[Detector Support]: Crash on startup with Quadro P2000 and ONNX TensorRT detector #20095
-
Describe the problem you are havingHello, I am having trouble getting Frigate to start with my Quadro P2000 GPU. The container crashes immediately after trying to initialize the detector. The error is always Could not determine exact line number: 'onnx', Message: Input should be a valid string. I have already done extensive troubleshooting, including: Correcting all config.yaml syntax for version 0.16. Confirming my docker-compose.yml is correct. Running a minimal config, which also failed. Frigate does run successfully when the detectors section is completely removed (using CPU detectors). The crash occurs with both the stable-tensorrt and stable Docker images. Here is my setup:
Version0.16.1-e664cb2 Frigate config filemqtt:
host: ip
user: user
password: password
topic_prefix: frigate
client_id: frigate
stats_interval: 60
detectors:
nvidia:
type: onnx
device: 0
provider:
- tensorrt
- cuda
ffmpeg:
hwaccel_args: preset-nvidia-h264
go2rtc:
streams:
1: rtsp://user:password@ip
2: rtsp://user:password@ip
3: rtsp://user:password@ip
cameras:
1
ffmpeg:
inputs:
- path: rtsp://127.0.0.1:8554/1
input_args: -rtsp_transport tcp
roles:
- record
- detect
detect:
width: 640
height: 640
fps: 5
2:
ffmpeg:
inputs:
- path: rtsp://127.0.0.1:8554/2
input_args: -rtsp_transport tcp
roles:
- record
- detect
detect:
width: 640
height: 640
fps: 5
3
ffmpeg:
inputs:
- path: rtsp://127.0.0.1:8554/3
input_args: -rtsp_transport tcp
roles:
- record
- detect
detect:
width: 640
height: 640
fps: 5
detect:
enabled: true
version: 0.16-0docker-compose file or Docker CLI commandservices:
frigate:
container_name: frigate
image: ghcr.io/blakeblackshear/frigate:stable
restart: unless-stopped
runtime: nvidia
environment:
- NVIDIA_VISIBLE_DEVICES=all
- NVIDIA_DRIVER_CAPABILITIES=compute,utility,video
- FRIGATE_RTSP_PASSWORD=pasword
shm_size: "1g"
volumes:
- /etc/localtime:/etc/localtime:ro
- /opt/frigate/config:/config
- /opt/frigate/storage:/media/frigate
- type: tmpfs
target: /tmp/cache
tmpfs:
size: 2000000000
ports:
- "5000:5000"
- "8554:8554"
- "8555:8555"
- "1984:1984"Relevant Frigate log outputs6-rc: info: service go2rtc-healthcheck successfully started
2025-09-16 10:44:34.727697457 [INFO] Starting NGINX...
2025-09-16 10:44:34.791984275 [INFO] Preparing Frigate...
2025-09-16 10:44:35.069470209 [INFO] Preparing new go2rtc config...
2025-09-16 10:44:35.084042707 [INFO] Starting Frigate...
s6-rc: info: service nginx successfully started
s6-rc: info: service certsync: starting
s6-rc: info: service certsync successfully started
s6-rc: info: service legacy-services: starting
2025-09-16 10:44:35.094827186 [INFO] Starting certsync...
s6-rc: info: service legacy-services successfully started
2025-09-16 10:44:35.390521068 [INFO] Starting go2rtc...
2025-09-16 10:44:35.444900381 10:44:35.444 INF go2rtc platform=linux/amd64 revision=fa580c5 version=1.9.9
2025-09-16 10:44:35.444919370 10:44:35.444 INF config path=/dev/shm/go2rtc.yaml
2025-09-16 10:44:35.445264845 10:44:35.445 INF [rtsp] listen addr=:8554
2025-09-16 10:44:35.445351652 10:44:35.445 INF [api] listen addr=:1984
2025-09-16 10:44:35.445613246 10:44:35.445 INF [webrtc] listen addr=:8555
2025-09-16 10:44:35.450519582 127.0.0.1 - - [16/Sep/2025:10:44:35 +0000] "" 400 0 "-" "-" "-"
2025-09-16 10:44:36.958235656 [2025-09-16 10:44:36] frigate.util.config INFO : Checking if frigate config needs migration...
2025-09-16 10:44:36.968882411 [2025-09-16 10:44:36] frigate.util.config INFO : frigate config does not need migration...
2025-09-16 10:44:36.982370836 *************************************************************
2025-09-16 10:44:36.982372862 *************************************************************
2025-09-16 10:44:36.982374754 *** Your config file is not valid! ***
2025-09-16 10:44:36.982377152 *** Please check the docs at ***
2025-09-16 10:44:36.982378207 *** https://docs.frigate.video/configuration/ ***
2025-09-16 10:44:36.982379512 *************************************************************
2025-09-16 10:44:36.982380513 *************************************************************
2025-09-16 10:44:36.982391013 *** Config Validation Errors ***
2025-09-16 10:44:36.982392135 *************************************************************
2025-09-16 10:44:36.982392798
2025-09-16 10:44:36.988629532 Could not determine exact line number: 'onnx'
2025-09-16 10:44:36.988631202 Message : Input should be a valid string
2025-09-16 10:44:36.988632036
2025-09-16 10:44:36.988633116 *************************************************************
2025-09-16 10:44:36.988634154 *** End Config Validation Errors ***
2025-09-16 10:44:36.988635169 *************************************************************
2025-09-16 10:44:37.285175860 [INFO] Service Frigate exited with code 1 (by signal 0)
s6-rc: info: service legacy-services: stopping
s6-rc: info: service legacy-services successfully stopped
s6-rc: info: service go2rtc-healthcheck: stopping
s6-rc: info: service certsync: stopping
2025-09-16 10:44:37.292513550 [INFO] The go2rtc-healthcheck service exited with code 256 (by signal 15)
2025-09-16 10:44:37.293299786 [INFO] Service CERTSYNC exited with code 256 (by signal 15)Install methodProxmox via Docker Object DetectorTensorRT Screenshots of the Frigate UI's System metrics pages
Any other information that may be helpfulHad to comment out the "detector" part from config to get it to boot for the screenshot. TTIA ! |
Beta Was this translation helpful? Give feedback.
Replies: 5 comments 1 reply
-
|
Your config is incorrect:
|
Beta Was this translation helpful? Give feedback.
-
|
@inkeep can provide you with the necessary suggestions and documentation links. |
Beta Was this translation helpful? Give feedback.
-
|
Your config is missing the required model section for the ONNX detector, and the provider field is not a valid option in Frigate’s detector configuration(1)(2)(3). The ONNX detector section should look like this, with a model block specifying the model path and parameters: detectors:
nvidia:
type: onnx
device: 0
model:
model_type: yolo-generic
width: 640
height: 640
input_tensor: nchw
input_dtype: float
path: /config/model_cache/yolo.onnx
labelmap_path: /labelmap/coco-80.txtYou can find more details and examples in the official documentation: Make sure your model path matches the actual file and that the width/height are correct for your model. Remove the provider key, as it is not used in Frigate’s detector config(2)(3). 📚 Sources: Have another question? Just tag @inkeep. |
Beta Was this translation helpful? Give feedback.
-
|
Ive adjusted as proposed: detectors: This is the error Im getting now: 2025-09-16 13:30:13.073295306 TypeError: Unable to load from type '<class 'NoneType'>' Running Yolov9-s model, tried 8 aswell but with same error. File permissions etc all checked and OK (1000). 2025-09-16 13:30:13.009375622 [2025-09-16 13:30:12] frigate.util.config INFO : Checking if frigate config needs migration... |
Beta Was this translation helpful? Give feedback.
-
|
You are using the type |
Beta Was this translation helpful? Give feedback.
-
|
Your error indicates that the ONNX detector is trying to load a model from a path that resolves to None, which usually means Frigate can't find the model file at the given location or the config structure is not correct(1). Double-check these points:
detectors:
onnx:
type: onnx
model:
model_type: yolox
width: 640
height: 640
input_tensor: nchw
input_pixel_format: bgr
path: /config/model_cache/onnx/yolov9-s.onnx
labelmap_path: /config/coco_labels.txt
model_type: yolo-generic
If you correct the config structure and ensure the file paths are valid inside the container, this should resolve the NoneType error. Reference: Supported Hardware(1). 📚 Sources: Have another question? Just tag @inkeep. |
Beta Was this translation helpful? Give feedback.
Your config is missing the required model section for the ONNX detector, and the provider field is not a valid option in Frigate’s detector configuration(1)(2)(3). The ONNX detector section should look like this, with a model block specifying the model path and parameters:
(2), (3)
You can find more details and examples in the official documentation:
Make sure your model path matches the actual file and that the width/h…