Model name for vLLM instantiation

[rpavlovicTT]

In run_vllm_server.py script vLLM server is configured with hf_model_name as

def model_setup(hf_model_id):
    # TODO: check HF repo access with HF_TOKEN supplied
    print(f"using model: {hf_model_id}")
    args = {
        "model": hf_model_id,
...

but in vLLM TT worker implementation we have for example:

        if ("meta-llama/Meta-Llama-3.1-8B" in self.model_config.model...

setup.sh generates following vars

$HF_MODEL_REPO_ID=meta-llama/Llama-3.1-8B-Instruct
$META_MODEL_NAME=Meta-Llama-3.1-8B-Instruct

but neither of them explicitly matches string expected by vLLM... so we should consolidate these naming conventions.

[tstescoTT]

FYI some WIP on this in https://github.com/tenstorrent/tt-inference-server/blob/tstesco/dev/setup.sh

Plan is to consolidate to HF repo ID, but we have some legacy model impls that differ. We can use HF model ID downstream of setup going forward.

[tstescoTT]

I think we should be handling the *-Instruct model versions explicitly because they have different handling for tokenization and chat templating.

[tstescoTT]

addressed in #88