Spaces:
Running
Running
yinanhe
commited on
Commit
•
2ae3b27
1
Parent(s):
ab968bc
[init]
Browse files- app.py +213 -0
- constants.py +55 -0
- requirements.txt +2 -0
app.py
ADDED
@@ -0,0 +1,213 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
__all__ = ['block', 'make_clickable_model', 'make_clickable_user', 'get_submissions']
|
2 |
+
import os
|
3 |
+
|
4 |
+
import gradio as gr
|
5 |
+
import pandas as pd
|
6 |
+
import json
|
7 |
+
import tempfile
|
8 |
+
|
9 |
+
from constants import *
|
10 |
+
from huggingface_hub import Repository
|
11 |
+
HF_TOKEN = os.environ.get("HF_TOKEN")
|
12 |
+
|
13 |
+
global data_component, filter_component
|
14 |
+
|
15 |
+
|
16 |
+
def upload_file(files):
|
17 |
+
file_paths = [file.name for file in files]
|
18 |
+
return file_paths
|
19 |
+
|
20 |
+
def add_new_eval(
|
21 |
+
input_file,
|
22 |
+
model_name_textbox: str,
|
23 |
+
revision_name_textbox: str,
|
24 |
+
model_link: str,
|
25 |
+
):
|
26 |
+
if input_file is None:
|
27 |
+
return "Error! Empty file!"
|
28 |
+
|
29 |
+
upload_data=json.loads(input_file)
|
30 |
+
submission_repo = Repository(local_dir=SUBMISSION_NAME, clone_from=SUBMISSION_URL, use_auth_token=HF_TOKEN, repo_type="dataset")
|
31 |
+
submission_repo.git_pull()
|
32 |
+
shutil.copyfile(CSV_DIR, os.path.join(SUBMISSION_NAME, f"{input_file}"))
|
33 |
+
|
34 |
+
csv_data = pd.read_csv(CSV_DIR)
|
35 |
+
|
36 |
+
if revision_name_textbox == '':
|
37 |
+
col = csv_data.shape[0]
|
38 |
+
model_name = model_name_textbox
|
39 |
+
else:
|
40 |
+
model_name = revision_name_textbox
|
41 |
+
model_name_list = csv_data['name']
|
42 |
+
name_list = [name.split(']')[0][1:] for name in model_name_list]
|
43 |
+
if revision_name_textbox not in name_list:
|
44 |
+
col = csv_data.shape[0]
|
45 |
+
else:
|
46 |
+
col = name_list.index(revision_name_textbox)
|
47 |
+
|
48 |
+
if model_link == '':
|
49 |
+
model_name = model_name # no url
|
50 |
+
else:
|
51 |
+
model_name = '[' + model_name + '](' + model_link + ')'
|
52 |
+
|
53 |
+
# add new data
|
54 |
+
new_data = [
|
55 |
+
model_name
|
56 |
+
]
|
57 |
+
for key in TASK_INFO:
|
58 |
+
if key in upload_data:
|
59 |
+
new_data.append(upload_data[key][0])
|
60 |
+
else:
|
61 |
+
new_data.append(0)
|
62 |
+
csv_data.loc[col] = new_data
|
63 |
+
csv_data = csv_data.to_csv(CSV_DIR, index=False)
|
64 |
+
submission_repo.push_to_hub()
|
65 |
+
return 0
|
66 |
+
|
67 |
+
def get_final_score(df):
|
68 |
+
# 分数计算公式
|
69 |
+
final_score = df.drop('name', axis=1).sum(axis=1)
|
70 |
+
# 将总分列放在第二列
|
71 |
+
df.insert(1, 'Final Score', final_score)
|
72 |
+
return df
|
73 |
+
|
74 |
+
def get_baseline_df():
|
75 |
+
submission_repo = Repository(local_dir=SUBMISSION_NAME, clone_from=SUBMISSION_URL, use_auth_token=HF_TOKEN, repo_type="dataset")
|
76 |
+
submission_repo.git_pull()
|
77 |
+
df = pd.read_csv(CSV_DIR)
|
78 |
+
df = get_final_score(df)
|
79 |
+
# calculate the final score
|
80 |
+
df = df.sort_values(by="Final Score", ascending=False)
|
81 |
+
present_columns = MODEL_INFO + checkbox_group.value
|
82 |
+
df = df[present_columns]
|
83 |
+
return df
|
84 |
+
|
85 |
+
def get_all_df():
|
86 |
+
submission_repo = Repository(local_dir=SUBMISSION_NAME, clone_from=SUBMISSION_URL, use_auth_token=HF_TOKEN, repo_type="dataset")
|
87 |
+
submission_repo.git_pull()
|
88 |
+
df = pd.read_csv(CSV_DIR)
|
89 |
+
df = get_final_score(df)
|
90 |
+
df = df.sort_values(by="Final Score", ascending=False)
|
91 |
+
return df
|
92 |
+
|
93 |
+
def on_filter_model_size_method_change(selected_columns):
|
94 |
+
updated_data = get_all_df()
|
95 |
+
|
96 |
+
# columns:
|
97 |
+
selected_columns = [item for item in TASK_INFO if item in selected_columns]
|
98 |
+
present_columns = MODEL_INFO + selected_columns
|
99 |
+
# print("selected_columns",'|'.join(selected_columns))
|
100 |
+
updated_data = updated_data[present_columns]
|
101 |
+
updated_data = updated_data.sort_values(by=selected_columns[0], ascending=False)
|
102 |
+
updated_headers = present_columns
|
103 |
+
update_datatype = [DATA_TITILE_TYPE[COLUMN_NAMES.index(x)] for x in updated_headers]
|
104 |
+
# print(updated_data,present_columns,update_datatype)
|
105 |
+
filter_component = gr.components.Dataframe(
|
106 |
+
value=updated_data,
|
107 |
+
headers=updated_headers,
|
108 |
+
type="pandas",
|
109 |
+
datatype=update_datatype,
|
110 |
+
interactive=False,
|
111 |
+
visible=True,
|
112 |
+
)
|
113 |
+
|
114 |
+
return filter_component#.value
|
115 |
+
|
116 |
+
block = gr.Blocks()
|
117 |
+
|
118 |
+
|
119 |
+
with block:
|
120 |
+
gr.Markdown(
|
121 |
+
LEADERBORAD_INTRODUCTION
|
122 |
+
)
|
123 |
+
with gr.Tabs(elem_classes="tab-buttons") as tabs:
|
124 |
+
with gr.TabItem("📊 VBench", elem_id="vbench-tab-table", id=1):
|
125 |
+
with gr.Row():
|
126 |
+
with gr.Accordion("Citation", open=False):
|
127 |
+
citation_button = gr.Textbox(
|
128 |
+
value=CITATION_BUTTON_TEXT,
|
129 |
+
label=CITATION_BUTTON_LABEL,
|
130 |
+
elem_id="citation-button",
|
131 |
+
lines=10,
|
132 |
+
)
|
133 |
+
|
134 |
+
gr.Markdown(
|
135 |
+
TABLE_INTRODUCTION
|
136 |
+
)
|
137 |
+
|
138 |
+
# selection for column part:
|
139 |
+
checkbox_group = gr.CheckboxGroup(
|
140 |
+
choices=TASK_INFO,
|
141 |
+
value=AVG_INFO,
|
142 |
+
label="Evaluation Dimension",
|
143 |
+
interactive=True,
|
144 |
+
)
|
145 |
+
|
146 |
+
data_component = gr.components.Dataframe(
|
147 |
+
value=get_baseline_df,
|
148 |
+
headers=COLUMN_NAMES,
|
149 |
+
type="pandas",
|
150 |
+
datatype=DATA_TITILE_TYPE,
|
151 |
+
interactive=False,
|
152 |
+
visible=True,
|
153 |
+
)
|
154 |
+
|
155 |
+
|
156 |
+
checkbox_group.change(fn=on_filter_model_size_method_change, inputs=[ checkbox_group], outputs=data_component)
|
157 |
+
|
158 |
+
# table 2
|
159 |
+
with gr.TabItem("📝 About", elem_id="mvbench-tab-table", id=2):
|
160 |
+
gr.Markdown(LEADERBORAD_INFO, elem_classes="markdown-text")
|
161 |
+
|
162 |
+
# table 3
|
163 |
+
with gr.TabItem("🚀 Submit here! ", elem_id="mvbench-tab-table", id=3):
|
164 |
+
gr.Markdown(LEADERBORAD_INTRODUCTION, elem_classes="markdown-text")
|
165 |
+
|
166 |
+
with gr.Row():
|
167 |
+
gr.Markdown(SUBMIT_INTRODUCTION, elem_classes="markdown-text")
|
168 |
+
|
169 |
+
with gr.Row():
|
170 |
+
gr.Markdown("# ✉️✨ Submit your model evaluation json file here!", elem_classes="markdown-text")
|
171 |
+
|
172 |
+
with gr.Row():
|
173 |
+
with gr.Column():
|
174 |
+
model_name_textbox = gr.Textbox(
|
175 |
+
label="Model name", placeholder="LaVie"
|
176 |
+
)
|
177 |
+
revision_name_textbox = gr.Textbox(
|
178 |
+
label="Revision Model Name", placeholder="LaVie"
|
179 |
+
)
|
180 |
+
|
181 |
+
with gr.Column():
|
182 |
+
model_link = gr.Textbox(
|
183 |
+
label="Model Link", placeholder="https://huggingface.co/decapoda-research/llama-7b-hf"
|
184 |
+
)
|
185 |
+
|
186 |
+
|
187 |
+
with gr.Column():
|
188 |
+
|
189 |
+
input_file = gr.components.File(label = "Click to Upload a json File", file_count="single", type='binary')
|
190 |
+
submit_button = gr.Button("Submit Eval")
|
191 |
+
|
192 |
+
submission_result = gr.Markdown()
|
193 |
+
submit_button.click(
|
194 |
+
add_new_eval,
|
195 |
+
inputs = [
|
196 |
+
input_file,
|
197 |
+
model_name_textbox,
|
198 |
+
revision_name_textbox,
|
199 |
+
model_link,
|
200 |
+
],
|
201 |
+
)
|
202 |
+
|
203 |
+
|
204 |
+
def refresh_data():
|
205 |
+
value1 = get_baseline_df()
|
206 |
+
return value1
|
207 |
+
|
208 |
+
with gr.Row():
|
209 |
+
data_run = gr.Button("Refresh")
|
210 |
+
data_run.click(on_filter_model_size_method_change, inputs=[checkbox_group], outputs=data_component)
|
211 |
+
|
212 |
+
|
213 |
+
block.launch()
|
constants.py
ADDED
@@ -0,0 +1,55 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import os
|
2 |
+
# this is .py for store constants
|
3 |
+
MODEL_INFO = ["name"]
|
4 |
+
TASK_INFO = ["Final Score",
|
5 |
+
"subject consistency",
|
6 |
+
"background consistency",
|
7 |
+
"temporal flickering",
|
8 |
+
"motion smoothness",
|
9 |
+
"dynamic degree",
|
10 |
+
"aesthetic quality",
|
11 |
+
"imaging quality",
|
12 |
+
"object class",
|
13 |
+
"multiple objects",
|
14 |
+
"human action",
|
15 |
+
"color",
|
16 |
+
"spatial relationship",
|
17 |
+
"scene",
|
18 |
+
"appearance style",
|
19 |
+
"temporal style",
|
20 |
+
"overall consistency"]
|
21 |
+
|
22 |
+
AVG_INFO = ["Final Score"]
|
23 |
+
|
24 |
+
DATA_TITILE_TYPE = ['markdown', 'number', 'number', 'number', 'number', 'number', 'number', 'number', 'number', 'number', 'number', 'number', 'number', 'number', 'number', 'number', 'number', 'number', 'number']
|
25 |
+
|
26 |
+
SUBMISSION_NAME = "vbench_leaderboard_submission"
|
27 |
+
SUBMISSION_URL = os.path.join("https://huggingface.co/datasets/Vchitect/", SUBMISSION_NAME)
|
28 |
+
CSV_DIR = "./vbench_leaderboard_submission/results.csv"
|
29 |
+
|
30 |
+
COLUMN_NAMES = MODEL_INFO + TASK_INFO
|
31 |
+
|
32 |
+
LEADERBORAD_INTRODUCTION = """# Vbench Leaderboard
|
33 |
+
|
34 |
+
🏆 Welcome to the leaderboard of the VBench! 🎦
|
35 |
+
|
36 |
+
Please follow the instructions in [Vbench](https://github.com/Vchitect/VBench?tab=readme-ov-file#usage) to upload the generated `result.json` file here. After clicking the `Submit Eval` button, click the `Refresh` button.
|
37 |
+
"""
|
38 |
+
|
39 |
+
SUBMIT_INTRODUCTION = """# Submit on VBench Benchmark Introduction
|
40 |
+
"""
|
41 |
+
|
42 |
+
TABLE_INTRODUCTION = """
|
43 |
+
"""
|
44 |
+
|
45 |
+
LEADERBORAD_INFO = """
|
46 |
+
VBench, a comprehensive benchmark suite for video generative models. We design a comprehensive and hierarchical Evaluation Dimension Suite to decompose "video generation quality" into multiple well-defined dimensions to facilitate fine-grained and objective evaluation. For each dimension and each content category, we carefully design a Prompt Suite as test cases, and sample Generated Videos from a set of video generation models. For each evaluation dimension, we specifically design an Evaluation Method Suite, which uses carefully crafted method or designated pipeline for automatic objective evaluation. We also conduct Human Preference Annotation for the generated videos for each dimension, and show that VBench evaluation results are well aligned with human perceptions. VBench can provide valuable insights from multiple perspectives.
|
47 |
+
"""
|
48 |
+
|
49 |
+
CITATION_BUTTON_LABEL = "Copy the following snippet to cite these results"
|
50 |
+
CITATION_BUTTON_TEXT = r"""@article{huang2023vbench,
|
51 |
+
title={{VBench}: Comprehensive Benchmark Suite for Video Generative Models},
|
52 |
+
author={Huang, Ziqi and He, Yinan and Yu, Jiashuo and Zhang, Fan and Si, Chenyang and Jiang, Yuming and Zhang, Yuanhan and Wu, Tianxing and Jin, Qingyang and Chanpaisit, Nattapol and Wang, Yaohui and Chen, Xinyuan and Wang, Limin and Lin, Dahua and Qiao, Yu and Liu, Ziwei},
|
53 |
+
journal={arXiv preprint arXiv:2311.17982},
|
54 |
+
year={2023}
|
55 |
+
}"""
|
requirements.txt
ADDED
@@ -0,0 +1,2 @@
|
|
|
|
|
|
|
1 |
+
gradio==3.23.0
|
2 |
+
pandas==2.0.0
|