app.py

import gradio as gr
from transformers import AutoModel, AutoTokenizer, AutoImageProcessor
import torch
import torchvision.transforms as T
from PIL import Image
import logging

logging.basicConfig(level=logging.INFO)
from torchvision.transforms.functional import InterpolationMode
import os
from huggingface_hub import login
hf_token = os.environ.get('hf_token', None)

# Define the path to your model
path = "h2oai/h2o-mississippi-2b"

# image preprocesing
IMAGENET_MEAN = (0.485, 0.456, 0.406)
IMAGENET_STD = (0.229, 0.224, 0.225)

def build_transform(input_size):
    MEAN, STD = IMAGENET_MEAN, IMAGENET_STD
    transform = T.Compose([
        T.Lambda(lambda img: img.convert('RGB') if img.mode != 'RGB' else img),
        T.Resize((input_size, input_size), interpolation=InterpolationMode.BICUBIC),
        T.ToTensor(),
        T.Normalize(mean=MEAN, std=STD)
    ])
    return transform

def find_closest_aspect_ratio(aspect_ratio, target_ratios, width, height, image_size):
    best_ratio_diff = float('inf')
    best_ratio = (1, 1)
    area = width * height
    for ratio in target_ratios:
        target_aspect_ratio = ratio[0] / ratio[1]
        ratio_diff = abs(aspect_ratio - target_aspect_ratio)
        if ratio_diff < best_ratio_diff:
            best_ratio_diff = ratio_diff
            best_ratio = ratio
        elif ratio_diff == best_ratio_diff:
            if area > 0.5 * image_size * image_size * ratio[0] * ratio[1]:
                best_ratio = ratio
    return best_ratio

def dynamic_preprocess(image, min_num=1, max_num=6, image_size=448, use_thumbnail=False):
    orig_width, orig_height = image.size
    aspect_ratio = orig_width / orig_height

    # calculate the existing image aspect ratio
    target_ratios = set(
        (i, j) for n in range(min_num, max_num + 1) for i in range(1, n + 1) for j in range(1, n + 1) if
        i * j <= max_num and i * j >= min_num)
    target_ratios = sorted(target_ratios, key=lambda x: x[0] * x[1])

    # find the closest aspect ratio to the target
    target_aspect_ratio = find_closest_aspect_ratio(
        aspect_ratio, target_ratios, orig_width, orig_height, image_size)

    # calculate the target width and height
    target_width = image_size * target_aspect_ratio[0]
    target_height = image_size * target_aspect_ratio[1]
    blocks = target_aspect_ratio[0] * target_aspect_ratio[1]

    # resize the image
    resized_img = image.resize((target_width, target_height))
    processed_images = []
    for i in range(blocks):
        box = (
            (i % (target_width // image_size)) * image_size,
            (i // (target_width // image_size)) * image_size,
            ((i % (target_width // image_size)) + 1) * image_size,
            ((i // (target_width // image_size)) + 1) * image_size
        )
        # split the image
        split_img = resized_img.crop(box)
        processed_images.append(split_img)
    assert len(processed_images) == blocks
    if use_thumbnail and len(processed_images) != 1:
        thumbnail_img = image.resize((image_size, image_size))
        processed_images.append(thumbnail_img)
    return processed_images, target_aspect_ratio

def dynamic_preprocess2(image, min_num=1, max_num=6, image_size=448, use_thumbnail=False, prior_aspect_ratio=None):
    orig_width, orig_height = image.size
    aspect_ratio = orig_width / orig_height

    # calculate the existing image aspect ratio
    target_ratios = set(
        (i, j) for n in range(min_num, max_num + 1) for i in range(1, n + 1) for j in range(1, n + 1) if
        i * j <= max_num and i * j >= min_num)
    target_ratios = sorted(target_ratios, key=lambda x: x[0] * x[1])

    new_target_ratios = []
    if prior_aspect_ratio is not None:
        for i in target_ratios:
            if prior_aspect_ratio[0]%i[0] != 0 and prior_aspect_ratio[1]%i[1] != 0:
                new_target_ratios.append(i)
            else:
                continue

    # find the closest aspect ratio to the target
    target_aspect_ratio = find_closest_aspect_ratio(
        aspect_ratio, new_target_ratios, orig_width, orig_height, image_size)

    # calculate the target width and height
    target_width = image_size * target_aspect_ratio[0]
    target_height = image_size * target_aspect_ratio[1]
    blocks = target_aspect_ratio[0] * target_aspect_ratio[1]

    # resize the image
    resized_img = image.resize((target_width, target_height))
    processed_images = []
    for i in range(blocks):
        box = (
            (i % (target_width // image_size)) * image_size,
            (i // (target_width // image_size)) * image_size,
            ((i % (target_width // image_size)) + 1) * image_size,
            ((i // (target_width // image_size)) + 1) * image_size
        )
        # split the image
        split_img = resized_img.crop(box)
        processed_images.append(split_img)
    assert len(processed_images) == blocks
    if use_thumbnail and len(processed_images) != 1:
        thumbnail_img = image.resize((image_size, image_size))
        processed_images.append(thumbnail_img)
    return processed_images
def load_image1(image_file, input_size=448, min_num=1, max_num=12):
    if isinstance(image_file, str):
        image = Image.open(image_file).convert('RGB')
    else:
        image = image_file
    transform = build_transform(input_size=input_size)
    images, target_aspect_ratio = dynamic_preprocess(image, image_size=input_size, use_thumbnail=True, min_num=min_num, max_num=max_num)
    pixel_values = [transform(image) for image in images]
    pixel_values = torch.stack(pixel_values)
    return pixel_values, target_aspect_ratio

def load_image2(image_file, input_size=448, min_num=1, max_num=12, target_aspect_ratio=None):
    
    if isinstance(image_file, str):
        image = Image.open(image_file).convert('RGB')
    else:
        image = image_file
    transform = build_transform(input_size=input_size)
    images = dynamic_preprocess2(image, image_size=input_size, use_thumbnail=True, min_num=min_num, max_num=max_num, prior_aspect_ratio=target_aspect_ratio)
    pixel_values = [transform(image) for image in images]
    pixel_values = torch.stack(pixel_values)
    return pixel_values

def load_image_msac(file_name):
    pixel_values, target_aspect_ratio = load_image1(file_name, min_num=1, max_num=6)
    pixel_values = pixel_values.to(torch.bfloat16).cuda()
    pixel_values2 = load_image2(file_name, min_num=3, max_num=6, target_aspect_ratio=target_aspect_ratio)
    pixel_values2 = pixel_values2.to(torch.bfloat16).cuda()
    pixel_values = torch.cat([pixel_values2[:-1], pixel_values[:-1], pixel_values2[-1:]], 0)
    return pixel_values
# Load the model and tokenizer
model = AutoModel.from_pretrained(
    path,
    torch_dtype=torch.bfloat16,
    low_cpu_mem_usage=True,
    trust_remote_code=True,
    use_auth_token=hf_token
).eval().cuda()

tokenizer = AutoTokenizer.from_pretrained(
    path, 
    trust_remote_code=True, 
    use_fast=False,
    use_auth_token=hf_token
)
tokenizer.pad_token = tokenizer.unk_token
tokenizer.eos_token = "<|end|>"
model.generation_config.pad_token_id = tokenizer.pad_token_id


def inference(image, user_message, temperature, top_p, max_new_tokens, chatbot,state, image_state):
    # if image is provided, store it in image_state:
    if chatbot is None:
        chatbot = []
        
    if image is not None:
        image_state = load_image_msac(image)
    else:
        # If image_state is None, then no image has been provided yet
        if image_state is None:
            chatbot.append(("System", "Please provide an image to start the conversation."))
            return chatbot, state, image_state, ""
        
    # Initialize history (state) if it's None
    if state is None:
        state = None  # model.chat function handles None as empty history        

    # Append user message to chatbot
    chatbot.append((user_message, None))

    # Set generation config
    do_sample = (float(temperature) != 0.0)    


    generation_config = dict(
        num_beams=1,
        max_new_tokens=int(max_new_tokens),
        do_sample=do_sample,
        temperature= float(temperature),
        top_p= float(top_p),
    )

    # Call model.chat with history
    response_text, new_state = model.chat(
        tokenizer,
        image_state,
        user_message,
        generation_config=generation_config,
        history=state,
        return_history=True
    )
    
    # update the satet with new_state
    state = new_state
    # Update chatbot with the model's response
    chatbot[-1] = (user_message, response_text)    
    
    return chatbot, state, image_state, ""

def regenerate_response(chatbot, temperature, top_p, max_new_tokens, state, image_state):

    # Check if there is a previous user message
    if chatbot is None or len(chatbot) == 0:
        chatbot = []
        chatbot.append(("System", "Nothing to regenerate. Please start a conversation first."))
        return chatbot, state, image_state
    
    # Check if there is a previous user message
    if state is None or image_state is None or len(state) == 0:
        chatbot.append(("System", "Nothing to regenerate. Please start a conversation first."))
        return chatbot, state, image_state
    
    # Get the last user message
    last_user_message, last_response = chatbot[-1]
    
    state = state[:-1]  # Remove last assistant's response from history
   
    if len(state) == 0:
        state = None
    # Set generation config
    do_sample = (float(temperature) != 0.0)    

    generation_config = dict(
        num_beams=1,
        max_new_tokens=int(max_new_tokens),
        do_sample=do_sample,
        temperature= float(temperature),
        top_p= float(top_p),
    )
    # Regenerate the response
    response_text, new_state = model.chat(
        tokenizer,
        image_state,
        last_user_message,
        generation_config=generation_config,
        history=state,  # Exclude last assistant's response
        return_history=True
    )
    
    # Update the state with new_state
    state = new_state
    
    # Update chatbot with the regenerated response
    chatbot.append((last_user_message, response_text))     
       
    return chatbot, state, image_state


def clear_all():
    return [], None, None, None  # Clear chatbot, state, image_state, image_input


# Build the Gradio interface
with gr.Blocks() as demo:
    gr.Markdown("# **H2O-Mississippi**")
    
    state= gr.State()
    image_state = gr.State()
    

    with gr.Row():
        # First column with image input
        with gr.Column(scale=1):
            image_input = gr.Image(type="pil", label="Upload an Image")
     
        # Second column with chatbot and user input
        with gr.Column(scale=2):    
            chatbot = gr.Chatbot(label="Conversation")
            user_input = gr.Textbox(label="What is your question", placeholder="Type your message here")

        
    with gr.Accordion('Parameters', open=False):
        with gr.Row():
            temperature_input = gr.Slider(
                minimum=0.0, 
                maximum=1.0, 
                step=0.1, 
                value=0.0, 
                interactive=True,
                label="Temperature")
            top_p_input = gr.Slider(
                minimum=0.0, 
                maximum=1.0, 
                step=0.1, 
                value=0.9,
                interactive=True, 
                label="Top P")
            max_new_tokens_input = gr.Slider(
                minimum=0, 
                maximum=4096, 
                step=64, 
                value=1024, 
                interactive=True,
                label="Max New Tokens (default: 1024)"
            )
    with gr.Row():
        submit_button = gr.Button("Submit")
        regenerate_button = gr.Button("Regenerate")
        clear_button = gr.Button("Clear")
        

    # When the submit button is clicked, call the inference function
    submit_button.click(
        fn=inference, 
        inputs=[image_input, user_input, temperature_input, top_p_input, max_new_tokens_input, chatbot, state, image_state], 
        outputs=[chatbot, state, image_state, user_input]
    )
    # When the regenerate button is clicked, re-run the last inference
    regenerate_button.click(
        fn=regenerate_response,
        inputs=[chatbot, temperature_input, top_p_input,max_new_tokens_input, state, image_state],
        outputs=[chatbot, state, image_state]
    )


    clear_button.click(
        fn=clear_all, 
        inputs=None, 
        outputs=[chatbot, state, image_state, image_input]
    )

demo.launch()