Ye Yuan

I'm not sure what you mean exactly by "single inference step". Could you elaborate?

Hi, I think to get the most likely prediction using AgentFormer, you would need to train a one-sample DLow model.

The value function also needs to be an RNN, or you can pass the rnn output from the policy to the value net.

I don't have any code for RNN yet. But I can imagine how it can be done. Suppose the agent collected a batch of three episodes of length 5,6,7, so...

I'm using M1 macbook pro and had no issue with running VTK on my machine. I am using miniforge with the installation instructions here: https://naolin.medium.com/conda-on-m1-mac-with-miniforge-bbc4e3924f2b

Have you installed the latest requirements.txt and install.sh?

@iulsko what OS you are using?

Check out solutions here: https://github.com/NVlabs/GLAMR/issues/10#issuecomment-1210789272 The issue is mainly with pyvista. You can try first to run their examples in the link above

are you able to pop a vis GUI with `--vis` flag when running the demo?

I'm perplexed that the window is completely black. The fitting process only runs once since it will use cached results saved when running first time.