Issue Using Qwen Model with Lotus (vLLM Deployment)

[KH-Khadija]

I am trying to use the Qwen2.5-14B-Instruct-GPTQ-Int8 model with Lotus and have it deployed via vLLM on a GPU. Based on the Lotus documentation, models deployed with vLLM should be usable like this:

from lotus.models import LM
lm = LM(
    model='hosted_vllm/Qwen/Qwen2.5-14B-Instruct-GPTQ-Int8',
    api_base='http://localhost:8010/v1',
)

However, when I try this, I get the following error:

Exception: This model isn't mapped yet. model=hosted_vllm/Qwen/Qwen2.5-14B-Instruct-GPTQ-Int8, custom_llm_provider=hosted_vllm. Add it here - https://github.com/BerriAI/litellm/blob/main/model_prices_and_context_window.json.

From my understanding, Lotus only supports models available in LiteLLM, so does this mean LiteLLM does not yet support Qwen models?

If that is the case and I am constrained to using Qwen, what are my options to work around this? Would I need to manually extend LiteLLM's supported models, or is there another way to integrate Qwen with Lotus?