llm_ops.py

import os 
import torch 
from langchain import HuggingFacePipeline
from transformers import AutoTokenizer
import transformers
from langchain import HuggingFaceHub

def get_openai_chat_model(API_key):
    try:
        from langchain.llms import OpenAI
    except ImportError as err:
        raise "{}, unable to load openAI. Please install openai and add OPENAIAPI_KEY"
    os.environ["OPENAI_API_KEY"] = API_key
    llm = OpenAI()
    return llm


def get_llama_model(temperature=0,api_key=None,max_tokens=2048):
    model_id = "meta-llama/Llama-2-7b-chat-hf"
    llm = None
    try:
        tokenizer = AutoTokenizer.from_pretrained(model_id)

        pipeline = transformers.pipeline("text-generation", 
            model=model_id,
            tokenizer=tokenizer,
            torch_dtype=torch.bfloat16,
            trust_remote_code=True,
            device_map="auto",
            max_length=1000,
            eos_token_id=tokenizer.eos_token_id
            )
        llm = HuggingFacePipeline(pipeline = pipeline, model_kwargs = {'temperature':temperature})
    except:
            raise "User not autorized to access the Model"
    return llm

def get_model_from_hub(api_key,temperature=0.1,max_tokens=2048,model_id="meta-llama/Llama-2-7b-chat-hf"):
    llm = HuggingFaceHub(huggingfacehub_api_token=api_key,
                                        repo_id=model_id,
                                        model_kwargs={"temperature": temperature, "max_new_tokens": max_tokens})
    return llm