feat(vllm): add vLLM module

[Josephrp]

solves #885

includes complete vLLM api , can be used for generation, embeddings , finetuning , dataset creation , testing experiment configurations , rl , and testing performance , ... etc.

[alexanderankin]

take a look at the example of a database container or the ollama container. they are intentionally minimal, giving you only what you need to get started. the purpose of this library is to provide temporary test infrastructure, not to implement every feature. if you can get this to the same amount of lines as ollama or less i may consider merging. looks like a promising project. if it implements the same api as ollama, maybe even lets consider making one container work for both images. not sure if this is possible/feasible.

[Josephrp]

i see why you're saying this , and i understand where you're coming from , however ... it will be easier for me to show you the necessity & use of a fully featured object with a demo , so i'll do that in the next two weeks as i test , use and build with testcontainers . Meanwhile i'll dig into your comments and see if i can get it to where you're saying it should be :-)

[kiview]

@Josephrp We are currently working on adding vLLM support into Docker Model Runner. So I think a much better approach for the project is adding integration with Docker Model Runner (similar to what Java and Go already did), which should also help with giving GPU access on e.g. macOS hosts (since we run inference natively).

https://java.testcontainers.org/modules/docker_model_runner/

[ericcurtin]

Thanks for highlighting this @kiview . Yes we are actively working on this, it is not ready, but please feel free to collaborate with us on this:

https://github.com/docker/inference-engine-vllm

vllm-project/vllm#26160