File .../llama_index/llms/vllm/base.py:292, in Vllm.complete(self, prompt, formatted, **kwargs)                                          
  --> 292     sampling_params = SamplingParams(**params)
  TypeError: Unexpected keyword argument 'best_of'

---

from llama_index.llms.vllm import Vllm                                                                                                 
  from vllm import SamplingParams                                                                                                          
  from llama_index.core.base.llms.types import CompletionResponse
                                                                                                                                           
  def patched_complete(self, prompt, formatted=False, **kwargs):                                                                           
      params = {
          "temperature": kwargs.get("temperature", 0.1),                                                                                   
          "top_p": kwargs.get("top_p", 0.9),                                                                                             
          "max_tokens": kwargs.get("max_tokens", 1024),
      }                                                                                                                                    
      sampling_params = SamplingParams(**params)
      output = self._client.generate(prompt, sampling_params)                                                                              
      return CompletionResponse(text=output[0].outputs[0].text.strip())                                                                    
                                              
  Vllm.complete = patched_complete

---

if self.best_of is not None:
      base_kwargs["best_of"] = self.best_of

---

pip install llama-index==0.14.20 llama-index-llms-vllm==0.7.0 vllm==0.19.0

---

from llama_index.llms.vllm import Vllm
                                                                                                                                           
     llm = Vllm("/models/your-model", vllm_kwargs={"gpu_memory_utilization": 0.6})                                                         
     response = llm.complete("Hello")

---

File .../llama_index/core/llms/callbacks.py:447, in llm_completion_callback.<locals>.wrap.<locals>.wrapped_llm_predict                   
      447     f_return_val = f(_self, *args, **kwargs)                                                                                     
                                                                                                                                           
  File .../llama_index/llms/vllm/base.py:287, in Vllm.complete(self, prompt, formatted, **kwargs)                                          
      287     params = {**self._model_kwargs, **kwargs}
      289     from vllm import SamplingParams                                                                                              
      291     # build sampling parameters                                                                                                  
  --> 292     sampling_params = SamplingParams(**params)                                                                                   
      293     outputs = self._client.generate([prompt], sampling_params)                                                                   
      294     return CompletionResponse(text=outputs[0].outputs[0].text)
                                                                                                                                           
  TypeError: Unexpected keyword argument 'best_of'