update leaderboard
Browse files
app.py
CHANGED
@@ -16,6 +16,107 @@ head_style = """
|
|
16 |
</style>
|
17 |
"""
|
18 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
19 |
with gr.Blocks(title="Math Leaderboard", head=head_style) as demo:
|
20 |
results = load_results()['results']
|
21 |
N_MODEL = len(results)
|
@@ -32,84 +133,8 @@ with gr.Blocks(title="Math Leaderboard", head=head_style) as demo:
|
|
32 |
|
33 |
with gr.Tabs(elem_classes='tab-buttons') as tabs:
|
34 |
with gr.TabItem('🏅 LMM Math Leaderboard', elem_id='main', id=0):
|
35 |
-
|
36 |
-
|
37 |
-
table = generate_table(results)
|
38 |
-
table['Rank'] = list(range(1, len(table) + 1))
|
39 |
-
|
40 |
-
type_map = check_box['type_map']
|
41 |
-
type_map['Rank'] = 'number'
|
42 |
-
|
43 |
-
checkbox_group = gr.CheckboxGroup(
|
44 |
-
choices=check_box['all'],
|
45 |
-
value=check_box['required'],
|
46 |
-
label='Evaluation Dimension',
|
47 |
-
interactive=True,
|
48 |
-
)
|
49 |
-
|
50 |
-
headers = ['Rank'] + check_box['essential'] + checkbox_group.value
|
51 |
-
with gr.Row():
|
52 |
-
model_name = gr.Textbox(
|
53 |
-
value='Input the Model Name (fuzzy)',
|
54 |
-
label='Model Name',
|
55 |
-
interactive=True,
|
56 |
-
visible=True)
|
57 |
-
model_size = gr.CheckboxGroup(
|
58 |
-
choices=MODEL_SIZE,
|
59 |
-
value=MODEL_SIZE,
|
60 |
-
label='Model Size',
|
61 |
-
interactive=True
|
62 |
-
)
|
63 |
-
model_type = gr.CheckboxGroup(
|
64 |
-
choices=MODEL_TYPE,
|
65 |
-
value=MODEL_TYPE,
|
66 |
-
label='Model Type',
|
67 |
-
interactive=True
|
68 |
-
)
|
69 |
-
|
70 |
-
data_component = gr.components.DataFrame(
|
71 |
-
value=table[headers],
|
72 |
-
type='pandas',
|
73 |
-
datatype=[type_map[x] for x in headers],
|
74 |
-
interactive=False,
|
75 |
-
visible=True)
|
76 |
-
|
77 |
-
def filter_df(fields, model_name, model_size, model_type):
|
78 |
-
results = load_results()['results']
|
79 |
-
headers = ['Rank'] + check_box['essential'] + fields
|
80 |
-
|
81 |
-
df = generate_table(results)
|
82 |
-
|
83 |
-
df['flag'] = [model_size_flag(x, model_size) for x in df['Param (B)']]
|
84 |
-
df = df[df['flag']]
|
85 |
-
df.pop('flag')
|
86 |
-
if len(df):
|
87 |
-
df['flag'] = [model_type_flag(df.iloc[i], model_type) for i in range(len(df))]
|
88 |
-
df = df[df['flag']]
|
89 |
-
df.pop('flag')
|
90 |
-
df['Rank'] = list(range(1, len(df) + 1))
|
91 |
-
default_val = 'Input the Model Name (fuzzy)'
|
92 |
-
if model_name != default_val:
|
93 |
-
print(model_name)
|
94 |
-
model_name = model_name.lower()
|
95 |
-
method_names = [x.split('</a>')[0].split('>')[-1].lower() for x in df['Method']]
|
96 |
-
flag = [model_name in name for name in method_names]
|
97 |
-
df['TEMP_FLAG'] = flag
|
98 |
-
df = df[df['TEMP_FLAG'] == True]
|
99 |
-
df.pop('TEMP_FLAG')
|
100 |
-
|
101 |
-
comp = gr.components.DataFrame(
|
102 |
-
value=df[headers],
|
103 |
-
type='pandas',
|
104 |
-
datatype=[type_map[x] for x in headers],
|
105 |
-
interactive=False,
|
106 |
-
visible=True)
|
107 |
-
return comp
|
108 |
-
|
109 |
-
for cbox in [checkbox_group, model_size, model_type]:
|
110 |
-
cbox.change(fn=filter_df, inputs=[checkbox_group, model_name, model_size, model_type], outputs=data_component)
|
111 |
-
model_name.submit(fn=filter_df, inputs=[checkbox_group, model_name, model_size, model_type], outputs=data_component)
|
112 |
-
|
113 |
for i, dataset in enumerate(DATASETS):
|
114 |
tab_name_map = {
|
115 |
'MathVista': 'MathVista (Test Mini)',
|
@@ -118,87 +143,7 @@ with gr.Blocks(title="Math Leaderboard", head=head_style) as demo:
|
|
118 |
|
119 |
with gr.TabItem(
|
120 |
f'📊 {dataset if dataset not in tab_name_map else tab_name_map[dataset]}', elem_id=dataset, id=i + 2):
|
121 |
-
|
122 |
-
s = structs[i]
|
123 |
-
s.table, s.check_box = BUILD_L2_DF(results, dataset)
|
124 |
-
s.type_map = s.check_box['type_map']
|
125 |
-
s.type_map['Rank'] = 'number'
|
126 |
-
|
127 |
-
s.checkbox_group = gr.CheckboxGroup(
|
128 |
-
choices=s.check_box['all'],
|
129 |
-
value=s.check_box['required'],
|
130 |
-
label=f'{dataset} CheckBoxes',
|
131 |
-
interactive=True,
|
132 |
-
)
|
133 |
-
s.headers = ['Rank'] + s.check_box['essential'] + s.checkbox_group.value
|
134 |
-
s.table['Rank'] = list(range(1, len(s.table) + 1))
|
135 |
-
|
136 |
-
with gr.Row():
|
137 |
-
s.model_name = gr.Textbox(
|
138 |
-
value='Input the Model Name (fuzzy)',
|
139 |
-
label='Model Name',
|
140 |
-
interactive=True,
|
141 |
-
visible=True)
|
142 |
-
s.model_size = gr.CheckboxGroup(
|
143 |
-
choices=MODEL_SIZE,
|
144 |
-
value=MODEL_SIZE,
|
145 |
-
label='Model Size',
|
146 |
-
interactive=True
|
147 |
-
)
|
148 |
-
s.model_type = gr.CheckboxGroup(
|
149 |
-
choices=MODEL_TYPE,
|
150 |
-
value=MODEL_TYPE,
|
151 |
-
label='Model Type',
|
152 |
-
interactive=True
|
153 |
-
)
|
154 |
-
s.data_component = gr.components.DataFrame(
|
155 |
-
value=s.table[s.headers],
|
156 |
-
type='pandas',
|
157 |
-
datatype=[s.type_map[x] for x in s.headers],
|
158 |
-
interactive=False,
|
159 |
-
visible=True)
|
160 |
-
s.dataset = gr.Textbox(value=dataset, label=dataset, visible=False)
|
161 |
-
|
162 |
-
def filter_df_l2(dataset_name, fields, model_name, model_size, model_type):
|
163 |
-
results = load_results()['results']
|
164 |
-
s = structs[DATASETS.index(dataset_name)]
|
165 |
-
headers = ['Rank'] + s.check_box['essential'] + fields
|
166 |
-
df = cp.deepcopy(s.table)
|
167 |
-
df['flag'] = [model_size_flag(x, model_size) for x in df['Param (B)']]
|
168 |
-
df = df[df['flag']]
|
169 |
-
df.pop('flag')
|
170 |
-
if len(df):
|
171 |
-
df['flag'] = [model_type_flag(df.iloc[i], model_type) for i in range(len(df))]
|
172 |
-
df = df[df['flag']]
|
173 |
-
df.pop('flag')
|
174 |
-
df['Rank'] = list(range(1, len(df) + 1))
|
175 |
-
default_val = 'Input the Model Name (fuzzy)'
|
176 |
-
if model_name != default_val:
|
177 |
-
print(model_name)
|
178 |
-
model_name = model_name.lower()
|
179 |
-
method_names = [x.split('</a>')[0].split('>')[-1].lower() for x in df['Method']]
|
180 |
-
flag = [model_name in name for name in method_names]
|
181 |
-
df['TEMP_FLAG'] = flag
|
182 |
-
df = df[df['TEMP_FLAG'] == True]
|
183 |
-
df.pop('TEMP_FLAG')
|
184 |
-
|
185 |
-
comp = gr.components.DataFrame(
|
186 |
-
value=df[headers],
|
187 |
-
type='pandas',
|
188 |
-
datatype=[s.type_map[x] for x in headers],
|
189 |
-
interactive=False,
|
190 |
-
visible=True)
|
191 |
-
return comp
|
192 |
-
|
193 |
-
for cbox in [s.checkbox_group, s.model_size, s.model_type]:
|
194 |
-
cbox.change(
|
195 |
-
fn=filter_df_l2,
|
196 |
-
inputs=[s.dataset, s.checkbox_group, s.model_name, s.model_size, s.model_type],
|
197 |
-
outputs=s.data_component)
|
198 |
-
s.model_name.submit(
|
199 |
-
fn=filter_df_l2,
|
200 |
-
inputs=[s.dataset, s.checkbox_group, s.model_name, s.model_size, s.model_type],
|
201 |
-
outputs=s.data_component)
|
202 |
|
203 |
with gr.Row():
|
204 |
with gr.Accordion('Citation', open=False):
|
|
|
16 |
</style>
|
17 |
"""
|
18 |
|
19 |
+
|
20 |
+
def math_main_tab(results):
|
21 |
+
_, check_box = BUILD_L1_DF(results)
|
22 |
+
table = generate_table(results)
|
23 |
+
table['Rank'] = list(range(1, len(table) + 1))
|
24 |
+
type_map = check_box['type_map']
|
25 |
+
type_map['Rank'] = 'number'
|
26 |
+
|
27 |
+
checkbox_group = gr.CheckboxGroup(choices=check_box['all'], value=check_box['required'], label='Evaluation Dimension')
|
28 |
+
|
29 |
+
headers = ['Rank'] + check_box['essential'] + checkbox_group.value
|
30 |
+
with gr.Row():
|
31 |
+
model_name = gr.Textbox(value='Input the Model Name (fuzzy)', label='Model Name')
|
32 |
+
model_size = gr.CheckboxGroup(choices=MODEL_SIZE, value=MODEL_SIZE, label='Model Size')
|
33 |
+
model_type = gr.CheckboxGroup(choices=MODEL_TYPE, value=MODEL_TYPE, label='Model Type')
|
34 |
+
|
35 |
+
data_component = gr.components.DataFrame(value=table[headers], datatype=[type_map[x] for x in headers])
|
36 |
+
|
37 |
+
def filter_df(fields, model_name, model_size, model_type):
|
38 |
+
results = load_results()['results']
|
39 |
+
headers = ['Rank'] + check_box['essential'] + fields
|
40 |
+
|
41 |
+
df = generate_table(results)
|
42 |
+
|
43 |
+
df['flag'] = [model_size_flag(x, model_size) for x in df['Param (B)']]
|
44 |
+
df = df[df['flag']]
|
45 |
+
df.pop('flag')
|
46 |
+
if len(df):
|
47 |
+
df['flag'] = [model_type_flag(df.iloc[i], model_type) for i in range(len(df))]
|
48 |
+
df = df[df['flag']]
|
49 |
+
df.pop('flag')
|
50 |
+
df['Rank'] = list(range(1, len(df) + 1))
|
51 |
+
default_val = 'Input the Model Name (fuzzy)'
|
52 |
+
if model_name != default_val:
|
53 |
+
method_names = [x.split('</a>')[0].split('>')[-1].lower() for x in df['Method']]
|
54 |
+
flag = [model_name.lower() in name for name in method_names]
|
55 |
+
df['TEMP'] = flag
|
56 |
+
df = df[df['TEMP'] == True]
|
57 |
+
df.pop('TEMP')
|
58 |
+
|
59 |
+
comp = gr.components.DataFrame(value=df[headers], datatype=[type_map[x] for x in headers])
|
60 |
+
return comp
|
61 |
+
|
62 |
+
for cbox in [checkbox_group, model_size, model_type]:
|
63 |
+
cbox.change(fn=filter_df, inputs=[checkbox_group, model_name, model_size, model_type], outputs=data_component)
|
64 |
+
model_name.submit(fn=filter_df, inputs=[checkbox_group, model_name, model_size, model_type], outputs=data_component)
|
65 |
+
|
66 |
+
|
67 |
+
def dataset_tab(results, struct, dataset):
|
68 |
+
s = struct
|
69 |
+
s.table, s.check_box = BUILD_L2_DF(results, dataset)
|
70 |
+
s.type_map = s.check_box['type_map']
|
71 |
+
s.type_map['Rank'] = 'number'
|
72 |
+
|
73 |
+
s.checkbox_group = gr.CheckboxGroup(choices=s.check_box['all'], value=s.check_box['required'], label=f'{dataset} CheckBoxes')
|
74 |
+
s.headers = ['Rank'] + s.check_box['essential'] + s.checkbox_group.value
|
75 |
+
s.table['Rank'] = list(range(1, len(s.table) + 1))
|
76 |
+
|
77 |
+
with gr.Row():
|
78 |
+
s.model_name = gr.Textbox(value='Input the Model Name (fuzzy)', label='Model Name')
|
79 |
+
s.model_size = gr.CheckboxGroup(choices=MODEL_SIZE, value=MODEL_SIZE, label='Model Size')
|
80 |
+
s.model_type = gr.CheckboxGroup(choices=MODEL_TYPE, value=MODEL_TYPE, label='Model Type')
|
81 |
+
|
82 |
+
s.data_component = gr.components.DataFrame(value=s.table[s.headers], datatype=[s.type_map[x] for x in s.headers])
|
83 |
+
s.dataset = gr.Textbox(value=dataset, label=dataset, visible=False)
|
84 |
+
|
85 |
+
def filter_df_l2(dataset_name, fields, model_name, model_size, model_type):
|
86 |
+
results = load_results()['results']
|
87 |
+
s = structs[DATASETS.index(dataset_name)]
|
88 |
+
headers = ['Rank'] + s.check_box['essential'] + fields
|
89 |
+
df = cp.deepcopy(s.table)
|
90 |
+
df['flag'] = [model_size_flag(x, model_size) for x in df['Param (B)']]
|
91 |
+
df = df[df['flag']]
|
92 |
+
df.pop('flag')
|
93 |
+
if len(df):
|
94 |
+
df['flag'] = [model_type_flag(df.iloc[i], model_type) for i in range(len(df))]
|
95 |
+
df = df[df['flag']]
|
96 |
+
df.pop('flag')
|
97 |
+
df['Rank'] = list(range(1, len(df) + 1))
|
98 |
+
default_val = 'Input the Model Name (fuzzy)'
|
99 |
+
if model_name != default_val:
|
100 |
+
method_names = [x.split('</a>')[0].split('>')[-1].lower() for x in df['Method']]
|
101 |
+
flag = [model_name.lower() in name for name in method_names]
|
102 |
+
df['TEMP'] = flag
|
103 |
+
df = df[df['TEMP'] == True]
|
104 |
+
df.pop('TEMP')
|
105 |
+
|
106 |
+
comp = gr.components.DataFrame(value=df[headers], datatype=[s.type_map[x] for x in headers])
|
107 |
+
return comp
|
108 |
+
|
109 |
+
for cbox in [s.checkbox_group, s.model_size, s.model_type]:
|
110 |
+
cbox.change(
|
111 |
+
fn=filter_df_l2,
|
112 |
+
inputs=[s.dataset, s.checkbox_group, s.model_name, s.model_size, s.model_type],
|
113 |
+
outputs=s.data_component)
|
114 |
+
s.model_name.submit(
|
115 |
+
fn=filter_df_l2,
|
116 |
+
inputs=[s.dataset, s.checkbox_group, s.model_name, s.model_size, s.model_type],
|
117 |
+
outputs=s.data_component)
|
118 |
+
|
119 |
+
|
120 |
with gr.Blocks(title="Math Leaderboard", head=head_style) as demo:
|
121 |
results = load_results()['results']
|
122 |
N_MODEL = len(results)
|
|
|
133 |
|
134 |
with gr.Tabs(elem_classes='tab-buttons') as tabs:
|
135 |
with gr.TabItem('🏅 LMM Math Leaderboard', elem_id='main', id=0):
|
136 |
+
math_main_tab(results)
|
137 |
+
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
138 |
for i, dataset in enumerate(DATASETS):
|
139 |
tab_name_map = {
|
140 |
'MathVista': 'MathVista (Test Mini)',
|
|
|
143 |
|
144 |
with gr.TabItem(
|
145 |
f'📊 {dataset if dataset not in tab_name_map else tab_name_map[dataset]}', elem_id=dataset, id=i + 2):
|
146 |
+
dataset_tab(results, structs[i], dataset)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
147 |
|
148 |
with gr.Row():
|
149 |
with gr.Accordion('Citation', open=False):
|