XXX failure of TensorRT X.Y when running XXX on GPU XXX

[loryruta]

Description

I'm having a similar problem in C++ using TensorRT and CUDA. My onnx model has 2 inputs and one output. My inference code is:

m_context = std::unique_ptr<nvinfer1::IExecutionContext>(m_engine->createExecutionContext());

// Set input dimensions (as input shape is dynamic)
m_context->setInputShape("im0", dims);
m_context->setInputShape("im1", dims);

status = m_context->setTensorAddress("im0", im0_device_ptr), assert(status);
status = m_context->setTensorAddress("im1", im1_device_ptr), assert(status);
status = m_context->setTensorAddress("disparity_map", out_disparity_map_device_ptr), assert(status);

status = m_context->enqueueV3(stream), assert(status);

The error I'm getting is:

[ERROR] [] TensorRT error: IExecutionContext::enqueueV3: Error Code 1: Cask (Cask Pooling Runner Execute Failure)

I've also tried changing enqueueV3 to executeV2 - still same issue.

I'm 99.99% sure im0_devce_ptr, im1_device_ptr and out_disparity_map_device_ptr are valid pointers to device memory.

The thing is, I have the exact similar code in Python + PyTorch and it works using the same engine.

Environment

TensorRT Version: 10.8

NVIDIA GPU: RTX 2060 SUPER

NVIDIA Driver Version: 560.35.05

CUDA Version: 12.6

CUDNN Version: -

Operating System: Ubuntu 24.04

Python Version (if applicable): 3.11

Tensorflow Version (if applicable):

PyTorch Version (if applicable): 2.5.1

Baremetal or Container (if so, version): Baremetal

Relevant Files

Model link:

https://drive.google.com/file/d/108rgI-m2-3Xg17vRDdvXf3yS-dX8fFHV/view?usp=drive_link

Steps To Reproduce

Commands or scripts:

Have you tried the latest release?:

Can this model run on other frameworks? For example run ONNX model with ONNXRuntime (polygraphy run <model.onnx> --onnxrt):