app.py

import gradio as gr
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer, StoppingCriteria
import gradio as gr
import spaces
import torch
import numpy as np
import torch
import torchvision.transforms as T
from PIL import Image
from torchvision.transforms.functional import InterpolationMode
from transformers import AutoModel, AutoTokenizer
from PIL import Image, ExifTags

from threading import Thread
import re
import time 
from PIL import Image
import torch
import spaces
import subprocess
import os

subprocess.run('pip install flash-attn --no-build-isolation', env={'FLASH_ATTENTION_SKIP_CUDA_BUILD': "TRUE"}, shell=True)

torch.set_default_device('cuda')


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=12, 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

def correct_image_orientation(image_path):
    # Mở ảnh
    image = Image.open(image_path)

    # Kiểm tra dữ liệu Exif (nếu có)
    try:
        exif = image._getexif()
        if exif is not None:
            for tag, value in exif.items():
                if ExifTags.TAGS.get(tag) == "Orientation":
                    # Sửa hướng dựa trên Orientation
                    if value == 3:
                        image = image.rotate(180, expand=True)
                    elif value == 6:
                        image = image.rotate(-90, expand=True)
                    elif value == 8:
                        image = image.rotate(90, expand=True)
                    break
    except Exception as e:
        print("Không thể xử lý Exif:", e)

    return image
    
def load_image(image_file, input_size=448, max_num=12):
    image = correct_image_orientation(image_file).convert('RGB')
    print("Image size: ", image.size)
    transform = build_transform(input_size=input_size)
    images = dynamic_preprocess(image, image_size=input_size, use_thumbnail=True, max_num=max_num)
    pixel_values = [transform(image) for image in images]
    pixel_values = torch.stack(pixel_values)
    return pixel_values

model = AutoModel.from_pretrained(
    "5CD-AI/Vintern-1B-v3_5",
    torch_dtype=torch.bfloat16,
    low_cpu_mem_usage=True,
    trust_remote_code=True,
).eval().cuda()
tokenizer = AutoTokenizer.from_pretrained("5CD-AI/Vintern-1B-v3_5", trust_remote_code=True, use_fast=False)
    
@spaces.GPU
def chat(message, history):
    print("history",history)
    print("message",message)

    if len(history) != 0 and len(message["files"]) != 0:
        return """Chúng tôi hiện chỉ hổ trợ 1 ảnh ở đầu ngữ cảnh! Vui lòng tạo mới cuộc trò chuyện.
We currently only support one image at the start of the context! Please start a new conversation."""
    
    if len(history) == 0 and len(message["files"]) != 0:
        if "path" in message["files"][0]:
            test_image = message["files"][0]["path"]
        else:
            test_image = message["files"][0]
        pixel_values = load_image(test_image, max_num=6).to(torch.bfloat16).cuda()
    elif len(history) == 0 and len(message["files"]) == 0:
        pixel_values = None     
    elif history[0][0][0] is not None and os.path.isfile(history[0][0][0]):
        test_image = history[0][0][0]
        pixel_values = load_image(test_image, max_num=6).to(torch.bfloat16).cuda()
    else:
        pixel_values = None 
        
    
    generation_config = dict(max_new_tokens= 700, do_sample=False, num_beams = 3, repetition_penalty=2.5)
    
    if len(history) == 0:
        if pixel_values is not None:
            question = '<image>\n'+message["text"]
        else:
            question = message["text"]
        response, conv_history = model.chat(tokenizer, pixel_values, question, generation_config, history=None, return_history=True)
    else:
        conv_history = []
        if history[0][0][0] is not None and os.path.isfile(history[0][0][0]):
            start_index = 1
        else:
            start_index = 0
        
        for i, chat_pair in enumerate(history[start_index:]):
            if i == 0 and start_index == 1:
                 conv_history.append(tuple(['<image>\n'+chat_pair[0],chat_pair[1]]))
            else:
                conv_history.append(tuple(chat_pair))

            
        print("conv_history",conv_history)
        question = message["text"]
        response, conv_history = model.chat(tokenizer, pixel_values, question, generation_config, history=conv_history, return_history=True)
        
    print(f'User: {question}\nAssistant: {response}')

    # return response
    buffer = ""
    for new_text in response:
      buffer += new_text
      generated_text_without_prompt = buffer[:]
      time.sleep(0.02)
      yield generated_text_without_prompt

CSS ="""
#component-10 {
  height: 70dvh !important;
  transform-origin: top; /* Đảm bảo rằng phần tử mở rộng từ trên xuống */
  border-style: solid;
  overflow: hidden;
  flex-grow: 1;
  min-width: min(160px, 100%);
  border-width: var(--block-border-width);
}

/* Đảm bảo ảnh bên trong nút hiển thị đúng cách cho các nút có aria-label chỉ định */
button.svelte-1lcyrx4[aria-label="user's message: a file of type image/jpeg, "] img.svelte-1pijsyv {
  width: 100%;
  object-fit: contain;
  height: 100%;
  border-radius: 13px; /* Thêm bo góc cho ảnh */
  max-width: 50vw;     /* Giới hạn chiều rộng ảnh */
}
/* Đặt chiều cao cho nút và cho phép chọn văn bản chỉ cho các nút có aria-label chỉ định */
button.svelte-1lcyrx4[aria-label="user's message: a file of type image/jpeg, "] {
  user-select: text;
  text-align: left;
  height: 300px;
}
/* Thêm bo góc và giới hạn chiều rộng cho ảnh không thuộc avatar container */
.message-wrap.svelte-1lcyrx4 > div.svelte-1lcyrx4 .svelte-1lcyrx4:not(.avatar-container) img {
  border-radius: 13px;
  max-width: 50vw;
}
.message-wrap.svelte-1lcyrx4 .message.svelte-1lcyrx4 img {
    margin: var(--size-2);
    max-height: 500px;
}
.image-preview-close-button {
  position: relative; /* Nếu cần định vị trí */
  width: 5%; /* Chiều rộng nút */
  height: 5%; /* Chiều cao nút */
  display: flex;
  justify-content: center;
  align-items: center;
  padding: 0; /* Để tránh ảnh hưởng từ padding mặc định */
  border: none; /* Tùy chọn để loại bỏ đường viền */
  background: none; /* Tùy chọn để loại bỏ nền */
}

.example-image-container.svelte-9pi8y1 {
    width: calc(var(--size-8) * 5);
    height: calc(var(--size-8) * 5);
    border-radius: var(--radius-lg);
    overflow: hidden;
    position: relative;
    margin-bottom: var(--spacing-lg);
}
"""

js = """
function forceLightTheme() {
    const url = new URL(window.location);

    // Cập nhật __theme thành light nếu giá trị không đúng
    if (url.searchParams.get('__theme') !== 'light') {
        url.searchParams.set('__theme', 'light');
        // Thay đổi URL mà không tải lại trang nếu cần
        window.history.replaceState({}, '', url.href);
    }

    // Đảm bảo document luôn áp dụng theme light
    document.documentElement.setAttribute('data-theme', 'light');
}
"""
from transformers import pipeline

pipe = pipeline("automatic-speech-recognition", model="openai/whisper-large-v3-turbo", torch_dtype=torch.float16, device="cuda:0")

@spaces.GPU
def transcribe_speech(filepath):
    output = pipe(
        filepath,
        max_new_tokens=256,
        generate_kwargs={
            "task": "transcribe",
        },  
        chunk_length_s=30,
        batch_size=1,
    )
    return output["text"]
        
demo = gr.Blocks(css=CSS,js=js, theme='NoCrypt/miku')

with demo:
    chat_demo_interface = gr.ChatInterface(
        fn=chat,
        description="""**Vintern-1B-v3.5** is the latest in the Vintern series, bringing major improvements over v2 across all benchmarks. This **continuous fine-tuning Version** enhances Vietnamese capabilities while retaining strong English performance. It excels in OCR, text recognition, and Vietnam-specific document understanding.""",
        examples=[{"text": "Hãy viết một email giới thiệu sản phẩm trong ảnh.", "files":["./demo_3.jpg"]},
                  {"text": "Trích xuất các thông tin từ ảnh trả về markdown.", "files":["./demo_1.jpg"]}, 
                  {"text": "Bạn là nhân viên marketing chuyên nghiệp. Hãy viết một bài quảng cáo dài trên mạng xã hội giới thiệu về cửa hàng.", "files":["./demo_2.jpg"]},
                  {"text": "Trích xuất thông tin kiện hàng trong ảnh và trả về dạng JSON.", "files":["./demo_4.jpg"]}],
        title="❄️ Vintern-1B-v3.5 Demo ❄️",
        multimodal=True,
        css=CSS,
        js=js,
        theme='NoCrypt/miku'
    )
        
    # mic_transcribe = gr.Interface(
    #     fn=transcribe_speech,
    #     inputs=gr.Audio(sources="microphone", type="filepath", editable=False),
    #     outputs=gr.components.Textbox(),
    # )

# chat_demo_interface.queue()
demo.queue().launch()