import transformers
import torch
from fastapi import FastAPI
from pydantic import BaseModel
import uvicorn


app = FastAPI()



model_id = "meta-llama/Meta-Llama-3-8B"

pipeline = transformers.pipeline(
    "text-generation", model=model_id, model_kwargs={"torch_dtype": torch.bfloat16}, device_map="auto"
)



class Item(BaseModel):
    prompt: str
    



def generate(item: Item):
    pipeline(item.prompt)
   

@app.post("/generate/")
async def generate_text(item: Item):
    return {"response": generate(item)}