samsonyilma
samsonyilma
I too observe similar behavior; different results with CUDAExecutionProvider vs TensorrtExecutionProvider when using onnxruntime-gpu==1.19.2 with tensorrt=10.4.0 I get same outputs (CUDA vs TensorRT providers) when using onnxruntime-gpu==1.17.1 with tensorrt==8.6.0 I...
> That seems like a clear-cut case. Our case involves proprietary models and C++ code that I can't share. We don't use the python bindings so unfortunately it would be...
I am using Ubuntu 22.04, python API with onnxruntime-gpu & tensorrt packages.