Christoph Reich

Maybe this repo can help you :) [D3Dnet](https://github.com/XinyiYing/D3Dnet)

I tried to reimplement version 2. In case you are interested, the (PyTorch) code is available [here](https://github.com/ChristophReich1996/Swin-Transformer-V2) :)

Hi @yuangui0316, thanks for your interest. In the current version, the size of the hidden dimensions of the meta-network for computing the positional encodings is 256. So loading the checkpoints...

Hi @834799106, Thanks for your interest in my reimplementation, and thanks for pointing out these issues. Feel free to leave a star there if you liked the code :) 1....

`deform_conv_cuda.cu` checks if the input tensor is contiguous stored in memory. Did you tried to call [`tensor.contiguous()`](https://pytorch.org/docs/stable/tensors.html?highlight=contiguous#torch.Tensor.contiguous) before calling the DCN module?

I implemented a [pure PyTorch 2D involution](https://github.com/ChristophReich1996/Involution) and faced a similar issue of Nans occurring during training when using the involution as a plug-in replacement for convolutions. In my case...

@memmelma and I implemented a pure PyTorch version of a [3D Involution](https://github.com/ChristophReich1996/Involution) which you could try out. However, the native PyTorch implementation is kinda GPU memory hungry due to the...

Thanks for your great work! I would also be very interested in the camera motion LoRA checkpoints. Looking forward :)

Hi everyone, sorry for my very late response! **The bug is now fixed**. Please note [timm](https://github.com/rwightman/pytorch-image-models) as always got a probably even better implementation and pre-trained weights! Highly recommend checking...

Hi @v3551G, thanks for your interest in our code! Yes, converting to ONNX might be problematic, I never tried the conversion. While we utilize a custom CUDA implementation of the...