Bagratuni commited on
Commit
1b75b9d
·
1 Parent(s): 2029a14
Files changed (1) hide show
  1. app.py +19 -19
app.py CHANGED
@@ -9,6 +9,8 @@ def display_table(exam_type):
9
  cols = df.columns.tolist()
10
  cols.insert(1, cols.pop(cols.index('Average')))
11
  df = df[cols]
 
 
12
  elif exam_type == "MMLU-Pro-Hy":
13
  df = pd.read_csv('mmlu_pro_hy_results.csv')
14
  subject_cols = ['Biology', 'Business', 'Chemistry', 'Computer Science', 'Economics', 'Engineering', 'Health', 'History', 'Law', 'Math', 'Other', 'Philosophy', 'Physics', 'Psychology']
@@ -19,13 +21,13 @@ def display_table(exam_type):
19
  cols.insert(1, cols.pop(cols.index('Average')))
20
  cols.append(cols.pop(cols.index('Other')))
21
  df = df[cols]
 
22
  return df
23
 
24
  def create_bar_chart(exam_type, plot_column):
25
  if exam_type == "Armenian Exams":
26
  df = pd.read_csv('unified_exam_results.csv')
27
  df = df.sort_values(by=[plot_column, 'Model'], ascending=[False, True]).reset_index(drop=True)
28
-
29
  x_col = plot_column
30
  title = f'{plot_column}'
31
  x_range_max = 20
@@ -37,7 +39,6 @@ def create_bar_chart(exam_type, plot_column):
37
  else:
38
  return "Distinction"
39
  df['Test Result'] = df[plot_column].apply(get_label)
40
-
41
  color_discrete_map = {
42
  "Fail": "#ff5f56",
43
  "Pass": "#ffbd2e",
@@ -51,27 +52,24 @@ def create_bar_chart(exam_type, plot_column):
51
  labels={x_col: 'Score', 'Model': 'Model'},
52
  title=title,
53
  orientation='h')
54
-
55
  fig.update_layout(
56
  xaxis=dict(range=[0, x_range_max]),
57
  title=dict(text=title, font=dict(size=16)),
58
  xaxis_title=dict(font=dict(size=12)),
59
  yaxis_title=dict(font=dict(size=12)),
60
- yaxis=dict(autorange="reversed")
 
61
  )
62
-
63
  return fig
64
-
65
  elif exam_type == "MMLU-Pro-Hy":
66
  df = pd.read_csv('mmlu_pro_hy_results.csv')
67
  subject_cols = ['Biology', 'Business', 'Chemistry', 'Computer Science', 'Economics', 'Engineering', 'Health', 'History', 'Law', 'Math', 'Other', 'Philosophy', 'Physics', 'Psychology']
68
  df['Average'] = df[subject_cols].mean(axis=1)
69
- df = df.sort_values(by='Average', ascending=False)
70
  df = df.drop(columns=['Accuracy'])
71
  x_col = plot_column
72
  title = f'{plot_column}'
73
  x_range_max = 1.0
74
-
75
  fig = px.bar(df,
76
  x=x_col,
77
  y='Model',
@@ -81,35 +79,37 @@ def create_bar_chart(exam_type, plot_column):
81
  title=title,
82
  orientation='h',
83
  range_color=[0,1])
84
-
85
  fig.update_layout(
86
  xaxis=dict(range=[0, x_range_max]),
87
  title=dict(text=title, font=dict(size=16)),
88
  xaxis_title=dict(font=dict(size=12)),
89
  yaxis_title=dict(font=dict(size=12)),
90
- yaxis=dict(autorange="reversed")
 
91
  )
92
-
93
  return fig
94
 
95
  with gr.Blocks() as app:
96
  with gr.Tabs():
97
  with gr.TabItem("Armenian Unified Exams"):
98
  gr.Markdown("# Armenian Unified Test Exams")
99
- gr.Markdown("### This benchmark contains results of various Language Models on Armenian Unified Test Exams for Armenian language and literature, Armenian history and mathematics. The scoring system is a 20-point scale, where 0-8 is a Fail, 8-18 is a Pass, and 18-20 is a Distinction.")
100
- # gr.Markdown("### Այս աղյուսակը պարունակում է տարբեր լեզվական մոդելների արդյունքները հայերեն լեզվի և գրականության, հայոց պատմության և մաթեմատիկայի միասնական քնությունների թեսթերի համար։ Գնահատման համակարգը 20 բալանոց սանդղակ է, որտեղ 0-8-ը նշանակում է Անբավարար, 8-18-ը՝ Բավարար, իսկ 18-20-ը՝ Գերազանց:")
101
-
 
 
102
  table_output_armenian = gr.DataFrame(value=lambda: display_table("Armenian Exams"))
103
  plot_column_dropdown = gr.Dropdown(choices=['Average', 'Armenian language and literature', 'Armenian history', 'Mathematics'], value='Average', label='Select Column to Plot')
104
  plot_output_armenian = gr.Plot(lambda column: create_bar_chart("Armenian Exams", column), inputs=plot_column_dropdown)
105
  with gr.TabItem("MMLU-Pro-Hy"):
106
  gr.Markdown("# MMLU-Pro Translated to Armenian (MMLU-Pro-Hy)")
107
- gr.Markdown("### This benchmark contains results of various Language Models on the MMLU-Pro benchmark, translated into Armenian. MMLU-Pro is a massive multi-task test in MCQA format. The scores represent accuracy.")
108
- # gr.Markdown("### Այս աղյուսակը պարունակում է տարբեր լեզվական մոդելների արդյունքները MMLU-Pro թեսթի համար, որը թարգմանվել է հայերեն: MMLU-Pro-ն իրենից ներկայացնում է : Միավորները ներկայացնում են ճշգրտությունը:")
109
-
 
 
110
  table_output_mmlu = gr.DataFrame(value=lambda: display_table("MMLU-Pro-Hy"))
111
- subject_cols = ['Biology', 'Business', 'Chemistry', 'Computer Science', 'Economics', 'Engineering', 'Health', 'History', 'Law', 'Math', 'Other', 'Philosophy', 'Physics', 'Psychology', 'Average']
112
  plot_column_dropdown_mmlu = gr.Dropdown(choices=subject_cols, value='Average', label='Select Column to Plot')
113
  plot_output_mmlu = gr.Plot(lambda column: create_bar_chart("MMLU-Pro-Hy", column), inputs=plot_column_dropdown_mmlu)
114
-
115
  app.launch(share=True, debug=True)
 
9
  cols = df.columns.tolist()
10
  cols.insert(1, cols.pop(cols.index('Average')))
11
  df = df[cols]
12
+ df.rename(columns={'Armenian language and literature': 'Armenian language\nand literature'}, inplace=True)
13
+ df = df.round(4)
14
  elif exam_type == "MMLU-Pro-Hy":
15
  df = pd.read_csv('mmlu_pro_hy_results.csv')
16
  subject_cols = ['Biology', 'Business', 'Chemistry', 'Computer Science', 'Economics', 'Engineering', 'Health', 'History', 'Law', 'Math', 'Other', 'Philosophy', 'Physics', 'Psychology']
 
21
  cols.insert(1, cols.pop(cols.index('Average')))
22
  cols.append(cols.pop(cols.index('Other')))
23
  df = df[cols]
24
+ df = df.round(4)
25
  return df
26
 
27
  def create_bar_chart(exam_type, plot_column):
28
  if exam_type == "Armenian Exams":
29
  df = pd.read_csv('unified_exam_results.csv')
30
  df = df.sort_values(by=[plot_column, 'Model'], ascending=[False, True]).reset_index(drop=True)
 
31
  x_col = plot_column
32
  title = f'{plot_column}'
33
  x_range_max = 20
 
39
  else:
40
  return "Distinction"
41
  df['Test Result'] = df[plot_column].apply(get_label)
 
42
  color_discrete_map = {
43
  "Fail": "#ff5f56",
44
  "Pass": "#ffbd2e",
 
52
  labels={x_col: 'Score', 'Model': 'Model'},
53
  title=title,
54
  orientation='h')
 
55
  fig.update_layout(
56
  xaxis=dict(range=[0, x_range_max]),
57
  title=dict(text=title, font=dict(size=16)),
58
  xaxis_title=dict(font=dict(size=12)),
59
  yaxis_title=dict(font=dict(size=12)),
60
+ yaxis=dict(autorange="reversed"),
61
+ autosize=True
62
  )
 
63
  return fig
 
64
  elif exam_type == "MMLU-Pro-Hy":
65
  df = pd.read_csv('mmlu_pro_hy_results.csv')
66
  subject_cols = ['Biology', 'Business', 'Chemistry', 'Computer Science', 'Economics', 'Engineering', 'Health', 'History', 'Law', 'Math', 'Other', 'Philosophy', 'Physics', 'Psychology']
67
  df['Average'] = df[subject_cols].mean(axis=1)
68
+ df = df.sort_values(by=plot_column, ascending=False).reset_index(drop=True)
69
  df = df.drop(columns=['Accuracy'])
70
  x_col = plot_column
71
  title = f'{plot_column}'
72
  x_range_max = 1.0
 
73
  fig = px.bar(df,
74
  x=x_col,
75
  y='Model',
 
79
  title=title,
80
  orientation='h',
81
  range_color=[0,1])
 
82
  fig.update_layout(
83
  xaxis=dict(range=[0, x_range_max]),
84
  title=dict(text=title, font=dict(size=16)),
85
  xaxis_title=dict(font=dict(size=12)),
86
  yaxis_title=dict(font=dict(size=12)),
87
+ yaxis=dict(autorange="reversed"),
88
+ autosize=True
89
  )
 
90
  return fig
91
 
92
  with gr.Blocks() as app:
93
  with gr.Tabs():
94
  with gr.TabItem("Armenian Unified Exams"):
95
  gr.Markdown("# Armenian Unified Test Exams")
96
+ gr.HTML(f"""
97
+ <div style="font-size: 16px;">
98
+ This benchmark contains results of various Language Models on Armenian Unified Test Exams for Armenian language and literature, Armenian history and mathematics. The scoring system is a 20-point scale, where 0-8 is a Fail, 8-18 is a Pass, and 18-20 is a Distinction.
99
+ </div>
100
+ """)
101
  table_output_armenian = gr.DataFrame(value=lambda: display_table("Armenian Exams"))
102
  plot_column_dropdown = gr.Dropdown(choices=['Average', 'Armenian language and literature', 'Armenian history', 'Mathematics'], value='Average', label='Select Column to Plot')
103
  plot_output_armenian = gr.Plot(lambda column: create_bar_chart("Armenian Exams", column), inputs=plot_column_dropdown)
104
  with gr.TabItem("MMLU-Pro-Hy"):
105
  gr.Markdown("# MMLU-Pro Translated to Armenian (MMLU-Pro-Hy)")
106
+ gr.HTML(f"""
107
+ <div style="font-size: 16px;">
108
+ This benchmark contains results of various Language Models on the MMLU-Pro benchmark, translated into Armenian. MMLU-Pro is a massive multi-task test in MCQA format. The scores represent accuracy.
109
+ </div>
110
+ """)
111
  table_output_mmlu = gr.DataFrame(value=lambda: display_table("MMLU-Pro-Hy"))
112
+ subject_cols = ['Average','Biology', 'Business', 'Chemistry', 'Computer Science', 'Economics', 'Engineering', 'Health', 'History', 'Law', 'Math', 'Philosophy', 'Physics', 'Psychology','Other']
113
  plot_column_dropdown_mmlu = gr.Dropdown(choices=subject_cols, value='Average', label='Select Column to Plot')
114
  plot_output_mmlu = gr.Plot(lambda column: create_bar_chart("MMLU-Pro-Hy", column), inputs=plot_column_dropdown_mmlu)
 
115
  app.launch(share=True, debug=True)