How to troubleshoot the error when running a tabby container in Docker Desktop?

I am getting the following error while running the tabby container in Docker Desktop:

docker run -it --gpus all -p 8080:8080 -v c:/tabby_home/.tabby:/data tabbyml/tabby serve --model TabbyML/StarCoder-1B --device CUDA
docker: Error response from daemon: failed to create task for container: failed to create shim task: OCI runtime create failed: runc create failed: unable to start container process: error during container init: error running hook #0: error running hook: exit status 1, stdout: , stderr: Auto-detected mode as 'legacy'
nvidia-container-cli: initialization error: load library failed: cannot open shared object file: no such file or directory: unknown.

Can someone help me to troubleshoot this issue?


Lakshmi Gowda

Asked on Jan 22, 2024

To troubleshoot the error when running a tabby container in Docker Desktop, you can follow these steps:

  1. Install the nvidia-container-toolkit library on the host computer. This library is used by the container as a link to the GPU.

  2. Make sure you have installed Docker from apt on Ubuntu, and not with snap. Docker needs direct access to the host computer.

  3. Verify the installation by running the following commands:

    • nvidia-smi inside the host to check if the NVIDIA GPU is detected.
    • nvcc --help inside the Docker container to check if the NVIDIA CUDA compiler is available.

If these steps do not resolve the issue, please provide more information about your setup and any other error messages you encounter.

Jan 22, 2024Edited by