Kevin Zakka

There's a few things you can try: * Increase `impratio` by one or two orders of magnitude * Switch to elliptic friction cones e.g. `` There are other things you...

Can you generate a tighter approximation with V-HACD?

@tudorjnu, I have indeed noticed that v4 produces almost flat meshes at times which makes qhull complain. Have you tried giving this [video](https://www.youtube.com/watch?v=Sa1NE-St8iA) a watch?

@bemoregt your image size is way bigger than MNIST (256x256) so there are few hyperparameters you'd have to tweak to improve performance. For example, you can try increasing the patch...

@duygusar @linzhiqiu the way I wrote the repository makes it so that you just have to modify the `data_loader.py` file to make it work for your own needs. The implementation...

Won't `l_t.detach()` stop the gradients for `l_t` and hence for `mu`? I mean, we'll still have gradient for `mu` from `log_pi` but there's a contribution from `l_t` as well. This...

@ipod825 I need to think about it some more. Empirically, I haven't seen a performance difference between the 2. I still reach ~1.3-1.4% error in about 30 epochs of training....

@ipod825 The PDF of a normal distribution is not bounded, so it is not guaranteed that `l_t` will never exceed [-1,1]. I was against using `torch.clamp` because it is not...

Any reason you're trying to install 2.1.1? I'd just go for the latest dm_control (1.0.2) and mujoco (2.1.5) which you can install with `pip install --upgrade dm_control`.

`intvelocity` is only available on the master branch of MuJoCo so it’s still an unreleased feature. As such, the pip version of MuJoCo as well as packages that depend on...