kz-transformers commited on
Commit
ea0ef47
·
verified ·
1 Parent(s): bd24c3f

Update src/leaderboard/build_leaderboard.py

Browse files
Files changed (1) hide show
  1. src/leaderboard/build_leaderboard.py +10 -10
src/leaderboard/build_leaderboard.py CHANGED
@@ -82,17 +82,17 @@ def build_leadearboard_df():
82
  f=json.load(open("genned.json",'r'))
83
  df = pd.DataFrame.from_records(f)
84
  print(f)
85
- if 'mmluproru' in list(df.columns):
86
- df['mmluproru'] = df['mmluproru'].fillna(0)
87
- else:
88
- df['mmluproru'] = 0
89
  print('LEADERBOARD DF: ', df.columns, df.shape)
90
- leaderboard_df = df[['model','kazakhdasturmc','model_dtype','ppl']]
91
- leaderboard_df['avg'] = leaderboard_df[['kazakhdasturmc']].mean(axis=1).values
92
- # print(leaderboard_df.columns)
93
- if len(leaderboard_df)>3:
94
- leaderboard_df = leaderboard_df[leaderboard_df['kazakhdasturmc']!=0]
95
- leaderboard_df = leaderboard_df[leaderboard_df['model']!='apsys/tlite-it-0.1']
 
 
 
96
 
97
  leaderboard_df.sort_values(by='avg',ascending=False,inplace=True,axis=0)
98
  numeric_cols = leaderboard_df.select_dtypes(include=['number']).columns
 
82
  f=json.load(open("genned.json",'r'))
83
  df = pd.DataFrame.from_records(f)
84
  print(f)
85
+
 
 
 
86
  print('LEADERBOARD DF: ', df.columns, df.shape)
87
+ leaderboard_df = df[['model', 'mmlu_translated_kk', 'kk_constitution_mc', 'kk_dastur_mc', 'kazakh_and_literature_unt_mc', 'kk_geography_unt_mc',
88
+ 'kk_world_history_unt_mc', 'kk_history_of_kazakhstan_unt_mc', 'kk_english_unt_mc', 'kk_biology_unt_mc',
89
+ 'kk_human_society_rights_unt_mc', 'model_dtype','ppl']]
90
+
91
+ leaderboard_df['avg'] = leaderboard_df[[
92
+ 'mmlu_translated_kk', 'kk_constitution_mc', 'kk_dastur_mc', 'kazakh_and_literature_unt_mc', 'kk_geography_unt_mc',
93
+ 'kk_world_history_unt_mc', 'kk_history_of_kazakhstan_unt_mc', 'kk_english_unt_mc', 'kk_biology_unt_mc',
94
+ 'kk_human_society_rights_unt_mc']].mean(axis=1).values
95
+
96
 
97
  leaderboard_df.sort_values(by='avg',ascending=False,inplace=True,axis=0)
98
  numeric_cols = leaderboard_df.select_dtypes(include=['number']).columns