Shichen Liu

This really helps! Thanks!!!!

@lizhenstat Thanks for addressing @LeighDavis 's questions. I currently have no ImageNet installed on my machine, so I am really glad you can help!

Hi, The code should be working on 2080Ti with pytorch 1.0. Could you provide the training command that you used?

Hi, I have updated and tested the code to support PyTorch 1.1.0. Please feel free to check it out!

Sorry for being late. Could you show me the command you use to train? The pytorch API changed a lot since 0.2, so I am planning to re-check the code...

Sorry for being late. As far as I know, you don't need the state dict of optimizer to do transfer learning. Just load the converted weight and train should be...

Our model is mainly designed for mobile devices, on which the actual inference time highly correlates with the theoretical complexity. However, the group convolution and index/shuffle operations are not efficiently...

Can you try removing "/home/wangls@1/anaconda3/envs/py37/lib/python3.7/site-packages/soft_renderer-1.0.0-py3.7-linux-x86_64.egg" folder and rebuilding it? It seems that the problem is due to version mismatch between already built modules and current cuda version.

Can you make sure your system cuda version is also 10.1? Because compiling cuda code does not necessarily use the cudatoolkit version.

Hi there, I would recommend you completely remove the previously built modules, for instance, ~/.local/python/xxx/softrasxxx.egg. and then rebuild the repo