NPU is not recognized in virtual environment

[LOUYWEI]

My device is 255H, my system is ubuntu-24.04, the kernel version is 6.11, and the commands lsmod | grep intel_vpu and modinfo intel_vpu can both show the existence of intel_vpu, and ls /dev/accel/accel0 can also see accel0, and I have also installed Level Zero through deb, why can't npu be recognized in my python virtual environment? The command python -c "from openvino import Core; print(Core().available_devices)" only outputs ['CPU', 'GPU'],What else do I need to do to correctly identify the NPU？

[jwludzik]

Hi, have you check your user access to accel0 device? I mean you have to be in "render" group. Check the first point in docs/overview.md#Troubleshooting

If it is not your case, then use strace to find a root cause:

$ strace -o strace.log --trace=file python -c "from openvino import Core; print(Core().available_devices)"
...
$ grep -E "(accel|libnpu_|libze_)" strace.log
# Below is partial output with most important stuff required by NPU driver
# Please, keep in mind that libze_intel_vpu.so.1 is old driver and should not be installed in system. libze_intel_npu.so.1 should be used
...
openat(AT_FDCWD, "/lib/x86_64-linux-gnu/libze_loader.so.1", O_RDONLY|O_CLOEXEC) = 3
....
openat(AT_FDCWD, "/usr/lib/x86_64-linux-gnu/libze_intel_vpu.so.1", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
...
openat(AT_FDCWD, "/lib/x86_64-linux-gnu/libze_intel_npu.so.1", O_RDONLY|O_CLOEXEC) = 3
...
openat(AT_FDCWD, "/dev/accel/accel0", O_RDWR|O_NOFOLLOW|O_CLOEXEC) = 3
...
openat(AT_FDCWD, "/dev/accel/accel63", O_RDWR|O_NOFOLLOW|O_CLOEXEC) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "/lib/x86_64-linux-gnu/libnpu_driver_compiler.so", O_RDONLY|O_CLOEXEC) = 3
...

Thanks to it you should be able to find an issue with your setup

[LOUYWEI]

@jwludzik Thanks for your help, I'll give it a try

[hteeyeoh]

Hi @jwludzik,

I came into this thread as i was doing the same thing after follow the troubleshooting steps mentioned above to include user in the "render" or "video" group. Now when i run ov.core.available_devices and print the returned device list:

print(core.available_devices)
['CPU', 'NPU']

I can now see the NPU listed in the returned list. So this mean my NPU device is ready to use right? Any additional steps needed if I wanna try run optimum with openvino for llm model and run on NPU device? Just to check

[jwludzik]

No, it should work out of the box. Please, check OpenVINO documentation for mode details how to run LLM in NPU: https://docs.openvino.ai/2025/openvino-workflow-generative/inference-with-genai/inference-with-genai-on-npu.html

[hteeyeoh]

No, it should work out of the box. Please, check OpenVINO documentation for mode details how to run LLM in NPU: https://docs.openvino.ai/2025/openvino-workflow-generative/inference-with-genai/inference-with-genai-on-npu.html

Alright. Thanks a lot. Will try it out.

[hteeyeoh]

No, it should work out of the box. Please, check OpenVINO documentation for mode details how to run LLM in NPU: https://docs.openvino.ai/2025/openvino-workflow-generative/inference-with-genai/inference-with-genai-on-npu.html

Hi @jwludzik , I tried with the url you shared it work. Another question, does openvino on npu only supported via method mentioned in the url which is using openvino_genai?

As currently my application are initially using HuggingFacePipeline.from_model_id to initialize/load the model with CPU/GPU. This is not working with NPU? Am I correct?

[jwludzik]

@hteeyeoh, I do not know. Best would be to seek help in OpenVINO. From me I can only share another OpenVINO documentation page: https://docs.openvino.ai/2025/openvino-workflow-generative/inference-with-optimum-intel.html#inference-with-optimum-intel . Hope it helps

[hteeyeoh]

@hteeyeoh, I do not know. Best would be to seek help in OpenVINO. From me I can only share another OpenVINO documentation page: https://docs.openvino.ai/2025/openvino-workflow-generative/inference-with-optimum-intel.html#inference-with-optimum-intel . Hope it helps

Hi @jwludzik , no issue. Just asking on the fly since having this thread. Ya the URL you shared it already included in our pipeline where convert the huggingface model to openvinoIR. We have been using this on CPU and GPU device and initialize them via langchain-huggingface pipeline module and include into our RAG chain. Now we are exploring on the possibilities on NPU device.