app.py

from fastapi import FastAPI, HTTPException
from pydantic import BaseModel
from transformers import AutoTokenizer, AutoModelForCausalLM, BitsAndBytesConfig
from huggingface_hub import login
import os
import torch
import uvicorn

login(os.getenv("HF_TOKEN"))

app = FastAPI(
    title="VexaAI Model-Platform: NVIDIA Nemotron-Nano-9B-V2",
    description="Self-hosted AI-Model NVIDIA Nemotron-Nano-9B-V2, powered by VexaAI.",
    version="0.9"
)

model_name = "nvidia/NVIDIA-Nemotron-Nano-9B-v2"

bnb_config = BitsAndBytesConfig(
    load_in_4bit=True,
    bnb_4bit_use_double_quant=True,
    bnb_4bit_quant_type="nf4",
    bnb_4bit_compute_dtype=torch.bfloat16
)

tokenizer = AutoTokenizer.from_pretrained(model_name, trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained(
    model_name,
    quantization_config=bnb_config,
    device_map="auto",
    trust_remote_code=True,
    torch_dtype=torch.bfloat16
)
model.eval()

class GenerateRequest(BaseModel):
    prompt: str
    max_new_tokens: int = 512
    temperature: float = 0.7

@app.post("/generate")
async def generate_text(request: GenerateRequest):
    try:
        inputs = tokenizer(request.prompt, return_tensors="pt").to(model.device)
        
        with torch.no_grad():
            outputs = model.generate(
                **inputs,
                max_new_tokens=request.max_new_tokens,
                temperature=request.temperature,
                do_sample=True,
                eos_token_id=tokenizer.eos_token_id,
                pad_token_id=tokenizer.eos_token_id
            )
        
        full_text = tokenizer.decode(outputs[0], skip_special_tokens=True)
        generated_text = full_text[len(tokenizer.decode(inputs["input_ids"][0], skip_special_tokens=True)):].strip()
        
        return {"generated_text": generated_text}
    except Exception as e:
        raise HTTPException(status_code=500, detail=f"VexaAI Model_Platform: HTTP/S error: {str(e)}")

@app.get("/")
async def root():
    return {"message": "To start generating text, use /generate."}

if __name__ == "__main__":
    uvicorn.run(app, host="0.0.0.0", port=7860)