Conversion from .etlt to TensorRT model fail

Robotics & Edge Computing Robotics - Isaac Isaac ROS
1

Hi, I’m following isaac_ros_rtdetr tutorial but I’m stuck at model conversion from .etlt to TensorRT model (see the output below).

Any suggestion on how to solve?

:/workspaces/isaac_ros-dev$ /opt/nvidia/tao/tao-converter -k sdetr -t fp16 -e ${ISAAC_ROS_WS}/isaac_ros_assets/models/synthetica_detr/sdetr_grasp.plan -p images,1x3x640x640,2x3x640x640,4x3x640x640 -p orig_target_sizes,1x2,2x2,4x2 ${ISAAC_ROS_WS}/isaac_ros_assets/models/synthetica_detr/sdetr_grasp.etlt
[INFO] [MemUsageChange] Init CUDA: CPU +13, GPU +0, now: CPU 19, GPU 139 (MiB)
[INFO] [MemUsageChange] Init builder kernel library: CPU +1445, GPU +268, now: CPU 1541, GPU 407 (MiB)
[libprotobuf ERROR google/protobuf/text_format.cc:298] Error parsing text-format onnx2trt_onnx.ModelProto: 1:1: Invalid control characters encountered in text.
[libprotobuf ERROR google/protobuf/text_format.cc:298] Error parsing text-format onnx2trt_onnx.ModelProto: 1:12: Invalid control characters encountered in text.
[libprotobuf ERROR google/protobuf/text_format.cc:298] Error parsing text-format onnx2trt_onnx.ModelProto: 1:17: Already saw decimal point or exponent; can't have another one.
[libprotobuf ERROR google/protobuf/text_format.cc:298] Error parsing text-format onnx2trt_onnx.ModelProto: 1:14: Message type "onnx2trt_onnx.ModelProto" has no field named "pytorch".
[ERROR] ModelImporter.cpp:733: Failed to parse ONNX model from file: /tmp/fileIBqR9A
[ERROR] Failed to parse the model, please check the encoding key to make sure it's correct
[ERROR] Number of optimization profiles does not match model input node number.
Aborted (core dumped)

Thank you!

2

Hi @marco.pastorio

To better support you, can you reply to some of my questions below?

I have currently tested on Jetson AGX Orin and x86, and I didn’t encounter any errors while running this step.

Best,
Raffaello

3

Sorry, I wrote it in the other topic but forgot to specify it here. My setup is:
Laptop with i9 processor, 32GB RAM, RTX4070 with 8GB RAM, Driver Version: 535.183.01, CUDA Version: 12.2

It seems to be within specs.

I can run correctly the Tutorial for cuMotion MoveIt Plugin with Isaac Sim, which seems quite resource-intensive, and other tutorials based on Isaac ROS Dev docker container, but I seem to struggle with those that involve AI models (like isaac_ros_foundationpose quickstart as I stated in the other topic).

UPDATE: isaac_ros_centerpose quickstart and isaac_ros_dope now run successfully, yesterday I had trouble with centerpose too. This one still give the conversion error though.