can not assign GPU device for operation about deepfacelab_linux

iperov updated the core and it doesnt work on linux anymore using nagadit's instructions. use this command instead to create the environment:

conda create -n deepfacelab -c main python=3.7 cudnn=7.6.5 cudatoolkit=10.1.243

and then downgrade tensorflow-gpu to 2.3.1

python -m pip install tensorflow-gpu==2.3.1

it should now work

[mostafakhaki]

Thanks, it works for me.

I use the following packages:
cudatoolkit=10.1.243
python=3.7
cudnn=7.6.5
tqdm
numpy==1.16.0
h5py==2.10.0
opencv-python==4.1.0.25
ffmpeg-python==0.1.17
scikit-image==0.14.2
scipy==1.4.1
colorama
tensorflow-gpu==2.3.1
pyqt5

[YoochanMyung]

OS: Ubuntu 20.04
NVIDIA-Driver : 510.47.03
CUDA version: 11.6
GPU: 3090Ti

The following settings work for the above environment.

conda create -n deepfacelab -c main python=3.7 cudnn=8 cudatoolkit=11
conda activate deepfacelab
python -m pip install -r ./DeepFaceLab/requirements-cuda.txt

[requirements-cuda.txt]
tqdm
numpy==1.19.3
numexpr
h5py==3.1.0
opencv-python==4.1.0.25
ffmpeg-python==0.1.17
scikit-image==0.14.2
scipy==1.4.1
colorama
tensorflow-gpu==2.5.0
pyqt5
tf2onnx==1.8.4*

[TroyHernandez]

numpy == 1.16.1

1.16.0 has a memory leak.

[Flatheadman]

It's not just always downgrade. sometimes you need upgrade.
My cuda 11.3 version needs a tensorflow 2.6.0 to use gpu.
Check your proper version from the table here: https://www.tensorflow.org/install/source

[musketeer38]

OS: Ubuntu 20.04 NVIDIA-Driver : 510.47.03 CUDA version: 11.6 GPU: 3090Ti

The following settings work for the above environment.

conda create -n deepfacelab -c main python=3.7 cudnn=8 cudatoolkit=11
conda activate deepfacelab
python -m pip install -r ./DeepFaceLab/requirements-cuda.txt

[requirements-cuda.txt]
tqdm
numpy==1.19.3
numexpr
h5py==3.1.0
opencv-python==4.1.0.25
ffmpeg-python==0.1.17
scikit-image==0.14.2
scipy==1.4.1
colorama
tensorflow-gpu==2.5.0
pyqt5
tf2onnx==1.8.4*

Thanks!! this procedure worked for me. With the original instructions, my GPU was not seen.
I just removed the specific version for tf2onnx.

My env:
GPU: Nvidia RTX 3060 laptop
SW: Ubuntu 22.04, Nvidia Driver Version: 515.65.01, CUDA Version: 11.7

[eliliam]

@YoochanMyung Conda doesn't seem to have opencv-python==4.1.0.25, should I install the latest version of opencb-python(4.6.0.66) instead?

Edit: I just realized I was on Python 3.9 instead of 3.7, is it possible to run DFL on Python 3.9?