How can I compile pytorch3d inference to run it in production?

[Alexankharin]

I've trained a pytirch3d model that I want to place inference on server with GPUs (AWS). It is highly desirable for me to keep as few dependencies as possible. Is it possible to convert the pytorch3d model (I've used only pointcloud rasterizer function from the library) into jit-traced/scripted model or to onnx format?

[SalehChoueib]

Hi Alexander,

I was wondering if you had any luck with inferencing a Pytorch3D model with onnx?

[saidElhadi]

Hi Salah

Were you able to inference a model built with pytorch2d using onnx?