Provide example chat template usage

[tstescoTT]

using vLLM endpoint API: v1/chat/completions

Client will respond:

 RESPONSE:
"As of transformers v4.44, default chat template is no longer allowed, so you must provide a chat template if the tokenizer does not define one."

[tstescoTT]

There are 2 ways to get the instruct template picked up by the vllm server so it can serve the /v1/chat/completions endpoint doing the instruct templating server side:

1. use instruct HF repo name in call to vllm.entrypoints.openai.api_server: https://github.com/tenstorrent/tt-inference-server/blob/main/vllm-tt-metal-llama3-70b/src/run_vllm_api_server.py#L45 instead of non-instruct repo (https://github.com/tenstorrent/vllm/blob/dev/examples/server_example_tt.py#L14C21-L14C50). I was able to repro the original issue running the meta-llama/Llama-3.1-70B which does not have the instruct template in the tokenizer config.
2. pass --chat-template arg to call to vllm.entrypoints.openai.api_server directly with raw string of the jinja template.

For reference, vllm server will also support /v1/completions endpoint that does not do any templating (e.g. if you want to do your own templating)

Example client usage with meta-llama/Llama-3.1-70B-Instruct and v1/chat/completions:

curl -s --no-buffer -X POST "http://localhost:7000/v1/chat/completions" \
    -H "Content-Type: application/json" \
    -H "Authorization: Bearer $TOKEN" \
    -d '{
        "model": "meta-llama/Llama-3.1-70B-Instruct",
        "messages": [
        {
            "role": "system",
            "content": "You are a helpful assistant."
        },
        {
            "role": "user",
            "content": "Hello!"
        }
        ],
        "max_tokens": 64,
        "stream": true
    }'