Skip to content
Merged
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
34 changes: 34 additions & 0 deletions docs/npu_accelerator.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,34 @@
# Using OpenVINO™ Toolkit containers with NPU accelerators

To use the NPU (Neural Processing Unit) Driver inside a container:
- Ensure that the corresponding firmware package `intel-fw-npu` is installed on the host.
- Add NPU device to the container, for example, by using the argument `--device=/dev/accel/accel0` when running your container.

See [NPU Driver release page](https://github.com/intel/linux-npu-driver/releases) for setup details.

## NPU device recovery

There is a known issue with NPU device recovery in the current NPU Driver.
Starting from NPU Linux driver release v1.13.0, a new NPU recovery behavior has been introduced.
Corresponding changes in Ubuntu kernels are expected with new kernel releases.

For more details, see the [known issue page](https://github.com/intel/linux-npu-driver/issues/87).

### Troubleshooting

If inference on the NPU crashes, you can manually reload the driver by running this command in the terminal:

```
sudo rmmod intel_vpu
sudo modprobe intel_vpu
```

## See also

[Intel® NPU Driver GitHub](https://github.com/intel/linux-npu-driver)

[Prebuilt images](#prebuilt-images)

[Working with OpenVINO Containers](docs/containers.md)

[Install Intel® Distribution of OpenVINO™ Toolkit From a Docker Image](https://docs.openvino.ai/2025/get-started/install-openvino/install-openvino-docker-linux.html)