[Config Support]: Trying to get 0.16 working with NVIDIA

[brandondb1]

Describe the problem you are having

I Have a GTX 750 that has successfully run Frigate Tensorrt for a while now. I upgraded to 0.16 and am now trying to get the ONNX detector working. I followed the directions to create the yolov7-320 model, and it created properly, but when I start Frigate I get an error:

Unsupported model IR version: 11, max supported IR version: 10

There doesn't seem to be an option in the python converter to specify this, and honestly I'm not sure if it's inherent in the yolo download or in the conversion. Any help would be appreciated as my Frigate will currently not start. I can roll it back but after spending the better part of half a day on this I'd just like to get it working.

Version

0.16

Frigate config file

mqtt:
  enabled: true
  host: 192.168.1.xx
  user: xxxxxx
  password: xxxxxxx
objects:
  track:
    - person
    - car
    - cat
go2rtc:
  streams:
    Driveway:
      - ffmpeg:http://192.168.1.xx/flv?port=1935&app=bcs&stream=channel0_main.bcs&user=admin&password=xxxx
    DrivewaySub:
      - ffmpeg:http://192.168.1.xx/flv?port=1935&app=bcs&stream=channel0_ext.bcs&user=admin&password=xxxxx
    BedroomStairs:
      - ffmpeg:http://192.168.1.xx/flv?port=1935&app=bcs&stream=channel0_main.bcs&user=admin&password=xxxx
    BedroomStairsSub:
      - ffmpeg:http://192.168.1.xx/flv?port=1935&app=bcs&stream=channel0_ext.bcs&user=admin&password=xxxx
    Porch:
      - ffmpeg:http://192.168.1.xx/flv?port=1935&app=bcs&stream=channel0_main.bcs&user=admin&password=xxxx
    PorchSub:
      - ffmpeg:http://192.168.1.xx/flv?port=1935&app=bcs&stream=channel0_ext.bcs&user=admin&password=xxxxx
    EastYard:
      - ffmpeg:http://192.168.1.xx/flv?port=1935&app=bcs&stream=channel0_main.bcs&user=admin&password=xxxx
    EastYardSub:
      - ffmpeg:http://192.168.1.xx/flv?port=1935&app=bcs&stream=channel0_ext.bcs&user=admin&password=xxxxx
    MainDeck:
      - rtsp://admin:xxxxxx@192.168.1.xx/cam/realmonitor?channel=1&subtype=00&authbasic=YWRtaW46YWRtaW4=
    MainDeckSub:
      - rtsp://admin:xxxxxx@192.168.1.xx/cam/realmonitor?channel=1&subtype=01&authbasic=YWRtaW46YWRtaW4=
    LivingRoom:
      - rtsp://admin:xxxxxx@192.168.1.xx:554/h264Preview_01_main
    LivingRoomSub:
      - rtsp://admin:xxxxxx@192.168.1.xx:554/h264Preview_01_sub


cameras:
  Porch:
    enabled: true
    ffmpeg:
      inputs:
        - path: rtsp://127.0.0.1:8554/Porch
          input_args: preset-rtsp-restream
          hwaccel_args: preset-nvidia-h264
          roles:
            - record
        - path: rtsp://127.0.0.1:8554/PorchSub
          input_args: preset-rtsp-restream
          hwaccel_args: preset-nvidia-h264
          roles:
            - detect
    detect:
      enabled: true
      fps: 5
    motion:
      threshold: 45
      contour_area: 30
      mask:
        - 0.602,0.165,0.604,0.27,0.664,0.275,0.665,0.169
        - 0.001,0.272,0.075,0.174,0.065,0,-0.001,-0.002
        - 0.69,0.518,0.955,0.588,0.904,0.856,0.696,0.803,0.661,0.599
      improve_contrast: true
    objects:
      track:
        - person
        - car

      mask: 0,-0.002,0,0.527,0.131,0.463,0.08,0
      filters:
        car: {}
    zones:
      porch_area:
        coordinates: 1,0.947,1,0.002,0.557,0.015,0.569,0.61,0.419,0.851
        inertia: 3
        loitering_time: 0
      garage_area:
        coordinates: 
          0,1,0.337,0.977,0.565,0.61,0.555,0,0.079,0,0.131,0.462,0,0.527
        inertia: 3
        loitering_time: 0
    record:
      enabled: true
      retain:
        days: 5
        mode: motion
      alerts:
        retain:
          days: 10
      detections:
        retain:
          days: 10
    live:
      streams:
        Porch: Porch
    review:
      alerts:
        required_zones:
          - garage_area
          - porch_area
      detections:
        required_zones:
          - porch_area
          - garage_area
  Driveway:
    enabled: true
    ffmpeg:
      inputs:
        - path: rtsp://127.0.0.1:8554/Driveway?video=copy
          input_args: preset-rtsp-restream
          hwaccel_args: preset-nvidia-h264
          roles:
            - record
        - path: rtsp://127.0.0.1:8554/DrivewaySub?video=copy
          input_args: preset-rtsp-restream
          hwaccel_args: preset-nvidia-h264
          roles:
            - detect
    detect:
      enabled: true
      fps: 5
      stationary:
        interval: 100
        threshold:
    objects:
      track:
        - person
        - car
    zones:
      driveway_area:
        coordinates: 
          0.257,0.378,0.547,0.353,0.777,0.358,1,0.369,1,0.382,1,0.594,0.912,0.817,0.814,1,0.592,1,0.392,0.967,0.182,0.927,0,0.848,0,0.397
        inertia: 3
        loitering_time: 0
    motion:
      mask: 0,0,1,0,1,0.367,0.79,0.359,0.563,0.351,0.262,0.379,0,0.395
      threshold: 50
      contour_area: 30
    record:
      enabled: true
      retain:
        days: 5
        mode: motion
      alerts:
        retain:
          days: 10
      detections:
        retain:
          days: 10
    live:
      streams:
        Driveway: Driveway
    review:
      alerts:
        required_zones: driveway_area
      detections:
        required_zones: driveway_area
  BedroomStairs:
    enabled: true
    ffmpeg:
      inputs:
        - path: rtsp://127.0.0.1:8554/BedroomStairs?video=copy&audio=aac
          input_args: preset-rtsp-restream
          hwaccel_args: preset-nvidia-h264
          roles:
            - record
        - path: rtsp://127.0.0.1:8554/BedroomStairsSub?video=copy
          input_args: preset-rtsp-restream
          hwaccel_args: preset-nvidia-h264
          roles:
            - detect
    objects:
      track:
        - person
    live:
      streams:
        BedroomStairs: BedroomStairs
    detect:
      enabled: true
      fps: 5
    motion:
      threshold: 50
      contour_area: 30
    record:
      enabled: true
      retain:
        days: 5
        mode: motion
      alerts:
        retain:
          days: 10
      detections:
        retain:
          days: 10
    birdseye:
      enabled: true
      mode: objects
  EastYard:
    enabled: true
    ffmpeg:
      inputs:
        - path: rtsp://127.0.0.1:8554/EastYard?video=copy&audio=aac
          input_args: preset-rtsp-restream
          hwaccel_args: preset-nvidia-h264
          roles:
            - record
        - path: rtsp://127.0.0.1:8554/EastYardSub?video=copy
          input_args: preset-rtsp-restream
          hwaccel_args: preset-nvidia-h264
          roles:
            - detect
    objects:
      track:
        - person
        - cat
    live:
      streams:
        EastYard: EastYard
    detect:
      enabled: true
      fps: 5
    motion:
      threshold: 50
      contour_area: 30
    record:
      enabled: true
      retain:
        days: 5
        mode: motion
      alerts:
        retain:
          days: 10
      detections:
        retain:
          days: 10
  LivingRoom:
    enabled: true
    ffmpeg:
      inputs:
        - path: rtsp://127.0.0.1:8554/LivingRoom?video=copy&audio=aac
          input_args: preset-rtsp-restream
          hwaccel_args: preset-nvidia-h264
          roles:
            - record
        - path: rtsp://127.0.0.1:8554/LivingRoomSub?video=copy
          input_args: preset-rtsp-restream
          hwaccel_args: preset-nvidia-h264
          roles:
            - detect
    detect:
      enabled: true
      fps: 5
    objects:
      track:
        - person
    record:
      enabled: true
      retain:
        days: 5
        mode: motion
      alerts:
        retain:
          days: 10
      detections:
        retain:
          days: 10
    live:
      streams:
        LivingRoom: LivingRoom
  MainDeck:
    enabled: true
    ffmpeg:
      inputs:
        - path: rtsp://127.0.0.1:8554/MainDeck
          input_args: preset-rtsp-restream
          hwaccel_args: preset-nvidia-h264
          roles:
            - record
        - path: rtsp://127.0.0.1:8554/MainDeckSub
          input_args: preset-rtsp-restream
          hwaccel_args: preset-nvidia-h264
          roles:
            - detect
    detect:
      enabled: true
      fps: 5
    motion:
      threshold: 50
      contour_area: 30
    objects:
      track:
        - person
        - cat
    record:
      enabled: true
      retain:
        days: 5
        mode: motion
      alerts:
        retain:
          days: 10
      detections:
        retain:
          days: 10
    live:
      streams:
        MainDeck: MainDeck
detectors:
  onnx:
    type: onnx

model:
  model_type: yolo-generic
  path: /config/model_cache/onnx/yolov7-320.onnx
  input_tensor: nchw
  input_dtype: float
  input_pixel_format: bgr
  width: 320
  height: 320


ui:
  timezone: America/New_York

auth:
  enabled: false


version: 0.16-0
detect:
  enabled: true

Relevant Frigate log output

2025-08-20 20:35:01.974538037  onnxruntime.capi.onnxruntime_pybind11_state.Fail: [ONNXRuntimeError] : 1 : FAIL : Load model from /config/model_cache/onnx/yolov7-320.onnx failed:/onnxruntime_src/onnxruntime/core/graph/model.cc:180 onnxruntime::Model::Model(onnx::ModelProto&&, const onnxruntime::PathString&, const onnxruntime::IOnnxRuntimeOpSchemaRegistryList*, const onnxruntime::logging::Logger&, const onnxruntime::ModelOptions&) Unsupported model IR version: 11, max supported IR version: 10

Relevant go2rtc log output

none

Frigate stats

No response

Operating system

Other

Install method

Docker Compose

docker-compose file or Docker CLI command

services:
  frigate:
    cap_add:
      - CAP_PERFMON
    container_name: frigate4
    deploy:
      resources:
        reservations:
          devices:
            - capabilities:
                - gpu
              count: 1
              driver: nvidia
    environment:
      - LOG_LEVEL=${LOG_LEVEL}
    image: ghcr.io/blakeblackshear/frigate:0.16.0-tensorrt
    ports:
      - '5000:5000'
      - '8554:8554'
      - 8555:8555/tcp
      - 8555:8555/udp
      - '1984:1984'
    restart: unless-stopped
    shm_size: 1024mb
    user: '0:0'
    volumes:
      - /mnt/Main/custom_apps/frigate/config:/config:rw
      - /mnt/Main/custom_apps/frigate/media:/media:rw
      - target: /tmp/cache
        tmpfs:
          size: 1000000000
        type: tmpfs

Object Detector

Other

Screenshots of the Frigate UI's System metrics pages

GTX 750

Any other information that may be helpful

Running on TrueNas Scale, latest version. Stopped working only upon upgrade to Frigate 0.16

[NickM-27]

The yolov7.onnx models are not optimized so they run slowly anyway. I would recommend trying to run YOLOx and see if that works

[brandondb1]

I'll give that a try, thanks. I didn't notice the pregenerated models initially...that would have saved a lot of time (I don't have any one machine with all of the required libraries, python, git, etc. installed on it so it's been a journey...)

[brandondb1]

I tried the yolox-m model and now I get:

onnxruntime.capi.onnxruntime_pybind11_state.Fail: [ONNXRuntimeError] : 1 : FAIL : Non-zero status code returned while running Slice node. Name:'Slice_34' Status Message: CUDA error cudaErrorNoKernelImageForDevice:no kernel image is available for execution on the device

I'm not sure what this error means, perhaps a driver incompatability? Truenas is running the slightly older 550.142 NVIDIA driver (new one is 570.xx I think).

[blairun]

I ran into the same issue. 0.15.2 works well, but 0.16.0 doesn't start up. My server has the latest Nvidia drivers and I'm passing through a 750ti to the container. I followed the new onnx documentation for the config and tried a few different models (yolox_.onnx, yolox_m.onnx, yolov9-s.onnx) but always get a similar error log. If these older GPUs are no longer supported, it would be helpful to have that documented somewhere.

[E:onnxruntime:, sequential_executor.cc:516 ExecuteKernel] Non-zero status code returned while running Slice node. Name:'Slice_34' Status Message: CUDA error cudaErrorNoKernelImageForDevice:no kernel image is available for execution on the device