Yarden As

Hi! I would be happy to try and help, any ideas where to start? Could [this](https://blog.evjang.com/2019/11/jaxpt.html) or [this](https://github.com/kach/softraxterizer) be of any use? I also like the idea of having a...

Thanks @erikfrey for the hints! So just trying to get a better understanding: basically we would need to implement functions like ```create_capsule``` ([e.g. here](https://github.com/google/brax/blob/8a3ea6065866f687cf657d4fc5c91d002a75f72f/brax/io/image.py#L69) and its implementation [here](https://github.com/erwincoumans/tinyrenderer/blob/0e8b3fe69f34f0b01683760c3eee29b077410631/tinyrenderer.cpp#L914)) which takes...

@erikfrey what's the timeline on the GPU rendering project? I might be able to contribute some parts.

@erikfrey hi! Are there any other things than joint stiffness that you would recommend checking/tuning after running mujoco_converter script on mujoco's .xml files? And also thanks for developing brax, it's...

@Rocamonde, do you still have this issue?

Hey @stergiosba, nice that you're working on S4s with equinox. I have my own implementation [here](https://github.com/yardenas/offline-meta-mbrl), which is also based on the original FLAX implementation. Happy to contribute here as...

@stergiosba, one thing that was a bit annoying was to pass the ssm around whenever I wanted to use the recurrent mode of S4. IIRC, in the original FLAX implementation...

@patrick-kidger, @stergiosba, I think I didn't explain myself good enough and there's a slight misunderstanding😅😅 Let me dive a bit deeper into the details, hopefully giving you a better idea...

@stergiosba, exactly, so whenever I use the RNN mode (typically for forecasting), I compute the ssm and then pass it to this function in each step in autoregression. This becomes...

@paganpasta yeah I got similar performance for both models (pytorch vs eqx one). I just tried porting my model to pytorch (collaborators still use torch) and a sanity test of...