Open
Description
Hi all,
The intelanalytics/ipex-llm-inference-cpp-xpu
docker image is ~27GB when unpacked, which is huge, and preventing me from deploying iGPU accelerated Ollama on my server.
There's a few obviously redundancies:
- There's two versions of drivers installed.
- Development packages like git, sudo, vim, less are installed.
- ~4gb of
nvidia-*
python packages are installed (a transitive dependency ofaccelerate
).
Would it be possible to provide more granular docker images, with these redundancies remove?
Thanks
Metadata
Metadata
Assignees
Labels
No labels