Jetson Orin Nano Jetpack6 : When generating model Yolov7-320-dla I get a generating error

[github665x2]

Describe the problem you are having

I run a Jetson Orin Nano 8gb. I created a docker-compose.yaml where I, as in de frigate documentation. add the parameter -dla to the model Yolo7-320. You get then Model=Yolo7-320-dla. The build process give's a error and later on exit with code 2 cause the *trt files cannot be found. If a build it without added -dla. The build process succesfully completes.

Version

0.16.2-4d58206

Frigate config file

mqtt:
  enabled: False

cameras:
  camera_garage: # <--- this will be changed to your actual camera later
    enabled: True
    ffmpeg:
      hwaccel_args: preset-jetson-h264
      inputs:
        - path: rtsp://192.168.1.21:554/axis-media/media.amp?streamprofile=h264_1024
          roles:
            - detect

birdseye:

  enabled: True
  mode: objects
         
detectors:
  tensorrt:
    type: tensorrt
    device: 0 #This is the default, select the first GPU

model:
  path: /config/model_cache/tensorrt/yolov7-320-dla.trt
  input_tensor: nchw
  input_pixel_format: rgb
  width: 320
  height: 320

detect:
  fps : 20
  width: 1280
  height: 720
  
objects:
  track:
    - person

docker-compose file or Docker CLI command

services:
 frigate:
    privileged: true
    environment:
      - YOLO_MODELS=yolov7-320-dla
      - USE_FP16=true
    container_name: frigate
    runtime: nvidia 
    devices:
      - /dev/video1:/dev/video1 
      - /dev/bus/usb:/dev/bus/usb
    volumes:
      - /home/develop/frigate/config:/config
      - /home/develop/frigate/storage:/media/frigate
      #- type: tmpfs # Optional: 1GB of memory, reduces SSD/SD Card wear
        #target: /tmp/cache
        #tmpfs:
          #size: 1000000000
    ports:
      - "5000:5000"
      - "8554:8554"
    image: ghcr.io/blakeblackshear/frigate:stable-tensorrt-jp6

Relevant Frigate log output

Generating the following TRT Models: yolov7-320-dla
s6-rc: info: service nginx-log successfully started
Downloading yolo weights
s6-rc: info: service certsync-log successfully started
s6-rc: info: service go2rtc-log successfully started
s6-rc: info: service go2rtc: starting
s6-rc: info: service frigate-log successfully started
s6-rc: info: service go2rtc successfully started
s6-rc: info: service go2rtc-healthcheck: starting
s6-rc: info: service go2rtc-healthcheck successfully started
2025-11-25 12:39:09.201585441  [INFO] Preparing new go2rtc config...
2025-11-25 12:39:09.811626229  [INFO] Starting go2rtc...
2025-11-25 12:39:09.938996789  12:39:09.938 INF go2rtc platform=linux/arm64 revision=fa580c5 version=1.9.9
2025-11-25 12:39:09.939013205  12:39:09.938 INF config path=/dev/shm/go2rtc.yaml
2025-11-25 12:39:09.939635403  12:39:09.939 INF [api] listen addr=:1984
2025-11-25 12:39:09.939668876  12:39:09.939 INF [rtsp] listen addr=:8554
2025-11-25 12:39:09.940340035  12:39:09.940 INF [webrtc] listen addr=:8555

Creating yolov7-320.cfg and yolov7-320.weights

Done.
2025-11-25 12:39:18.514674431  [INFO] Starting go2rtc healthcheck service...

Generating yolov7-320-dla.trt. This may take a few minutes.

/usr/local/src/tensorrt_demos/yolo/onnx_to_tensorrt.py:149: DeprecationWarning: Use network created with NetworkDefinitionCreationFlag::EXPLICIT_BATCH flag instead.
  builder.max_batch_size = MAX_BATCH_SIZE
/usr/local/src/tensorrt_demos/yolo/onnx_to_tensorrt.py:151: DeprecationWarning: Use set_memory_pool_limit instead.
  config.max_workspace_size = 1 << 30
/usr/local/src/tensorrt_demos/yolo/onnx_to_tensorrt.py:174: DeprecationWarning: Use build_serialized_network instead.
  engine = builder.build_engine(network, config)
ERROR: failed to build the TensorRT engine!
[11/25/2025-12:40:52] [TRT] [W] 066_maxpool: DLA only supports windows in the range of [1-8].
[11/25/2025-12:40:52] [TRT] [W] Layer '066_maxpool' (POOLING): Unsupported on DLA. Switching this layer's device type to GPU.
[11/25/2025-12:40:52] [TRT] [W] 068_maxpool: DLA only supports windows in the range of [1-8].
[11/25/2025-12:40:52] [TRT] [W] Layer '068_maxpool' (POOLING): Unsupported on DLA. Switching this layer's device type to GPU.
[11/25/2025-12:40:52] [TRT] [W] Layer '(Unnamed Layer* 389) [PluginV2IOExt]' (PLUGIN_V2): Unsupported on DLA. Switching this layer's device type to GPU.
[11/25/2025-12:40:52] [TRT] [W] Layer '(Unnamed Layer* 390) [PluginV2IOExt]' (PLUGIN_V2): Unsupported on DLA. Switching this layer's device type to GPU.
[11/25/2025-12:40:52] [TRT] [W] Layer '(Unnamed Layer* 391) [PluginV2IOExt]' (PLUGIN_V2): Unsupported on DLA. Switching this layer's device type to GPU.
[11/25/2025-12:40:55] [TRT] [W] Dimension: 1 (33600) exceeds maximum allowed size for DLA: 8192
[11/25/2025-12:40:55] [TRT] [W] Validation failed for DLA layer: (Unnamed Layer* 392) [Concatenation]. Switching to GPU fallback.
[11/25/2025-12:41:07] [TRT] [E] 2: Assertion !n->candidateRequirements.empty() failed. No supported formats for {ForeignNode[001_convolutional + 001_convolutional_bn...063_convolutional_swish]}
[11/25/2025-12:41:07] [TRT] [E] 2: [optimizer.cpp::getFormatRequirements::3154] Error Code 2: Internal Error (Assertion !n->candidateRequirements.empty() failed. No supported formats for {ForeignNode[001_convolutional + 001_convolutional_bn...063_convolutional_swish]})
Loading the ONNX file...
Adding yolo_layer plugins.
Adding a concatenated output as "detections".
Naming the input tensort as "input".
Building the TensorRT engine.  This would take a while...
(Use "--verbose" or "-v" to enable verbose logging.)
Using DLA core 0.
Available tensorrt models:
ls: cannot access '*.trt': No such file or directory
s6-rc: warning: unable to start service trt-model-prepare: command exited 2

Install method

Docker Compose

Object Detector

TensorRT

Screenshots of the Frigate UI's System metrics pages

The server is not starting up.

Any other information that may be helpful

No response

[inkeep-ai-assistant[bot]]

Based on the error logs, the DLA (Deep Learning Accelerator) model generation is failing due to TensorRT engine build errors. The logs show several DLA compatibility issues:

1. DLA Layer Incompatibilities: Multiple layers are being automatically switched from DLA to GPU because they're unsupported on DLA(1)

   • Pooling layers with windows outside the 1-8 range
   • PluginV2 layers
   • Concatenation layers exceeding DLA's maximum dimension size (8192)

2. Critical Build Failure: The final error Assertion !n->candidateRequirements.empty() failed. No supported formats indicates TensorRT cannot find a valid execution path for the model on DLA(1)

Solution: The YOLOv7-320 model architecture has layers that are incompatible with the Jetson Orin Nano's DLA. Instead of using the -dla suffix, you should:

1. Use GPU-only execution by removing the -dla suffix from your model name(1):

environment:
  - YOLO_MODELS=yolov7-320
  - USE_FP16=true

model:
  path: /config/model_cache/tensorrt/yolov7-320.trt

2. Alternative approach: Consider using ONNX instead, which has improved performance in recent versions(1). You can configure an ONNX detector by using the -tensorrt-jp6 image (which you're already using) and setting up an ONNX model configuration(1).

The DLA is designed for specific layer types and has dimensional constraints that the YOLOv7-320 architecture exceeds. Running on GPU will still provide hardware acceleration without the DLA-specific limitations(1).

**

📚 Sources:

1. Supported Hardware

---

Have another question? Just tag @inkeep.