[tests] Update Dockerfile to use cuda 12.2

[younesbelkada]

An attempt to fix the current example tests being skipped - according to the logs:

./../../../opt/conda/envs/peft/lib/python3.8/site-packages/torch/cuda/__init__.py:138
  /opt/conda/envs/peft/lib/python3.8/site-packages/torch/cuda/__init__.py:138: UserWarning: CUDA initialization: The NVIDIA driver on your system is too old (found version 11080). Please update your GPU driver by downloading and installing a new version from the URL: http://www.nvidia.com/Download/index.aspx Alternatively, go to: https://pytorch.org/ to install a PyTorch version that has been compiled with your version of the CUDA driver. (Triggered internally at ../c10/cuda/CUDAFunctions.cpp:108.)
    return torch._C._cuda_getDeviceCount() > 0

../../../../opt/conda/envs/peft/lib/python3.8/site-packages/bitsandbytes/cextension.py:34
  /opt/conda/envs/peft/lib/python3.8/site-packages/bitsandbytes/cextension.py:34: UserWarning: The installed version of bitsandbytes was compiled without GPU support. 8-bit optimizers, 8-bit multiplication, and GPU quantization are unavailable.
    warn("The installed version of bitsandbytes was compiled without GPU support. "

Will investigate more and report back here

cc @BenjaminBossan

[HuggingFaceDocBuilderDev]

The documentation is not available anymore as the PR was closed or merged.

[younesbelkada]

After some investigation, that seems to be indeed the culprit, I think torch 2.1 somehow fails with old CUDA versions leading to torch.cuda.is_available() to be set to False. I also had to add NVIDIA_DISABLE_REQUIRE: "1" (ref: NVIDIA/nvidia-container-toolkit#148) so that the docker image can run without the check of nvidia versions

[BenjaminBossan]

Thanks so much for investigating and fixing the issue, LGTM