Spaces:
Running
Running
Update description
Browse files- pages/fev_bench.py +16 -2
pages/fev_bench.py
CHANGED
|
@@ -11,7 +11,6 @@ from streamlit.elements.lib.column_types import ColumnConfig
|
|
| 11 |
from src.strings import (
|
| 12 |
CITATION_FEV,
|
| 13 |
CITATION_HEADER,
|
| 14 |
-
FEV_BENCHMARK_BASIC_INFO,
|
| 15 |
FEV_BENCHMARK_DETAILS,
|
| 16 |
PAIRWISE_BENCHMARK_DETAILS,
|
| 17 |
get_pivot_legend,
|
|
@@ -42,6 +41,21 @@ GROUP_TYPES = ["Full (100 tasks)", "Mini (20 tasks)", "By frequency", "By domain
|
|
| 42 |
FREQUENCY_OPTIONS = list(FREQUENCY_GROUPS.keys())
|
| 43 |
DOMAIN_OPTIONS = list(DOMAIN_GROUPS.keys())
|
| 44 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 45 |
# Mapping from UI selections to table directory names
|
| 46 |
GROUP_DIR_MAPPING = {
|
| 47 |
"Full (100 tasks)": "full",
|
|
@@ -117,7 +131,7 @@ with cols[1] as main_container:
|
|
| 117 |
pairwise_df = get_pairwise(selected_metric, group_dir)
|
| 118 |
|
| 119 |
st.markdown("## :material/trophy: Leaderboard", unsafe_allow_html=True)
|
| 120 |
-
st.markdown(
|
| 121 |
df_styled = format_leaderboard(metric_df)
|
| 122 |
st.dataframe(
|
| 123 |
df_styled,
|
|
|
|
| 11 |
from src.strings import (
|
| 12 |
CITATION_FEV,
|
| 13 |
CITATION_HEADER,
|
|
|
|
| 14 |
FEV_BENCHMARK_DETAILS,
|
| 15 |
PAIRWISE_BENCHMARK_DETAILS,
|
| 16 |
get_pivot_legend,
|
|
|
|
| 41 |
FREQUENCY_OPTIONS = list(FREQUENCY_GROUPS.keys())
|
| 42 |
DOMAIN_OPTIONS = list(DOMAIN_GROUPS.keys())
|
| 43 |
|
| 44 |
+
def get_subset_description(group_type: str, subgroup: str | None, num_tasks: int) -> str:
|
| 45 |
+
"""Generate a description of the current subset."""
|
| 46 |
+
base = f"Results for various forecasting models on **{num_tasks} tasks**"
|
| 47 |
+
if group_type == "Full (100 tasks)":
|
| 48 |
+
subset_desc = "from the full **fev-bench** benchmark"
|
| 49 |
+
elif group_type == "Mini (20 tasks)":
|
| 50 |
+
subset_desc = "from the **fev-bench-mini** subset"
|
| 51 |
+
elif group_type == "By frequency":
|
| 52 |
+
subset_desc = f"with **{subgroup.lower()}** frequency"
|
| 53 |
+
else: # By domain
|
| 54 |
+
subset_desc = f"from the **{subgroup}** domain"
|
| 55 |
+
paper_link = "[fev-bench: A Realistic Benchmark for Time Series Forecasting](https://arxiv.org/abs/2509.26468)"
|
| 56 |
+
return f"{base} {subset_desc}, as described in the paper {paper_link}."
|
| 57 |
+
|
| 58 |
+
|
| 59 |
# Mapping from UI selections to table directory names
|
| 60 |
GROUP_DIR_MAPPING = {
|
| 61 |
"Full (100 tasks)": "full",
|
|
|
|
| 131 |
pairwise_df = get_pairwise(selected_metric, group_dir)
|
| 132 |
|
| 133 |
st.markdown("## :material/trophy: Leaderboard", unsafe_allow_html=True)
|
| 134 |
+
st.markdown(get_subset_description(selected_group_type, selected_subgroup, len(task_list)), unsafe_allow_html=True)
|
| 135 |
df_styled = format_leaderboard(metric_df)
|
| 136 |
st.dataframe(
|
| 137 |
df_styled,
|