app.py

import gradio as gr
import torch
import transformers
from transformers import AutoModelForCausalLM, AutoTokenizer
from PIL import Image
import warnings

# disable some warnings
transformers.logging.set_verbosity_error()
transformers.logging.disable_progress_bar()
warnings.filterwarnings('ignore')

# set device
torch.set_default_device('cuda')  # or 'cpu'

model_name = 'cognitivecomputations/dolphin-vision-7b'

# create model
model = AutoModelForCausalLM.from_pretrained(
    model_name,
    torch_dtype=torch.float16,
    device_map='auto',
    trust_remote_code=True)
tokenizer = AutoTokenizer.from_pretrained(
    model_name,
    trust_remote_code=True)

def inference(prompt, image):
    messages = [
        {"role": "user", "content": f'<image>\n{prompt}'}
    ]
    text = tokenizer.apply_chat_template(
        messages,
        tokenize=False,
        add_generation_prompt=True
    )

    text_chunks = [tokenizer(chunk).input_ids for chunk in text.split('<image>')]
    input_ids = torch.tensor(text_chunks[0] + [-200] + text_chunks[1], dtype=torch.long).unsqueeze(0)

    image_tensor = model.process_images([image], model.config).to(dtype=model.dtype)

    # generate
    output_ids = model.generate(
        input_ids,
        images=image_tensor,
        max_new_tokens=2048,
        use_cache=True)[0]

    return tokenizer.decode(output_ids[input_ids.shape[1]:], skip_special_tokens=True).strip()

with gr.Blocks() as demo:
    with gr.Row():
        with gr.Column():
            prompt_input = gr.Textbox(label="Prompt", placeholder="Describe this image in detail")
            image_input = gr.Image(label="Image", type="pil")
            submit_button = gr.Button("Submit")
        with gr.Column():
            output_text = gr.Textbox(label="Output")

    submit_button.click(fn=inference, inputs=[prompt_input, image_input], outputs=output_text)

demo.launch()