blakeblackshear.frigate/docker/rocm
2024-01-26 11:11:56 +02:00
..
migraphx ROCm AMD/GPU based build and detector, WIP 2024-01-25 20:14:22 +02:00
Dockerfile docker/rocm: make imported model files readable by all 2024-01-26 10:54:24 +02:00
README.md docker/rocm: readme about running on AMD GPUs 2024-01-26 11:11:56 +02:00
rocm-pin-600 ROCm AMD/GPU based build and detector, WIP 2024-01-25 20:14:22 +02:00
rocm.hcl ROCm AMD/GPU based build and detector, WIP 2024-01-25 20:14:22 +02:00
rocm.list ROCm AMD/GPU based build and detector, WIP 2024-01-25 20:14:22 +02:00
rocm.mk ROCm AMD/GPU based build and detector, WIP 2024-01-25 20:14:22 +02:00

AMD/ROCm GPU based detector

Docker settings

AMD/ROCm needs access to /dev/kfd and /dev/dri. When running as user also needs the video group. Sometimes also needs the renderandssh/_ssh` groups. For reference/comparison see running ROCm PyTorch Docker image.

$ docker run -it --device=/dev/kfd --device=/dev/dri --group-add video \
    ...

Frigate configuration

An example of a working configuration:

model:
  path: /yolov8n_320x320.onnx
  labelmap_path: /yolov8s_labels.txt
  model_type: yolov8
detectors:
  rocm:
    type: rocm