app.py

from transformers import AutoProcessor
from unsloth import UnslothModel
from PIL import Image
import gradio as gr
import torch

model, tokenizer = UnslothModel.from_pretrained(
    model_name="unsloth/Llama-3.2-11B-Vision-bnb-4bit",
    adapter_path="ArnavLatiyan/my-lora-leaf-model",
    load_in_4bit=True
)
processor = AutoProcessor.from_pretrained("unsloth/llama-3.2-vision-11b")

def predict(image):
    inputs = processor(images=image, return_tensors="pt").to("cuda" if torch.cuda.is_available() else "cpu")
    prompt = tokenizer("What disease is this?", return_tensors="pt").to(inputs["pixel_values"].device)
    inputs.update(prompt)
    outputs = model.generate(**inputs, max_new_tokens=50)
    return tokenizer.decode(outputs[0], skip_special_tokens=True)

gr.Interface(fn=predict, inputs="image", outputs="text").launch()