app.py

"""
Adapted from: https://github.com/Vision-CAIR/MiniGPT-4/blob/main/demo.py
"""
import argparse
import os
import random

import numpy as np
import torch
import torch.backends.cudnn as cudnn
import gradio as gr

from video_llama.common.config import Config
from video_llama.common.dist_utils import get_rank
from video_llama.common.registry import registry
from video_llama.conversation.conversation_video import Chat, Conversation, default_conversation,SeparatorStyle,conv_llava_llama_2
import decord
decord.bridge.set_bridge('torch')

#%%
# imports modules for registration
from video_llama.datasets.builders import *
from video_llama.models import *
from video_llama.processors import *
from video_llama.runners import *
from video_llama.tasks import *

#%%
def parse_args():
    parser = argparse.ArgumentParser(description="Demo")
    parser.add_argument("--cfg-path", default='eval_configs/video_llama_eval_withaudio.yaml', help="path to configuration file.")
    parser.add_argument("--gpu-id", type=int, default=0, help="specify the gpu to load the model.")
    parser.add_argument("--model_type", type=str, default='vicuna', help="The type of LLM")
    parser.add_argument(
        "--options",
        nargs="+",
        help="override some settings in the used config, the key-value pair "
        "in xxx=yyy format will be merged into config file (deprecate), "
        "change to --cfg-options instead.",
    )
    args = parser.parse_args()
    return args


def setup_seeds(config):
    seed = config.run_cfg.seed + get_rank()

    random.seed(seed)
    np.random.seed(seed)
    torch.manual_seed(seed)

    cudnn.benchmark = False
    cudnn.deterministic = True


# ========================================
#             Model Initialization
# ========================================

print('Initializing Chat')
args = parse_args()
cfg = Config(args)

model_config = cfg.model_cfg
model_config.device_8bit = args.gpu_id
model_cls = registry.get_model_class(model_config.arch)
model = model_cls.from_config(model_config).to('cuda:{}'.format(args.gpu_id))
model.eval()
vis_processor_cfg = cfg.datasets_cfg.webvid.vis_processor.train
vis_processor = registry.get_processor_class(vis_processor_cfg.name).from_config(vis_processor_cfg)
chat = Chat(model, vis_processor, device='cuda:{}'.format(args.gpu_id))
print('Initialization Finished')

# ========================================
#             Gradio Setting
# ========================================

def gradio_ask(user_message, chatbot, chat_state):
    print("building prompt...")
    if len(user_message) == 0:
        return gr.update(interactive=True, placeholder='Input should not be empty!'), chatbot, chat_state
    chat.ask(user_message, chat_state)
    chatbot = chatbot + [[user_message, None]]
    return '', chatbot, chat_state


def gradio_answer(chatbot, chat_state, img_list, num_beams, temperature):
    print("generating answer...")
    llm_message = chat.answer(conv=chat_state,
                              img_list=img_list,
                              num_beams=1,
                              temperature=temperature,
                              max_new_tokens=240,
                              max_length=511)[0]
    chatbot[-1][1] = llm_message
    print(chat_state.get_prompt())
    print(chat_state)
    return chatbot, chat_state, img_list

def gradio_reset(chat_state, img_list):
    if chat_state is not None:
        chat_state.messages = []
    if img_list is not None:
        img_list = []
    return None, gr.update(value=None, interactive=True), gr.update(value=None, interactive=True), gr.update(placeholder='Please upload your video first', interactive=False),gr.update(value="Upload & Start Chat", interactive=True), chat_state, img_list

def upload_imgorvideo(gr_video, gr_img, text_input,chatbot,audio_flag):
    if args.model_type == 'vicuna':
        chat_state = default_conversation.copy()
    else:
        chat_state = conv_llava_llama_2.copy()
    if gr_img is None and gr_video is None:
        return None, None, None, gr.update(interactive=True), chat_state, None
    elif gr_video is not None:
        print(gr_video)
        chatbot = [((gr_video,), None)]
        chat_state = default_conversation.copy()
        chat_state = Conversation(
            system= "You are able to understand the visual content that the user provides."
           "Follow the instructions carefully and explain your answers in detail.",
            roles=("Human", "Assistant"),
            messages=[],
            offset=0,
            sep_style=SeparatorStyle.SINGLE,
            sep="###",
        )
        img_list = []
        llm_message = chat.upload_video(gr_video, chat_state, img_list)
        llm_message = chat.ask(text_input, chat_state)
        llm_message = chat.answer(conv=chat_state,
                              img_list=img_list,
                              num_beams=1,
                              temperature=1.0,
                              max_new_tokens=240,
                              max_length=511)[0]
        print(llm_message)
        output = [[llm_message]]
        return llm_message, output
    elif gr_img is not None:
        print(gr_img)
        chatbot = [((gr_img,), None)]
        chat_state = Conversation(
            system= "You are able to understand the visual content that the user provides."
           "Follow the instructions carefully and explain your answers in detail.",
            roles=("Human", "Assistant"),
            messages=[],
            offset=0,
            sep_style=SeparatorStyle.SINGLE,
            sep="###",
        )
        img_list = []
        llm_message = chat.upload_img(gr_img, chat_state, img_list)
        return gr.update(interactive=False), gr.update(interactive=False), gr.update(interactive=True, placeholder='Type and press Enter'), gr.update(value="Start Chatting", interactive=False), chat_state, img_list,chatbot
    else:
        # img_list = []
        return gr.update(interactive=False), gr.update(interactive=False, placeholder='Currently, only one input is supported'), gr.update(value="Currently, only one input is supported", interactive=False), chat_state, None,chatbot


title = """
<h1 align="center"><a href="https://github.com/DAMO-NLP-SG/Video-LLaMA"><img src="https://s1.ax1x.com/2023/05/22/p9oQ0FP.jpg", alt="Video-LLaMA" border="0" style="margin: 0 auto; height: 200px;" /></a> </h1>

<h1><img src="https://upload.wikimedia.org/wikipedia/commons/thumb/5/51/IBM_logo.svg/1000px-IBM_logo.svg.png", alt="Video-LLaMA" border="0" style="margin: 0 auto; height: 200px;" /></h1>

<h1 align="center">Video-LLaMA-2: An Instruction-tuned Audio-Visual Language Model for Video Understanding</h1>

<h5 align="center">  Introduction: Video-LLaMA is a multi-model large language model that achieves video-grounded conversations between humans and computers \
    by connecting language decoder with off-the-shelf unimodal pre-trained models. </h5> 


Current online demo uses the 7B version of Video-LLaMA-2 due to resource limitations of running on a Nvidia A10.

From the IBM Generative AI Italy team who better adapted the model for LLAMA-2-7B. For any issue contact [email protected]
"""

cite_markdown = ("""
## Citation
If you find our project useful, hope you can star our repo and cite our paper as follows:
```
@article{damonlpsg2023videollama,
  author = {Zhang, Hang and Li, Xin and Bing, Lidong},
  title = {Video-LLaMA: An Instruction-tuned Audio-Visual Language Model for Video Understanding},
  year = 2023,
  journal = {arXiv preprint arXiv:2306.02858}
  url = {https://arxiv.org/abs/2306.02858}
}
""")

case_note_upload = ("""
### We provide some examples at the bottom of the page. Simply click on them to try them out directly.
""")

#TODO show examples below

with gr.Blocks() as demo:
    gr.Markdown(title)

    with gr.Row():
        with gr.Column(scale=0.5):
            video = gr.Video()
            image = gr.Image(type="filepath")
            gr.Markdown(case_note_upload)

            upload_button = gr.Button(value="Upload & Start Chat", interactive=True, variant="primary")
            clear = gr.Button("Restart")
            
            num_beams = gr.Slider(
                minimum=1,
                maximum=10,
                value=1,
                step=1,
                interactive=True,
                label="beam search numbers)",
            )
            
            temperature = gr.Slider(
                minimum=0.1,
                maximum=2.0,
                value=1.0,
                step=0.1,
                interactive=True,
                label="Temperature",
            )

            audio = gr.Checkbox(interactive=True, value=False, label="Audio")
        with gr.Column():
            chat_state = gr.State()
            img_list = gr.State()
            chatbot = gr.Chatbot(label='Video-LLaMA')
            text_input = gr.Textbox(label='User', placeholder='Upload your image/video first, or directly click the examples at the bottom of the page.', interactive=False)
            output = gr.Textbox(label='Output')
            
        
    gr.Markdown(cite_markdown)
    #upload_button.click(upload_imgorvideo, inputs=[video, image, text_input], outputs=[chat_state,chatbot])
    
    text_input.submit(upload_imgorvideo, inputs=[video, image, text_input], outputs=[output])
    #clear.click(gradio_reset, [chat_state, img_list], [chatbot, video, image, text_input, upload_button, chat_state, img_list], queue=False)
    
demo.queue().launch(debug=True)


# %%