Async vllm (#693)
* fix pyproject
* async
* wip logprobs
* moved adapter/delta cleanup to avoid messing up the pipeline
* cleaner shutdown
* remove useless waits
* now we manage async models at the pipeline level to have a single event loop
* Update src/lighteval/models/vllm/vllm_model.py
Co-authored-by: lewtun <lewis.c.tunstall@gmail.com>
* removing dead code from PR
* updated docstring
* lol you need to iterate on the generator to actually get it to generate
* remove dead code
* Update src/lighteval/models/vllm/vllm_model.py
Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
* add warning about sampling for now
* mini fix
---------
Co-authored-by: lewtun <lewis.c.tunstall@gmail.com>
Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
