use.py

import torch
from transformers import AutoModelForCausalLM, AutoTokenizer

def run_pin_inference(prompt, model_id="LH-Tech-AI/Pin-Tiny", subfolder="Pin-25M"):
    # 1. Device Setup
    device = "cuda" if torch.cuda.is_available() else "cpu"
    print(f"Using device: {device}")

    # 2. Load tokenizer and model
    tokenizer = AutoTokenizer.from_pretrained("gpt2")
    tokenizer.pad_token = tokenizer.eos_token
    
    model = AutoModelForCausalLM.from_pretrained(model_id, subfolder=subfolder).to(device)

    # 3. Format prompt
    formatted_prompt = f"[INST] {prompt} [/INST]"
    inputs = tokenizer(formatted_prompt, return_tensors="pt").to(device)

    # 4. Generate
    with torch.no_grad():
        outputs = model.generate(
            **inputs,
            max_new_tokens=64,
            temperature=0.7,
            do_sample=True,
            pad_token_id=tokenizer.eos_token_id,
            eos_token_id=tokenizer.encode("[")[0] 
        )

    # 5. Decode & Cleanup
    full_text = tokenizer.decode(outputs[0], skip_special_tokens=True)
    
    if "[/INST]" in full_text:
        response = full_text.split("[/INST]")[-1].split("[INST]")[0].strip()
    else:
        response = full_text
        
    return response

# --- Sample test ---
if __name__ == "__main__":
    user_query = "What is the weather like today?"
    answer = run_pin_inference(user_query, model_id="LH-Tech-AI/Pin", subfolder="Pin-25M")
    
    print(f"\nUser: {user_query}")
    print(f"Pin: {answer}")