Michael Feil

Have you tried to change the `_setup_viewing_camera` in the `env.render` method? Particularly `self._render_width = 320` and `self._render_height = 240 ` https://github.com/rr-learning/CausalWorld/blob/548e66c36fba01125cf6290992dfd833ae42709b/causal_world/envs/causalworld.py#L550-L581

Maybe somebody can verify - does it also happen if you change the user? ``` #./Dockerfile User someotheruser # change from root to other user ```

@sam-h-bean Not to easy, i tried that in a similar project (https://github.com/michaelfeil/infinity) - thats written in pure python, and a bit slower (2.5x less throughput). Some starting points: Here the...

Update, the following function does not seem to have such a behavior. ```python def tokenize(example, rank: int = 0): # global tokenizer_tinyllama gc.collect() # chat = [ # {"role": "user",...

@jamescalam Does __call__ support async calls?

I would disagree. Ray has extended serving support, and extends towards LLMs and streaming. Also there is tools like https://ray-project.github.io/kuberay/components/operator/ Perhaps isolate ray or make it optional.

@lanking520 Missed that point and got you a bit wrong. Thanks for additional explanations.

@flexchar Are you still having this issue? I suspect ```pip install --upgrade torch``` will overwrite the pinned version - this should be the cause.

@fdegier Does it make sense to add instructions in the markdown on how to configure the GitLab VS Code extension?

FYI, I just created a small project called [**infinity**](https://github.com/michaelfeil/infinity). Its a lightweight async implementation using fastapi and pydantic for input validation, `torch` and `ctranslate2` under the hood, performs dynamic batching...