Vincent Moens

If it can help, I wrote a module that makes the Forward macro usable for functions with an arbitrary number of scalar arguments (10 in the example here). Is probably...

I don't know whether it tackles the problem as you'd have wished, but there are a few ad hoc solutions to that: First, compute the vector * matrix product using...

Perhaps you have more than one key? Have you tried setting `export MJKEY_PATH=path/to/.mujoco/some/custom/key.txt`?

I still get the same behaviour with `DISABLE_RENDER_THREAD_OFFLOADING=1` :/

Hi @saran-t I've been trying hard to reproduce this but it seems to only happen after the code reaches a certain level of complexity (e.g. gpus are used for training...

Here's one [0e88eac27f1d01bfa1d260d52c051ab5fe514859](https://github.com/nicolas-dufour/torchrl/commit/0e88eac27f1d01bfa1d260d52c051ab5fe514859) Here's the command line ``` conda create -n mbrl_dmcontrol3 python=3.10 conda activate mbrl_dmcontrol3 pip install dm_control module load cuda/11.6 nccl/2.12.7-cuda.11.6 nccl_efa/1.15.1-nccl.2.12.7-cuda.11.6 pip3 install torch torchvision torchaudio --extra-index-url...

> If it's consistently reproducible, a messy repro case will be better than not having one at all, so please do give us that anyway. got it! > Also, are...

Let me write a single-gpu example for you

It is running in a much more stable way than it used to. No noisy pixel, and runs that used to collapse after a couple of iterations are now running...

> @vmoens Hey Vincent, I've adapted the maml_torchrl script using lastest torchrl and corrected my own MAML-RL implementation, but the results are still not aligned. Mind taking a look why...