app.py

import torch
import gradio as gr
from transformers import AutoTokenizer, AutoModelForCausalLM, BitsAndBytesConfig

base_model = "stevenArtificial/Babaru-Llama-3.2-1B-Instruct"

tokenizer = AutoTokenizer.from_pretrained(base_model)

quantization_config = BitsAndBytesConfig(
    load_in_4bit=True,
    bnb_4bit_quant_type="nf4",
    bnb_4bit_use_double_quant=True,
    bnb_4bit_compute_dtype=torch.float16
)

model = AutoModelForCausalLM.from_pretrained(
    base_model,
    quantization_config=quantization_config,
    device_map="auto"
)

def generate(prompt):
    inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
    outputs = model.generate(**inputs, max_new_tokens=256, do_sample=True, temperature=0.7)
    return tokenizer.decode(outputs[0], skip_special_tokens=True)

iface = gr.Interface(
    fn=generate,
    inputs=gr.Textbox(lines=4, label="Prompt"),
    outputs=gr.Textbox(label="Response"),
    title="Babaru LLaMA-3.2-1B-Instruct Chatbot",
    description="Interactive demo using Babaru fine-tuned LLaMA-3.2-1B-Instruct by stevenArtificial."
)

iface.launch()