refactor: refactor the envs

app.py

@@ -8,15 +8,13 @@ from src.about import (
 from src.benchmarks import (
     qa_benchmark_dict,
     long_doc_benchmark_dict,
-    METRIC_LIST,
-    DEFAULT_METRIC_QA,
-    DEFAULT_METRIC_LONG_DOC
+    METRIC_LIST
 )
 from src.display.css_html_js import custom_css
 from src.envs import (
     API,
     EVAL_RESULTS_PATH,
-    REPO_ID
+    REPO_ID, DEFAULT_METRIC_QA, DEFAULT_METRIC_LONG_DOC
 )
 from src.loaders import (
     load_eval_results

src/benchmarks.py

@@ -3,9 +3,6 @@ from enum import Enum
 
 from air_benchmark.tasks.tasks import BenchmarkTable
 
-DEFAULT_METRIC_QA = "ndcg_at_10"
-DEFAULT_METRIC_LONG_DOC = "recall_at_10"
-
 METRIC_LIST = [
     "ndcg_at_1",
     "ndcg_at_3",

src/envs.py

@@ -31,3 +31,5 @@ BENCHMARK_VERSION_LIST = [
 ]
 
 LATEST_BENCHMARK_VERSION = BENCHMARK_VERSION_LIST[-1]
+DEFAULT_METRIC_QA = "ndcg_at_10"
+DEFAULT_METRIC_LONG_DOC = "recall_at_10"

src/loaders.py

@@ -3,7 +3,7 @@ from typing import List
 
 import pandas as pd
 
-from src.benchmarks import DEFAULT_METRIC_QA, DEFAULT_METRIC_LONG_DOC
+from src.envs import DEFAULT_METRIC_QA, DEFAULT_METRIC_LONG_DOC
 from src.display.columns import COL_NAME_REVISION, COL_NAME_TIMESTAMP, COL_NAME_IS_ANONYMOUS
 
 from src.models import FullEvalResult, LeaderboardDataStore