Suggest way to use latest pytorch (2.2.2) about gpu-jupyter

[fzyzcjy]

Thanks! I will have a look at it when next time working on infra.

[ChristophSchranz]

Hi @fzyzcjy ,
the suggested way to update a package within the container is pip install [package]==[version].
The flag -U will upgrade the package to the latest version that might not support the current CUDA version pip flag -U.

[fzyzcjy]

Thank you! Then another question is, how should I upgrade CUDA - should I wait for the image or there are other methods?

[benz0li]

@fzyzcjy You could use my/b-data's GPU accelerated Python images:

glcr.b-data.ch/jupyterlab/cuda/python/scipy:3.12

• derived from nvidia/cuda:12.4.1-devel-ubuntu22.04
  • including a custom cuDNN 8 installation

Install TensorFlow: pip install tensorflow

Install PyTorch: pip install torch torchvision torchaudio

---

What is different about my/b-data's images:

1. Multi-arch: linux/amd64, linux/arm64/v8
2. Derived from nvidia/cuda:12.4.1-devel-ubuntu22.04
   • including development libraries and headers
3. TensortRT and TensorRT plugin libraries
   • including development libraries and headers
4. IDE: code-server next to JupyterLab
5. Just Python – no Conda / Mamba

---

There are also Python 3.11 images, which are still derived from nvidia/cuda:11.8.0-cudnn8-devel-ubuntu22.04:

glcr.b-data.ch/jupyterlab/cuda/python/scipy:3.11

• derived from nvidia/cuda:11.8.0-cudnn8-devel-ubuntu22.04

Install TensorFlow: pip install tensorflow==2.14.1

Install PyTorch: pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu118

---

Cross reference:

• b-data/jupyterlab-python-docker-stack#8