Izan Catalán

 Hi @jcwchen I solved the problem getting the input and output names and using them on "model.run" That way, I can get the outpus which ara in [1,1000] shape....

Hi @gramalingam, yes my use case is to use the models with a feature extraction as research to understand its behaviour. Any information would help.

@thakkarV Where can I find the preconditions for conv3d? And in addition, you said that it is a 2.x implementation of conv. Is there any other example for conv2d with...

@thakkarV those examples are for a NVIDIA H100 Tensor Core GPU (SM90), my gpu is a NVIDIA A100 Tensor Core GPU and does not work with them (sm80), that is...

@thakkarV maybe not first-class API, but some source code, as is explained in [Convolution Readme](https://github.com/NVIDIA/cutlass/blob/main/media/docs/implicit_gemm_convolution.md), and more specifically in [device convolution](https://github.com/NVIDIA/cutlass/blob/main/tools/util/include/cutlass/util/reference/device/convolution.h)?

Not yet. Still waiting for help

Hi @hwu36, I will explain to you my final goal; maybe you will have a better idea of what I am trying to do because I have seen you have...

@hwu36 Yes, that seems like the easiest thing to do. However, I have a couple of questions: 1) Adding an extra parameter to the kernel would mean changing the definition...

@hwu36 can you please take a look to this new issue which is related: https://github.com/NVIDIA/cutlass/issues/2166

@jackkosaian are they available in Python? I couldn't check that as I have written in a [new post](https://github.com/NVIDIA/cutlass/issues/1990)