latest-aio-gpu-intel-f32 docker not working on the arc a750

[maxvaneck]

LocalAI version:
docker image with the tag :latest-aio-gpu-intel-f32

Environment, CPU architecture, OS, and Version:
Linux Zana 6.6.68-Unraid #1 SMP PREEMPT_DYNAMIC Tue Dec 31 13:42:37 PST 2024 x86_64 11th Gen Intel(R) Core(TM) i5-11400 @ 2.60GHz GenuineIntel GNU/Linux

64 GB RAM

Describe the bug
gpu does not work nor do i get any output from chat. however the tag[master-vulkan-ffmpeg-core] works just fine with the gpu

To Reproduce
just run the docker latest-aio-gpu-intel-f32 while passing through /dev/dri and use localai

Expected behavior
gpu accelerated chatting

Logs

logs.txt

Additional context
i use the a750 with 8 GB vram which is not a lot but at least something

[kettly1260]

I am using latest-aio-gpu-intel-f16 docker with I5-12400 Igpu, and the model cannot be loaded normally, but it is normal to switch to pure CPU

5:19PM INF [/build/backend/python/coqui/run.sh] Fails: failed to load model with internal loader: backend not found: /tmp/localai/backend_data/backend-assets/grpc/build/backend/python/coqui/run.sh
5:19PM INF [/build/backend/python/autogptq/run.sh] Attempting to load
5:19PM INF BackendLoader starting backend=/build/backend/python/autogptq/run.sh modelID=deepseek-r1-distill-qwen-7b o.model=DeepSeek-R1-Distill-Qwen-7B-Q4_K_M.gguf
5:19PM INF [/build/backend/python/autogptq/run.sh] Fails: failed to load model with internal loader: backend not found: /tmp/localai/backend_data/backend-assets/grpc/build/backend/python/autogptq/run.sh
5:19PM INF [/build/backend/python/vllm/run.sh] Attempting to load
5:19PM INF BackendLoader starting backend=/build/backend/python/vllm/run.sh modelID=deepseek-r1-distill-qwen-7b o.model=DeepSeek-R1-Distill-Qwen-7B-Q4_K_M.gguf
5:19PM INF [/build/backend/python/vllm/run.sh] Fails: failed to load model with internal loader: backend not found: /tmp/localai/backend_data/backend-assets/grpc/build/backend/python/vllm/run.sh

[maxvaneck]

i am using an intel arc a750 as it shows in the logs

[KimSHenriksen]

Have you looked at #3967? In the log.txt you have a similar error

stderr No kernel named ZTSZZL17rms_norm_f32_syclPKfPfiifPN4sycl3_V15queueEiENKUlRNS3_7handlerEE0_clES7_EUlNS3_7nd_itemILi3EEEE was foundException

[maxvaneck]

it seems they have problems with outdated dependecies. i'll look if a apt update/upgrade inside the container fixes stuff for me

[KimSHenriksen]

If the host's oneapi works with your arc gpu it might be possible to add '-v /opt/intel/oneapi:/opt/intel/oneapi' to the docker container and use the host's library. That'll save the update/rebuild step.

[maxvaneck]

that would work, unfortunately i use unraid which has no package manager and the root filesystem resets every reboot except some config files of course

[maxvaneck]

i updated the packages inside the container which was quite a lot but no dice