app.py

import gradio as gr
from transformers import AutoTokenizer, AutoModelForCausalLM
from peft import PeftModel
import torch

# Base model (4-bit LLaMA)
base_model_id = "unsloth/Llama-3.2-3B-Instruct-bnb-4bit"

# Your LoRA adapter repo
adapter_id = "Christi049/meal-gen-adapter"

# Load tokenizer
tokenizer = AutoTokenizer.from_pretrained(base_model_id)

# Load base model
base_model = AutoModelForCausalLM.from_pretrained(
    base_model_id,
    torch_dtype=torch.float16,
    device_map="auto"
)

# Apply LoRA adapter
model = PeftModel.from_pretrained(base_model, adapter_id)

def generate_meal(prompt):
    inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
    outputs = model.generate(**inputs, max_new_tokens=300, temperature=0.7)
    return tokenizer.decode(outputs[0], skip_special_tokens=True)

iface = gr.Interface(
    fn=generate_meal,
    inputs=gr.Textbox(label="Meal Request", placeholder="e.g., Generate a 7-day vegetarian meal plan"),
    outputs=gr.Textbox(label="Generated Meal Plan"),
    title="Weekly Meal Generator"
)

iface.launch()