Matteo Bettini

> Thanks a lot @matteobettini ! This is really helpful to get started with the changes of vectorising some properties :) I'll likely continue further development in my fork but...

One tricky thing that might be worth being careful about is uninitended broadcasting: For example if a property that was a float is now a tensor, its shape might have...

> Yeah, great point on the broadcasting. I just ran into this. I agree with your reshaping suggestion, I think making sure everything is 2D in the form of `batch_dim,...

Hello. Thanks for this question as this is a point I feel it is good to clarify and improve upon. ## The current situation Currently, as you say, there are...

Hello, mps is not yet fully supported as a torch device so we haven't tested vmas on it. My suggestion is to use the cpu on mac untill they fix...

The whole simulator is based on pytorch so I do not think it is feasible. I’ll look into testing vmas with mps and fixing what I can fix and will...

There are a lot of operators which are not yet supperted, like the 'norm' one. These are core to vmas and are everywhere. ` UserWarning: The operator 'aten::linalg_vector_norm' is not...

The simulator now seems to run fine with device="mps", the problem is just that many operations are extremely slow due to not being supported and falling back to cpu. Like...

This is amazing! The public wandb project with possibility to filter through releases would be super cool. It would be cool if these scripts autogenerated an output file and automatically...

> So pretty much all the same except that there's no Agent dim in the hidden state when it's shared and centralized since we flatten / expand the input /...