vLLM inference example

[3manifold]

Can you provide a vLLM inference example without the need to run a server? For example, something similar to:

model = LLM(
    model=model_args.model_name,
    runner="pooling",
    dtype="bfloat16",
    max_model_len=8192,
# ....

)
# ....
emb = model.embed(model_input)
return F.normalize(emb[0].outputs.embedding, p=2, dim=-1) 

Is that correct?