Spaces:
Restarting
Restarting
kz-transformers
commited on
Update src/leaderboard/build_leaderboard.py
Browse files
src/leaderboard/build_leaderboard.py
CHANGED
@@ -64,10 +64,10 @@ def download_dataset(repo_id, local_dir, repo_type="dataset", max_attempts=3, ba
|
|
64 |
|
65 |
def download_openbench():
|
66 |
# download prev autogenerated leaderboard files
|
67 |
-
download_dataset("
|
68 |
|
69 |
# download answers of different models that we trust
|
70 |
-
download_dataset("
|
71 |
|
72 |
|
73 |
def build_leadearboard_df():
|
@@ -84,11 +84,11 @@ def build_leadearboard_df():
|
|
84 |
df['mmluproru'] = df['mmluproru'].fillna(0)
|
85 |
else:
|
86 |
df['mmluproru'] = 0
|
87 |
-
leaderboard_df = df[['model','
|
88 |
-
leaderboard_df['avg'] = leaderboard_df[['
|
89 |
# print(leaderboard_df.columns)
|
90 |
if len(leaderboard_df)>3:
|
91 |
-
leaderboard_df = leaderboard_df[leaderboard_df['
|
92 |
leaderboard_df = leaderboard_df[leaderboard_df['model']!='apsys/tlite-it-0.1']
|
93 |
|
94 |
leaderboard_df.sort_values(by='avg',ascending=False,inplace=True,axis=0)
|
|
|
64 |
|
65 |
def download_openbench():
|
66 |
# download prev autogenerated leaderboard files
|
67 |
+
download_dataset("kz-transformers/s-shlepa-metainfo", DATA_PATH)
|
68 |
|
69 |
# download answers of different models that we trust
|
70 |
+
download_dataset("kz-transformers/s-openbench-eval", "m_data")
|
71 |
|
72 |
|
73 |
def build_leadearboard_df():
|
|
|
84 |
df['mmluproru'] = df['mmluproru'].fillna(0)
|
85 |
else:
|
86 |
df['mmluproru'] = 0
|
87 |
+
leaderboard_df = df[['model','kazakhdasturmc','model_dtype','ppl']]
|
88 |
+
leaderboard_df['avg'] = leaderboard_df[['kazakhdasturmc']].mean(axis=1).values
|
89 |
# print(leaderboard_df.columns)
|
90 |
if len(leaderboard_df)>3:
|
91 |
+
leaderboard_df = leaderboard_df[leaderboard_df['kazakhdasturmc']!=0]
|
92 |
leaderboard_df = leaderboard_df[leaderboard_df['model']!='apsys/tlite-it-0.1']
|
93 |
|
94 |
leaderboard_df.sort_values(by='avg',ascending=False,inplace=True,axis=0)
|