Add toggleable metric charts

app.py

@@ -13,6 +13,7 @@ import numpy as np
 import pandas as pd
 import plotly.express as px
 import plotly.graph_objects as go
+from plotly.subplots import make_subplots
 from sklearn.base import clone
 from sklearn.compose import ColumnTransformer
 from sklearn.datasets import fetch_california_housing, fetch_openml, load_breast_cancer, load_diabetes, load_digits, load_iris, load_wine, make_classification
@@ -37,6 +38,8 @@ APP_TITLE = "tabBench"
 TABFM_MODEL_ID = "google/tabfm-1.0.0-pytorch"
 RANDOM_STATE = 42
 CANDY_DATA_URL = "https://raw.githubusercontent.com/fivethirtyeight/data/master/candy-power-ranking/candy-data.csv"
+GOOGLE_COLORS = ["#4285F4", "#DB4437", "#F4B400", "#0F9D58", "#A142F4", "#00ACC1"]
+METRIC_CHOICES = ["accuracy", "f1_weighted", "roc_auc", "rmse", "mae", "r2", "seconds"]
 
 
 @dataclass(frozen=True)
@@ -609,19 +612,83 @@ def benchmark_frame(
     return results[["rank", "model", "status", *metric_cols]], preview, summary
 
 
-def metric_chart(results: pd.DataFrame) -> go.Figure:
+def metric_chart(results: pd.DataFrame, selected_metrics: list[str] | None = None, chart_style: str = "Line") -> go.Figure:
     if results is None or results.empty:
         return go.Figure()
-    long_cols = [c for c in ["accuracy", "f1_weighted", "roc_auc", "r2"] if c in results.columns and results[c].notna().any()]
-    if long_cols:
-        long = results.melt(id_vars=["model"], value_vars=long_cols, var_name="metric", value_name="score")
-        fig = px.bar(long, x="model", y="score", color="metric", barmode="group", color_discrete_sequence=["#f97316", "#8b5cf6", "#14b8a6", "#2563eb"])
-        fig.update_yaxes(range=[0, 1])
-    else:
-        long_cols = [c for c in ["rmse", "mae"] if c in results.columns and results[c].notna().any()]
-        long = results.melt(id_vars=["model"], value_vars=long_cols, var_name="metric", value_name="score")
-        fig = px.bar(long, x="model", y="score", color="metric", barmode="group", color_discrete_sequence=["#f97316", "#8b5cf6"])
-    fig.update_layout(template="plotly_white", height=360, margin=dict(l=20, r=20, t=25, b=20), legend_title_text="")
+    selected_metrics = selected_metrics or METRIC_CHOICES
+    metric_cols = [c for c in selected_metrics if c in results.columns and results[c].notna().any()]
+    if not metric_cols:
+        metric_cols = [c for c in METRIC_CHOICES if c in results.columns and results[c].notna().any()]
+    if not metric_cols:
+        return go.Figure()
+
+    clean = results.sort_values("rank") if "rank" in results.columns else results.copy()
+    x_labels = clean["model"].astype(str).tolist()
+    if chart_style == "Radar":
+        normalized = clean[["model", *metric_cols]].copy()
+        for metric in metric_cols:
+            values = pd.to_numeric(normalized[metric], errors="coerce")
+            lo, hi = values.min(), values.max()
+            if pd.isna(lo) or pd.isna(hi) or hi == lo:
+                normalized[metric] = 0.5
+            elif metric in {"rmse", "mae", "seconds"}:
+                normalized[metric] = 1 - ((values - lo) / (hi - lo))
+            else:
+                normalized[metric] = (values - lo) / (hi - lo)
+        fig = go.Figure()
+        theta = metric_cols + [metric_cols[0]]
+        for idx, row in normalized.iterrows():
+            values = [row[m] for m in metric_cols] + [row[metric_cols[0]]]
+            fig.add_trace(
+                go.Scatterpolar(
+                    r=values,
+                    theta=theta,
+                    fill="toself",
+                    name=str(row["model"]),
+                    line=dict(color=GOOGLE_COLORS[idx % len(GOOGLE_COLORS)], width=2),
+                    opacity=0.78,
+                )
+            )
+        fig.update_layout(
+            template="plotly_white",
+            height=420,
+            margin=dict(l=35, r=35, t=35, b=25),
+            polar=dict(radialaxis=dict(visible=True, range=[0, 1])),
+            legend_title_text="Model",
+            title="Normalized metric shape (higher is better)",
+        )
+        return fig
+
+    fig = make_subplots(
+        rows=len(metric_cols),
+        cols=1,
+        shared_xaxes=True,
+        vertical_spacing=0.08,
+        subplot_titles=[metric.replace("_", " ").upper() for metric in metric_cols],
+    )
+    for idx, metric in enumerate(metric_cols, start=1):
+        fig.add_trace(
+            go.Scatter(
+                x=x_labels,
+                y=clean[metric],
+                mode="lines+markers",
+                name=metric,
+                line=dict(color=GOOGLE_COLORS[(idx - 1) % len(GOOGLE_COLORS)], width=3, shape="spline"),
+                marker=dict(size=9, line=dict(color="white", width=1.5)),
+                hovertemplate=f"<b>%{{x}}</b><br>{metric}: %{{y:.4f}}<extra></extra>",
+            ),
+            row=idx,
+            col=1,
+        )
+        if metric in {"accuracy", "f1_weighted", "roc_auc", "r2"}:
+            fig.update_yaxes(range=[min(0, float(clean[metric].min()) - 0.05), 1.02], row=idx, col=1)
+    fig.update_layout(
+        template="plotly_white",
+        height=max(320, 185 * len(metric_cols)),
+        margin=dict(l=30, r=20, t=40, b=35),
+        showlegend=False,
+        hovermode="x unified",
+    )
     return fig
 
 
@@ -640,20 +707,46 @@ def time_chart(results: pd.DataFrame) -> go.Figure:
     return fig
 
 
-def run_catalog(dataset_name: str, sample_size: int, test_percent: int, seed: int, selected_models: list[str], include_tabfm: bool):
+def run_catalog(
+    dataset_name: str,
+    sample_size: int,
+    test_percent: int,
+    seed: int,
+    selected_models: list[str],
+    include_tabfm: bool,
+    selected_metrics: list[str],
+    chart_style: str,
+):
     spec = get_spec(dataset_name)
     df = get_dataset(dataset_name, sample_size, seed)
     results, preview, summary = benchmark_frame(df, spec.target, spec.task, sample_size, test_percent / 100, seed, selected_models, include_tabfm)
-    return summary, results.round(4), metric_chart(results), time_chart(results), preview
+    return summary, results.round(4), metric_chart(results, selected_metrics, chart_style), time_chart(results), preview
 
 
-def run_upload(file, target: str, task: str, sample_size: int, test_percent: int, seed: int, selected_models: list[str], include_tabfm: bool):
+def run_upload(
+    file,
+    target: str,
+    task: str,
+    sample_size: int,
+    test_percent: int,
+    seed: int,
+    selected_models: list[str],
+    include_tabfm: bool,
+    selected_metrics: list[str],
+    chart_style: str,
+):
     if file is None:
         raise gr.Error("Upload a CSV file first.")
     df = pd.read_csv(file.name)
     selected_task = None if task == "Auto" else task.lower()
     results, preview, summary = benchmark_frame(df, target, selected_task, sample_size, test_percent / 100, seed, selected_models, include_tabfm)
-    return summary, results.round(4), metric_chart(results), time_chart(results), preview
+    return summary, results.round(4), metric_chart(results, selected_metrics, chart_style), time_chart(results), preview
+
+
+def redraw_metric_chart(results: pd.DataFrame, selected_metrics: list[str], chart_style: str):
+    if results is None or len(results) == 0:
+        return go.Figure()
+    return metric_chart(pd.DataFrame(results), selected_metrics, chart_style)
 
 
 def catalog_table() -> pd.DataFrame:
@@ -714,6 +807,8 @@ def build_app() -> gr.Blocks:
                             seed = gr.Number(value=42, precision=0, label="Random seed")
                             models = gr.CheckboxGroup(DEFAULT_MODELS, value=["RandomForest", "HistGradientBoosting", "XGBoost", "Dummy"], label="Baselines")
                             include_tabfm = gr.Checkbox(value=False, label="Run TabFM live")
+                            metric_toggles = gr.CheckboxGroup(METRIC_CHOICES, value=["rmse", "mae", "r2", "accuracy", "f1_weighted", "roc_auc"], label="Chart metrics")
+                            chart_style = gr.Radio(["Line", "Radar"], value="Line", label="Chart style")
                             run_btn = gr.Button("Run benchmark", variant="primary")
                         with gr.Column(scale=3):
                             summary = gr.Markdown()
@@ -722,8 +817,11 @@ def build_app() -> gr.Blocks:
                                 chart = gr.Plot(label="Metric comparison")
                                 speed = gr.Plot(label="Speed")
                             preview = gr.Dataframe(label="Held-out preview", interactive=False)
-                    run_btn.click(run_catalog, [dataset, sample, test_pct, seed, models, include_tabfm], [summary, leaderboard, chart, speed, preview])
-                    demo.load(run_catalog, [dataset, sample, test_pct, seed, models, include_tabfm], [summary, leaderboard, chart, speed, preview])
+                    run_inputs = [dataset, sample, test_pct, seed, models, include_tabfm, metric_toggles, chart_style]
+                    run_btn.click(run_catalog, run_inputs, [summary, leaderboard, chart, speed, preview])
+                    metric_toggles.change(redraw_metric_chart, [leaderboard, metric_toggles, chart_style], chart)
+                    chart_style.change(redraw_metric_chart, [leaderboard, metric_toggles, chart_style], chart)
+                    demo.load(run_catalog, run_inputs, [summary, leaderboard, chart, speed, preview])
                 with gr.Tab("Upload Dataset"):
                     with gr.Row():
                         with gr.Column(scale=1, elem_classes=["panel"]):
@@ -735,6 +833,8 @@ def build_app() -> gr.Blocks:
                             upload_seed = gr.Number(value=42, precision=0, label="Random seed")
                             upload_models = gr.CheckboxGroup(DEFAULT_MODELS, value=["RandomForest", "HistGradientBoosting", "XGBoost", "Dummy"], label="Baselines")
                             upload_tabfm = gr.Checkbox(value=False, label="Run TabFM live")
+                            upload_metric_toggles = gr.CheckboxGroup(METRIC_CHOICES, value=["rmse", "mae", "r2", "accuracy", "f1_weighted", "roc_auc"], label="Chart metrics")
+                            upload_chart_style = gr.Radio(["Line", "Radar"], value="Line", label="Chart style")
                             upload_btn = gr.Button("Run uploaded dataset", variant="primary")
                         with gr.Column(scale=3):
                             upload_summary = gr.Markdown()
@@ -745,9 +845,11 @@ def build_app() -> gr.Blocks:
                             upload_preview = gr.Dataframe(label="Held-out preview", interactive=False)
                     upload_btn.click(
                         run_upload,
-                        [file, target, task, upload_sample, upload_test_pct, upload_seed, upload_models, upload_tabfm],
+                        [file, target, task, upload_sample, upload_test_pct, upload_seed, upload_models, upload_tabfm, upload_metric_toggles, upload_chart_style],
                         [upload_summary, upload_leaderboard, upload_chart, upload_speed, upload_preview],
                     )
+                    upload_metric_toggles.change(redraw_metric_chart, [upload_leaderboard, upload_metric_toggles, upload_chart_style], upload_chart)
+                    upload_chart_style.change(redraw_metric_chart, [upload_leaderboard, upload_metric_toggles, upload_chart_style], upload_chart)
                 with gr.Tab("Dataset Catalog"):
                     gr.Dataframe(catalog_table(), interactive=False, label="Included benchmark catalog")
                     gr.Markdown(