search

hzwer / Practical-RIFE

We are developing more practical frame interpolation approach.
MIT License
542 stars 60 forks source link

Torch_TensorRT inference? #73

Closed AIManifest closed 2 months ago

AIManifest commented 2 months ago

Hello,

I've recently stumbled across the torch_tensorrt.compile method for PyTorch models, but after a few hours of testing, I noticed that I could not run inference with these methods because the IFNet inference, namely when the block method is called, creates new shapes for the inputs, so passing the inputs parameter into torch_tensorrt.compile always returns an error. I was wondering if you had any idea how to implement this? I tried vsrife and vsgan-tensorrt-docker methods, but I think they have different structures from this repository. Any guidance would be appreciated. Thank you~

AIManifest commented 2 months ago

Update, about a week ago, through my research, I noticed that some models may not be compatible or May not work right out of the box with this request. Closing this.