Add single command LLM deployment

[mreso]

Description

This PR adds a feature to TorchServe to deploy LLM with a single command.
Is uses a new ts.launcher interface to start and stop torchserve and adds ts.llm_launcher to launch an LLM model given its hugginface hub identifier using our new vllm integration.
It also adds a docker image based on our gpu image to easily run this without installing torchserve.

Type of change

Please delete options that are not relevant.

• New feature (non-breaking change which adds functionality)
• This change requires a documentation update

Feature/Issue validation/testing

Please describe the Unit or Integration tests that you ran to verify your changes and relevant result summary. Provide instructions so it can be reproduced.
Please also list any relevant details for your test configuration.

• Test A

docker run --rm -ti --gpus all -e HUGGING_FACE_HUB_TOKEN=$token -p 8080:8080 -v data:/data ts/llm --model_id meta-llama/Meta-Llama-3-8B-Instruct --disable_token
curl -X POST -d '{"prompt":"Hello, my name is", "max_new_tokens": 50}' --header "Content-Type: application/json" "http://localhost:8080/predictions/model"

Logs for Test A

{"text": " [", "tokens": 510}{"text": "Your", "tokens": 7927}{"text": " Name", "tokens": 4076}{"text": "].", "tokens": 948}{"text": " I", "tokens": 358}{"text": " am", "tokens": 1097}{"text": " a", "tokens": 264}{"text": " [", "tokens": 510}{"text": "Your", "tokens": 7927}{"text": " Profession", "tokens": 50311}{"text": "/", "tokens": 14}{"text": "Student", "tokens": 14428}{"text": "]", "tokens": 60}{"text": " and", "tokens": 323}{"text": " I", "tokens": 358}{"text": " am", "tokens": 1097}

Checklist:

• Did you have fun?
• Have you added tests that prove your fix is effective or that this feature works?
• Has code been commented, particularly in hard-to-understand areas?
• Have you made corresponding changes to the documentation?

[agunapal]

Looks good!
Had one comment on the token. Please check and change if you think it makes sense

[agunapal]

Is this needed -e HUGGING_FACE_HUB_TOKEN=$token Why not directly set export HUGGING_FACE_HUB_TOKEN= <HUGGINGFACE_HUB_TOKEN>

Also from the security POV, does the existing command print the token?

[mreso]

AFAIK the docker will not pick up env variables from the calling environment. So you would still have

export HUGGING_FACE_HUB_TOKEN= <HUGGINGFACE_HUB_TOKEN>
docker run --rm -ti --gpus all -e HUGGING_FACE_HUB_TOKEN=$HUGGING_FACE_HUB_TOKEN ...

which is even longer and it comes down to the same process. For a RL deployment the token variable would be set through a secret and so the token would not show up in any of the logs.