Spaces:
Running
on
CPU Upgrade
Running
on
CPU Upgrade
Apply formatters to src/populate.py
Browse files- src/populate.py +15 -12
src/populate.py
CHANGED
|
@@ -13,29 +13,32 @@ def get_leaderboard_df(results_path: str, requests_path: str, cols: list, benchm
|
|
| 13 |
raw_data = get_raw_eval_results(results_path, requests_path)
|
| 14 |
all_data_json = [v.to_dict() for v in raw_data]
|
| 15 |
|
| 16 |
-
|
| 17 |
-
|
| 18 |
df = pd.DataFrame.from_records(all_data_json)
|
| 19 |
-
|
| 20 |
score_cols = [
|
| 21 |
-
|
| 22 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 23 |
]
|
| 24 |
-
|
| 25 |
existing_score_cols = [col for col in score_cols if col in df.columns]
|
| 26 |
-
|
| 27 |
# スコア列を100で割り、.4f形式でフォーマット
|
| 28 |
-
df[existing_score_cols] = (df[existing_score_cols] / 100).applymap(lambda x: f
|
| 29 |
df = df.sort_values(by=[AutoEvalColumn.AVG.name], ascending=False)
|
| 30 |
df = df[cols].round(decimals=2)
|
| 31 |
-
|
| 32 |
# filter out if any of the benchmarks have not been produced
|
| 33 |
df = df[has_no_nan_values(df, benchmark_cols)]
|
| 34 |
|
| 35 |
-
df[
|
| 36 |
-
|
| 37 |
-
return df
|
| 38 |
|
|
|
|
| 39 |
|
| 40 |
|
| 41 |
def get_evaluation_queue_df(save_path: str, cols: list) -> list[pd.DataFrame]:
|
|
|
|
| 13 |
raw_data = get_raw_eval_results(results_path, requests_path)
|
| 14 |
all_data_json = [v.to_dict() for v in raw_data]
|
| 15 |
|
|
|
|
|
|
|
| 16 |
df = pd.DataFrame.from_records(all_data_json)
|
| 17 |
+
|
| 18 |
score_cols = [
|
| 19 |
+
"ALT E to J BLEU",
|
| 20 |
+
"ALT J to E BLEU",
|
| 21 |
+
"WikiCorpus E to J BLEU",
|
| 22 |
+
"WikiCorpus J to E BLEU",
|
| 23 |
+
"XL-Sum JA BLEU",
|
| 24 |
+
"XL-Sum ROUGE1",
|
| 25 |
+
"XL-Sum ROUGE2",
|
| 26 |
+
"XL-Sum ROUGE-Lsum",
|
| 27 |
]
|
| 28 |
+
|
| 29 |
existing_score_cols = [col for col in score_cols if col in df.columns]
|
| 30 |
+
|
| 31 |
# スコア列を100で割り、.4f形式でフォーマット
|
| 32 |
+
df[existing_score_cols] = (df[existing_score_cols] / 100).applymap(lambda x: f"{x:.4f}")
|
| 33 |
df = df.sort_values(by=[AutoEvalColumn.AVG.name], ascending=False)
|
| 34 |
df = df[cols].round(decimals=2)
|
| 35 |
+
|
| 36 |
# filter out if any of the benchmarks have not been produced
|
| 37 |
df = df[has_no_nan_values(df, benchmark_cols)]
|
| 38 |
|
| 39 |
+
df["Model"] = df["Model"].apply(lambda x: f'[{x.split("/")[-1]}]({x})' if isinstance(x, str) else x)
|
|
|
|
|
|
|
| 40 |
|
| 41 |
+
return df
|
| 42 |
|
| 43 |
|
| 44 |
def get_evaluation_queue_df(save_path: str, cols: list) -> list[pd.DataFrame]:
|