Myelin error when load ONNX model, myelin::symbolic_value_t::operator size_t() const: Assertion `is_const()' failed

We’re preparing to use Orin with TensorRT8.3 for better performance, but it failed when creating a Builder. Error occurred in myelin which is close-sourced, we’ve gotten stuck here for days and had no clue where to start debugging, so ask for your help.

Please note that I am using orin, which only goes wrong at fp16, and int8 can pass normally.

/root/gpgpu/MachineLearning/myelin/src/compiler/…/./include/utils/value.h:394: myelin::symbolic_value_t::operator size_t() const: Assertion `is_const()’ failed.

Environment

**TensorRT Version :tensorrt 8.3
**GPU Type :Orin
**Nvidia Driver Version :515
**CUDA Version :11.4
**CUDNN Version :8.2

Relevant Files

onnx:

log:

Steps To Reproduce

trtexec --onnx=detr3d-transformer_opt1_5_1_8_rm.onnx --workspace=4096 --fp16 --verbose

Hi,
Request you to share the ONNX model and the script if not shared already so that we can assist you better.
Alongside you can try few things:

  1. validating your model with the below snippet

check_model.py

import sys
import onnx
filename = yourONNXmodel
model = onnx.load(filename)
onnx.checker.check_model(model).
2) Try running your model with trtexec command.

In case you are still facing issue, request you to share the trtexec “”–verbose"" log for further debugging
Thanks!

Models and logs are already available

Hi,

We couldn’t reproduce the issue on TensorRT version 8.4 GA version.
Please use the latest version.
https://developer.nvidia.com/nvidia-tensorrt-8x-download

Thank you.

Is it true that you are using orin drive os 6.03 and tensort8.4?

I use x86 v100 to be able to pass directly, and I have to be on the drive os to reproduce the problem

The version number of x86 v100 is exactly the same as that of drive OS 6.0, which is trt8.3

I used Ubuntu OS. In the recent version, we have fixed known issues.
We always recommend you to please use the latest TensorRT version.

Thank you.