app.py

import os
from fastapi import FastAPI, HTTPException
from pydantic import BaseModel
from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

# Disable hf_transfer
os.environ["HF_HUB_ENABLE_HF_TRANSFER"] = "false"

app = FastAPI()

# Load your fine-tuned model and tokenizer
model_name = "OnlyCheeini/greesychat-turbo"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(model_name, torch_dtype=torch.float16).to("cuda")

class OpenAIRequest(BaseModel):
    model: str
    prompt: str
    max_tokens: int = 64
    temperature: float = 0.7
    top_p: float = 0.9

class OpenAIResponse(BaseModel):
    choices: list

@app.post("/v1/completions", response_model=OpenAIResponse)
async def generate_text(request: OpenAIRequest):
    if request.model != model_name:
        raise HTTPException(status_code=400, detail="Model not found")

    inputs = tokenizer(request.prompt, return_tensors="pt").to("cuda")
    outputs = model.generate(
        **inputs,
        max_length=inputs['input_ids'].shape[1] + request.max_tokens,
        temperature=request.temperature,
        top_p=request.top_p,
    )

    generated_text = tokenizer.decode(outputs[0], skip_special_tokens=True)
    return OpenAIResponse(choices=[{"text": generated_text}])

if __name__ == "__main__":
    import uvicorn
    uvicorn.run(app, host="0.0.0.0", port=8000)