import os
from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

MODEL_NAME = "bigcode/starcoderbase-1b"
HF_TOKEN = os.getenv("HUGGINGFACE_TOKEN")

# Force CPU mode
device = "cpu"  # Change this from "cuda"

tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME, token=HF_TOKEN)
model = AutoModelForCausalLM.from_pretrained(
    MODEL_NAME,
    token=HF_TOKEN,
    torch_dtype=torch.float16,  # Keep memory low
    device_map="auto",  # Still allows auto placement
    trust_remote_code=True
).to(device)

def generate_code(prompt: str, max_tokens: int = 256):
    inputs = tokenizer(prompt, return_tensors="pt").to(device)
    output = model.generate(**inputs, max_new_tokens=max_tokens)
    return tokenizer.decode(output[0], skip_special_tokens=True)