow.py", line 58, in
from tensorflow.python.pywrap_tensorflow_internal import *
File "C:\Users\mdinoulis\AppData\Roaming\Python\Python36\site-packages\tensorflow\python\pywrap_tensorflow_internal.py", line 28, in
_pywrap_tensorflow_internal = swig_import_helper()
File "C:\Users\mdinoulis\AppData\Roaming\Python\Python36\site-packages\tensorflow\python\pywrap_tensorflow_internal.py", line 24, in swig_import_helper
_mod = imp.load_module('_pywrap_tensorflow_internal', fp, pathname, description)
File "imp.py", line 243, in load_module
File "imp.py", line 343, in load_dynamic
ImportError: DLL load failed: The specified module could not be found.
Failed to load the native TensorFlow runtime.
See https://www.tensorflow.org/install/errors
for some common reasons and solutions. Include the entire stack trace
above this error message when asking for help.

[getnamo]

There are two ways you can work around the embedded python build to use the latest python setup of your choice:

1. Change UseThirdPartyPython https://github.com/getnamo/UnrealEnginePython/blob/master/Source/UnrealEnginePython/UnrealEnginePython.Build.cs#L13 to false and specify your python install location in e.g. https://github.com/getnamo/UnrealEnginePython/blob/master/Source/UnrealEnginePython/UnrealEnginePython.Build.cs#L28 and compile. This will use your specified python installation instead of the one included in the plugin, which can have whatever setup you're can run locally.
2. Use the remote variant of the plugin: https://github.com/getnamo/machine-learning-remote-ue4 which can communicate to a python installation anywhere (local folder, or remote server). Your python installation just has to support: https://github.com/getnamo/ml-remote-server wrapper (e.g. python 3.7)

Most likely reason for current blocker on the embedded build is the raw embedded python 3.6 zip which doesn't appear to support newer Tensorflow than r1.12 (https://github.com/getnamo/tensorflow-ue4/blob/master/Content/Scripts/upymodule.json). An update to this to support e.g. TF2.0 out of the box would be the ideal fix and add support for newer cuda.