main.py

from fastapi import FastAPI
from pydantic import BaseModel
import requests
from ctransformers import AutoModelForCausalLM

llms = {
"tinnyllama":{"name": "TheBloke/TinyLlama-1.1B-1T-OpenOrca-GGUF", "file":"tinyllama-1.1b-1t-openorca.Q4_K_M.gguf", "suffix":"<|im_end|><|im_start|>assistant", "prefix":"<|im_start|>system You are a helpful assistant  <|im_end|><|im_start|>user"},
"orca2":{"name": "TheBloke/Orca-2-7B-GGUF", "file":"orca-2-7b.Q4_K_M.gguf", "suffix":"<|im_end|><|im_start|>assistant", "prefix":"<|im_start|>system You are a helpful assistant<|im_end|><|im_start|>user "},
"zephyr":{"name": "TheBloke/zephyr-7B-beta-GGUF", "file":"zephyr-7b-beta.Q4_K_M.gguf", "suffix":"</s><|assistant|>", "prefix":"<|system|>You are a helpful assistant</s><|user|> "},
"mixtral":{"name": "TheBloke/Mistral-7B-Instruct-v0.1-GGUF", "file":"mistral-7b-instruct-v0.1.Q4_K_M.gguf", "suffix":"[/INST]", "prefix":"<s>[INST] "},
"llama2":{"name": "TheBloke/Llama-2-7B-Chat-GGUF", "file":"llama-2-7b-chat.Q4_K_M.gguf", "suffix":"[/INST]", "prefix":"[INST] <<SYS>> You are a helpful assistant <</SYS>>"},
"llama2":{"name": "TheBloke/SOLAR-10.7B-Instruct-v1.0-GGUF", "file":"solar-10.7b-instruct-v1.0.Q4_K_M.gguf", "suffix":"\n### Assistant:\n", "prefix":"### User:\n"}
}

for k in llms.keys():
    AutoModelForCausalLM.from_pretrained(llms[k]['name'], model_file=llms[k]['file'])

#Pydantic object
class validation(BaseModel):
    prompt: str
    llm: str
#Fast API
app = FastAPI()

@app.post("/llm_on_cpu")
async def stream(item: validation):
    
    

    prefix=llms[item.llm]['prefix']
    suffix=llms[item.llm]['suffix']
    user="""
    {prompt}"""
    
    llm = AutoModelForCausalLM.from_pretrained(llms[item.llm]['name'], model_file=llms[item.llm]['file'])

    prompt = f"{prefix}{user.replace('{prompt}', item.prompt)}{suffix}"
    return llm(prompt)