![](https://i.imgur.com/9UQ54uc.gif"){kind=small}
thiccboygamerTTV
New member
0
0%
Months of Service
0%
Hello.
Download Miniconda or Anaconda and open the prompt (command line).
Create a new Conda environment with the following command. Our Python version should be 3.10/3.9/3.8. I have tried 3.10 and 3.8, and we will continue with 3.10 in this guide. I recommend continuing with 3.10 to avoid Numpy compatibility issues.
conda create -n gpu python=3.10
We have created a new Conda environment named GPU.
Enter this environment by using the command conda activate gpu.
Install the necessary packages/libraries to use Nvidia CUDA with the command conda install -c conda-forge cudatoolkit=11.2 cudnn=8.1.0.
Upgrade pip with the command pip install --upgrade pip.
Install the latest supported version of TensorFlow for Python 3.10 with the command pip install tensorflow.
Download the prerelease version of TensorFlow with pip install tf-nightly. I'm not certain if this is necessary, but I recommend installing it.
Install DirectML plugins with the command pip install tensorflow-directml-plugin. A brief description of what it is: a hardware accelerator package that enables GPU recognition.
To check if the GPU is detected, you can use the following command. The output should look like this:
python -c "import tensorflow as tf; print(tf.config.list_physical_devices('GPU'))"
I recommend having a good internet connection, as TensorFlow is quite large in size.
I also installed PyTorch on the same environment. The installation command is as follows:
conda install pytorch torchvision torchaudio cudatoolkit=11.3 -c pytorch
At the moment, both are working and detecting the GPU.
I mostly use PyTorch, and it is compatible with Py 3.12. Therefore, I have separate environments for TensorFlow and Torch, and I switch between them depending on which one I am using.
If you're using PyCharm, you need to select this environment as the new interpreter.
Wishing you abundant neurons.
Download Miniconda or Anaconda and open the prompt (command line).
Create a new Conda environment with the following command. Our Python version should be 3.10/3.9/3.8. I have tried 3.10 and 3.8, and we will continue with 3.10 in this guide. I recommend continuing with 3.10 to avoid Numpy compatibility issues.
conda create -n gpu python=3.10
We have created a new Conda environment named GPU.
Enter this environment by using the command conda activate gpu.
Install the necessary packages/libraries to use Nvidia CUDA with the command conda install -c conda-forge cudatoolkit=11.2 cudnn=8.1.0.
Upgrade pip with the command pip install --upgrade pip.
Install the latest supported version of TensorFlow for Python 3.10 with the command pip install tensorflow.
Download the prerelease version of TensorFlow with pip install tf-nightly. I'm not certain if this is necessary, but I recommend installing it.
Install DirectML plugins with the command pip install tensorflow-directml-plugin. A brief description of what it is: a hardware accelerator package that enables GPU recognition.
To check if the GPU is detected, you can use the following command. The output should look like this:
python -c "import tensorflow as tf; print(tf.config.list_physical_devices('GPU'))"
I recommend having a good internet connection, as TensorFlow is quite large in size.
I also installed PyTorch on the same environment. The installation command is as follows:
conda install pytorch torchvision torchaudio cudatoolkit=11.3 -c pytorch
At the moment, both are working and detecting the GPU.
I mostly use PyTorch, and it is compatible with Py 3.12. Therefore, I have separate environments for TensorFlow and Torch, and I switch between them depending on which one I am using.
If you're using PyCharm, you need to select this environment as the new interpreter.
Wishing you abundant neurons.