File size: 10,612 Bytes
b2afdba
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
"""
Adapted from: https://github.com/Vision-CAIR/MiniGPT-4/blob/main/demo.py
"""
import argparse
import os
import sys
import random

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

from global_local.common.config import Config
from global_local.common.dist_utils import get_rank
from global_local.common.registry import registry
from global_local.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 global_local.datasets.builders import *
from global_local.models import *
from global_local.processors import *
from global_local.runners import *
from global_local.tasks import *

#%%
def parse_args():
    parser = argparse.ArgumentParser(description="Demo")
    #parser.add_argument("--cfg-path", required=True, help="path to configuration file.")
    parser.add_argument("--cfg-path", type=str, default='./eval_configs/conversation_demo.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='llama_v2', help="specify LLM")
    parser.add_argument('--pretrained_weight_path', type=str, default="./ckpt/finetuned_model.pth", metavar='PWP',
                    help='path to pretrained weight path')
    parser.add_argument('--num_frames_per_clip', type=int, default=16, metavar='NPPC',
                    help='specify how frames to use per clip')
    parser.add_argument('--num_segments', type=int, default=4, metavar='NS',
                        help='specify number of video segments')
    parser.add_argument('--hierarchical_agg_function', type=str, default="without-top-final-global-prompts-region-segment-full-dis-spatiotemporal-prompts-attn-early-attn-linear-learned", metavar='HAF',
                        help='specify function to merge global and clip visual representations')

    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.num_frames_per_clip = args.num_frames_per_clip
model.num_segments = args.num_segments
model.hierarchical_agg_function = args.hierarchical_agg_function
model.global_region_embed_weight = None

model.initialize_visual_agg_function()

best_checkpoint = torch.load(args.pretrained_weight_path, map_location='cpu')['model_state_dict']
pretrained_dict = {}
for k, v in best_checkpoint.items():
    pretrained_dict[k.replace('module.', '')] = v

model_dict = model.state_dict()
model_dict.update(pretrained_dict)
model.load_state_dict(model_dict)
model.cuda().eval()

#vis_processor_cfg = cfg.datasets_cfg.cc_sbu_align.vis_processor.train
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_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, chat_state,chatbot):
    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_img is not None and gr_video is None:
        print(gr_img)
        chatbot = chatbot + [((gr_img,), None)]
        chat_state.system =  "You are able to understand the visual content that the user provides. Follow the instructions carefully and explain your answers in detail."
        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
    elif gr_video is not None and gr_img is None:
        print(gr_video)
        chatbot = chatbot + [((gr_video,), None)]
        chat_state.system =  "You are able to understand the visual content that the user provides. Follow the instructions carefully and explain your answers in detail."
        img_list = []
        llm_message = chat.upload_video_without_audio(gr_video, 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

def gradio_ask(user_message, chatbot, chat_state):
    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):
    llm_message = chat.answer(conv=chat_state,
                              img_list=img_list,
                              num_beams=num_beams,
                              temperature=temperature,
                              max_new_tokens=300,
                              max_length=2000)[0]
    chatbot[-1][1] = llm_message
    print(chat_state.get_prompt())
    print(chat_state)
    return chatbot, chat_state, img_list

title = """
<h1 align="center">Global-Local QFormer for Long Video Understanding with LLMs</h1>

<h5 align="center">  Introduction: We introduce a Global-Local QFormer video model that is connected with a Large Language Model to understand \
                    and answer questions about long videos. </h5> 

<div style='display:flex; gap: 0.25rem; '>
<a href='https://huggingface.co/spaces/rxtan/rxtan/Global-Local-QFormer-Video-LLM'><img src='https://img.shields.io/badge/%F0%9F%A4%97%20Hugging%20Face-Spaces-blue'></a> 
<a href=''><img src='https://img.shields.io/badge/Paper-PDF-red'></a>
</div>


Thank you for using the Global-Local QFormer Demo Page! If you have any questions or feedback, feel free to contact us. 
Current online demo uses the 7B version of Llama-2 due to resource limitations.


"""

Note_markdown = ("""
### We note that our Global-Local QFormer model may be limited at understanding videos from rare domains. Due to the pretraining data, the \
    model may be susceptible to hallucinations
We would like to acknowledge the Video-LLama repository which we copied the demo layout from.

**Boston University**
""")

cite_markdown = ("""
""")

#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")
            gr.Markdown(Note_markdown)
        with gr.Column():
            chat_state = gr.State()
            img_list = gr.State()
            chatbot = gr.Chatbot(label='Global-Local QFormer')
            text_input = gr.Textbox(label='User', placeholder='Please upload your video first.', interactive=False)
            

    '''with gr.Column():
        gr.Examples(examples=[
            [f"examples/skateboarding_dog.mp4", "What is the dog doing? "],
            [f"examples/birthday.mp4", "What is the boy doing? "],
            [f"examples/IronMan.mp4", "Is the guy in the video Iron Man? "],
        ], inputs=[video, text_input])'''
        
    gr.Markdown(cite_markdown)
    upload_button.click(upload_imgorvideo, [video, text_input, chat_state,chatbot], [video, text_input, upload_button, chat_state, img_list,chatbot])
    
    text_input.submit(gradio_ask, [text_input, chatbot, chat_state], [text_input, chatbot, chat_state]).then(
        gradio_answer, [chatbot, chat_state, img_list, num_beams, temperature], [chatbot, chat_state, img_list]
    )
    clear.click(gradio_reset, [chat_state, img_list], [chatbot, video, text_input, upload_button, chat_state, img_list], queue=False)
    
#demo.launch(share=False, enable_queue=True, debug=True)
demo.launch(share=False, debug=True)