Deploy Space: update src

src/display/css_html_js.py

@@ -1,8 +1,74 @@
 # Gradio custom styles for the TSFM Realworld Bench leaderboard.
 custom_css = """
-.markdown-text { font-size: 16px !important; }
-.tab-buttons button { font-size: 20px; }
+:root {
+    --bench-ink: #18202f;
+    --bench-muted: #5b6575;
+    --bench-line: #d7dce5;
+    --bench-panel: #f8fafc;
+    --bench-blue: #2563eb;
+    --bench-teal: #0f766e;
+}
+.gradio-container {
+    max-width: 1320px !important;
+    margin: 0 auto !important;
+}
+.markdown-text {
+    font-size: 16px !important;
+    color: var(--bench-muted);
+}
+.tab-buttons button {
+    font-size: 16px !important;
+    font-weight: 650 !important;
+    border-radius: 6px 6px 0 0 !important;
+}
 #leaderboard-table { margin-top: 15px; }
+.summary-grid {
+    display: grid;
+    grid-template-columns: repeat(4, minmax(0, 1fr));
+    gap: 12px;
+    margin: 18px 0 22px;
+}
+.summary-card {
+    border: 1px solid var(--bench-line);
+    border-radius: 8px;
+    background: linear-gradient(180deg, #ffffff 0%, var(--bench-panel) 100%);
+    padding: 14px 16px;
+    box-shadow: 0 1px 2px rgba(15, 23, 42, 0.05);
+}
+.summary-value {
+    color: var(--bench-ink);
+    font-size: 28px;
+    line-height: 1.1;
+    font-weight: 780;
+}
+.summary-value.small {
+    font-size: 18px;
+    line-height: 1.35;
+}
+.summary-label {
+    margin-top: 5px;
+    color: var(--bench-muted);
+    font-size: 12px;
+    text-transform: uppercase;
+    letter-spacing: 0.04em;
+    font-weight: 700;
+}
+.dataframe {
+    border-radius: 8px !important;
+    overflow: hidden !important;
+    border: 1px solid var(--bench-line) !important;
+}
+.dataframe table {
+    font-size: 14px !important;
+}
+.dataframe th {
+    background: #eef2f7 !important;
+    color: var(--bench-ink) !important;
+    font-weight: 750 !important;
+}
+.dataframe td {
+    color: #263244 !important;
+}
 .domain-heading {
     margin-top: 1.75rem;
     margin-bottom: 0.5rem;
@@ -12,14 +78,19 @@ custom_css = """
 .dataset-heading {
     margin-top: 2rem;
     margin-bottom: 0.75rem;
-    font-size: 1.45rem !important;
+    font-size: 1.25rem !important;
     font-weight: 700 !important;
     line-height: 1.35;
-    color: #e2e8f0;
+    color: var(--bench-ink);
 }
 .dataset-heading p {
-    font-size: 1.45rem !important;
+    font-size: 1.25rem !important;
     font-weight: 700 !important;
     margin: 0;
 }
+@media (max-width: 820px) {
+    .summary-grid {
+        grid-template-columns: repeat(2, minmax(0, 1fr));
+    }
+}
 """

src/utils.py

@@ -10,6 +10,8 @@ import numpy as np
 import pandas as pd
 from scipy import stats
 
+from src.display.formatting import format_timestamp_utc8
+
 METRIC_COLUMNS = [
     "eval_metrics/MSE[mean]",
     "eval_metrics/MSE[0.5]",
@@ -123,16 +125,6 @@ def load_all_results(root_dir: str = "results") -> pd.DataFrame:
     df = df.sort_values(by=["model", "dataset"]).reset_index(drop=True)
     df["model"] = df["model"].replace(LEGACY_MODEL_NAMES)
     
-    ALLOWED_MODELS = {
-        "Toto-2.0-1B",
-        "Moirai-1.1-Large",
-        "Chronos-2",
-        "Timer-S1",
-        "TiRex",
-        "TimesFM-2.5"
-    }
-    df = df[df["model"].isin(ALLOWED_MODELS)]
-
     parts = df["dataset"].astype(str).str.split("/", expand=True)
     if parts.shape[1] < 3:
         for idx in range(parts.shape[1], 3):
@@ -364,6 +356,40 @@ def group_datasets_by_domain(
     return ordered
 
 
+def build_leaderboard_summary_html(root_dir: str = "results") -> str:
+    df = prepare_results_df(root_dir)
+    if df.empty:
+        return """
+        <div class="summary-grid">
+          <div class="summary-card"><div class="summary-value">0</div><div class="summary-label">Models</div></div>
+          <div class="summary-card"><div class="summary-value">0</div><div class="summary-label">Datasets</div></div>
+          <div class="summary-card"><div class="summary-value">0</div><div class="summary-label">Domains</div></div>
+          <div class="summary-card"><div class="summary-value">n/a</div><div class="summary-label">Last Refresh</div></div>
+        </div>
+        """
+
+    models = df["model"].nunique()
+    datasets = df["dataset"].nunique()
+    domains = df["domain"].nunique() if "domain" in df.columns else 0
+    last_refresh = "n/a"
+    status_path = Path(root_dir) / "online_status.json"
+    if status_path.exists():
+        try:
+            status = json.loads(status_path.read_text())
+            last_refresh = format_timestamp_utc8(status.get("finished_at", ""))
+        except Exception:
+            pass
+
+    return f"""
+    <div class="summary-grid">
+      <div class="summary-card"><div class="summary-value">{models}</div><div class="summary-label">Models</div></div>
+      <div class="summary-card"><div class="summary-value">{datasets}</div><div class="summary-label">Datasets</div></div>
+      <div class="summary-card"><div class="summary-value">{domains}</div><div class="summary-label">Domains</div></div>
+      <div class="summary-card wide"><div class="summary-value small">{last_refresh}</div><div class="summary-label">Last Refresh</div></div>
+    </div>
+    """
+
+
 def dataset_tab_label(dataset: str, root_dir: str = "results") -> str:
     raw_domain = lookup_dataset_domain(dataset, root_dir)
     domain = display_domain(raw_domain)
@@ -638,10 +664,9 @@ def _today_start_utc8() -> datetime:
 
 def get_dataset_last_eval_time(ds_base: str, root_dir: str = "results") -> datetime | None:
     latest_dt: datetime | None = None
-    ALLOWED_SLUGS = {"chronos_2", "moirai_1_1_large", "timer_s1", "timesfm_2_5", "tirex", "toto_2_0_1b"}
     # Scan model directories
     for model_dir in Path(root_dir).iterdir():
-        if not model_dir.is_dir() or model_dir.name not in ALLOWED_SLUGS:
+        if not model_dir.is_dir():
             continue
         forecasts_dir = model_dir / "forecasts"
         if not forecasts_dir.is_dir():
@@ -716,9 +741,8 @@ def make_domain_pie_chart(root_dir: str = "results") -> object | None:
     # Fallback to cumulative snapshots count (no today filter) if empty
     if not timestamp_counts:
         seen_datasets: set[str] = set()
-        ALLOWED_SLUGS = {"chronos_2", "moirai_1_1_large", "timer_s1", "timesfm_2_5", "tirex", "toto_2_0_1b"}
         for model_dir in sorted(Path(root_dir).iterdir()):
-            if not model_dir.is_dir() or model_dir.name not in ALLOWED_SLUGS:
+            if not model_dir.is_dir():
                 continue
             forecasts_dir = model_dir / "forecasts"
             if not forecasts_dir.is_dir():