Add more models and show TabFM status

README.md

@@ -30,7 +30,7 @@ The app includes:
 - A benchmark catalog with 10+ common tabular tasks, including Titanic-style survival, housing prices, fraud detection, recipe ratings, Halloween candy ranking, and classic sklearn datasets.
 - A leaderboard with model metrics, timing, and task-aware ranking.
 - Charts for accuracy/F1/AUC or RMSE/MAE/R2.
-- Controls for sample size, train/test split, random seed, model selection, and TabFM inclusion.
+- Controls for sample size, train/test split, random seed, model selection, chart metrics, and TabFM inclusion.
 - A CSV upload flow so visitors can run the same arena on their own dataset.
 
 The built-in catalog uses real open datasets where possible:
@@ -42,6 +42,8 @@ The built-in catalog uses real open datasets where possible:
 
 TabFM is loaded through the public Google Research package when available. The app keeps graceful fallbacks so the Space still works on CPU-only or dependency-constrained runtimes. Live TabFM runs are opt-in because the model weights are large and CPU-only inference can be slow.
 
+Model choices include Logistic/Ridge, RandomForest, ExtraTrees, GradientBoosting, HistGradientBoosting, AdaBoost, KNN, SVM, XGBoost, LightGBM, Dummy, and optional live TabFM.
+
 Note: TabFM's weights use their own non-commercial license. Review the upstream model license before using this Space commercially.
 
 Links:

app.py

@@ -18,7 +18,18 @@ 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
 from sklearn.dummy import DummyClassifier, DummyRegressor
-from sklearn.ensemble import HistGradientBoostingClassifier, HistGradientBoostingRegressor, RandomForestClassifier, RandomForestRegressor
+from sklearn.ensemble import (
+    AdaBoostClassifier,
+    AdaBoostRegressor,
+    ExtraTreesClassifier,
+    ExtraTreesRegressor,
+    GradientBoostingClassifier,
+    GradientBoostingRegressor,
+    HistGradientBoostingClassifier,
+    HistGradientBoostingRegressor,
+    RandomForestClassifier,
+    RandomForestRegressor,
+)
 from sklearn.impute import SimpleImputer
 from sklearn.linear_model import LogisticRegression, Ridge
 from sklearn.metrics import (
@@ -30,8 +41,10 @@ from sklearn.metrics import (
     roc_auc_score,
 )
 from sklearn.model_selection import train_test_split
+from sklearn.neighbors import KNeighborsClassifier, KNeighborsRegressor
 from sklearn.pipeline import Pipeline
 from sklearn.preprocessing import OneHotEncoder, StandardScaler
+from sklearn.svm import SVC, SVR
 
 
 APP_TITLE = "tabBench"
@@ -716,14 +729,24 @@ def available_baselines(task: str) -> dict[str, object]:
         models: dict[str, object] = {
             "Logistic": Pipeline([("prep", make_preprocessor(pd.DataFrame(), True)), ("model", LogisticRegression(max_iter=800))]),
             "RandomForest": Pipeline([("prep", make_preprocessor(pd.DataFrame())), ("model", RandomForestClassifier(n_estimators=80, min_samples_leaf=2, n_jobs=-1, random_state=RANDOM_STATE))]),
+            "ExtraTrees": Pipeline([("prep", make_preprocessor(pd.DataFrame())), ("model", ExtraTreesClassifier(n_estimators=120, min_samples_leaf=2, n_jobs=-1, random_state=RANDOM_STATE))]),
+            "GradientBoosting": Pipeline([("prep", make_preprocessor(pd.DataFrame())), ("model", GradientBoostingClassifier(n_estimators=100, learning_rate=0.06, random_state=RANDOM_STATE))]),
             "HistGradientBoosting": Pipeline([("prep", make_preprocessor(pd.DataFrame())), ("model", HistGradientBoostingClassifier(max_iter=120, random_state=RANDOM_STATE))]),
+            "AdaBoost": Pipeline([("prep", make_preprocessor(pd.DataFrame())), ("model", AdaBoostClassifier(n_estimators=80, learning_rate=0.08, random_state=RANDOM_STATE))]),
+            "KNN": Pipeline([("prep", make_preprocessor(pd.DataFrame(), True)), ("model", KNeighborsClassifier(n_neighbors=7))]),
+            "SVM": Pipeline([("prep", make_preprocessor(pd.DataFrame(), True)), ("model", SVC(C=1.0, probability=True, random_state=RANDOM_STATE))]),
             "Dummy": Pipeline([("prep", make_preprocessor(pd.DataFrame())), ("model", DummyClassifier(strategy="most_frequent"))]),
         }
     else:
         models = {
             "Ridge": Pipeline([("prep", make_preprocessor(pd.DataFrame(), True)), ("model", Ridge(alpha=1.0))]),
             "RandomForest": Pipeline([("prep", make_preprocessor(pd.DataFrame())), ("model", RandomForestRegressor(n_estimators=80, min_samples_leaf=2, n_jobs=-1, random_state=RANDOM_STATE))]),
+            "ExtraTrees": Pipeline([("prep", make_preprocessor(pd.DataFrame())), ("model", ExtraTreesRegressor(n_estimators=120, min_samples_leaf=2, n_jobs=-1, random_state=RANDOM_STATE))]),
+            "GradientBoosting": Pipeline([("prep", make_preprocessor(pd.DataFrame())), ("model", GradientBoostingRegressor(n_estimators=100, learning_rate=0.06, random_state=RANDOM_STATE))]),
             "HistGradientBoosting": Pipeline([("prep", make_preprocessor(pd.DataFrame())), ("model", HistGradientBoostingRegressor(max_iter=120, random_state=RANDOM_STATE))]),
+            "AdaBoost": Pipeline([("prep", make_preprocessor(pd.DataFrame())), ("model", AdaBoostRegressor(n_estimators=80, learning_rate=0.08, random_state=RANDOM_STATE))]),
+            "KNN": Pipeline([("prep", make_preprocessor(pd.DataFrame(), True)), ("model", KNeighborsRegressor(n_neighbors=7))]),
+            "SVM": Pipeline([("prep", make_preprocessor(pd.DataFrame(), True)), ("model", SVR(C=1.0))]),
             "Dummy": Pipeline([("prep", make_preprocessor(pd.DataFrame())), ("model", DummyRegressor(strategy="median"))]),
         }
     if importlib.util.find_spec("xgboost"):
@@ -743,6 +766,26 @@ def available_baselines(task: str) -> dict[str, object]:
                     ("model", XGBRegressor(n_estimators=80, max_depth=4, learning_rate=0.08, random_state=RANDOM_STATE)),
                 ]
             )
+    if importlib.util.find_spec("lightgbm"):
+        try:
+            from lightgbm import LGBMClassifier, LGBMRegressor
+
+            if task == "classification":
+                models["LightGBM"] = Pipeline(
+                    [
+                        ("prep", make_preprocessor(pd.DataFrame())),
+                        ("model", LGBMClassifier(n_estimators=120, learning_rate=0.06, random_state=RANDOM_STATE, verbose=-1)),
+                    ]
+                )
+            else:
+                models["LightGBM"] = Pipeline(
+                    [
+                        ("prep", make_preprocessor(pd.DataFrame())),
+                        ("model", LGBMRegressor(n_estimators=120, learning_rate=0.06, random_state=RANDOM_STATE, verbose=-1)),
+                    ]
+                )
+        except Exception:
+            pass
     return models
 
 
@@ -860,6 +903,9 @@ def benchmark_frame(
             rows.append({"model": name, "status": "ok", "seconds": time.perf_counter() - start, **metrics})
         except Exception as exc:
             rows.append({"model": name, "status": f"failed: {exc}", "seconds": time.perf_counter() - start})
+    for name in selected_models:
+        if name not in models:
+            rows.append({"model": name, "status": "unavailable in this runtime", "seconds": 0.0, "rank_score": np.nan})
 
     if include_tabfm:
         start = time.perf_counter()
@@ -883,6 +929,8 @@ def benchmark_frame(
         except Exception as exc:
             rows.append({"model": "TabFM", "status": f"unavailable: {exc}", "seconds": time.perf_counter() - start})
             notes.append(f"TabFM did not run in this environment. On Spaces, keep Python 3.11 and allow the GitHub dependency plus model download for `{TABFM_MODEL_ID}`.")
+    else:
+        rows.append({"model": "TabFM", "status": "skipped - enable Run TabFM live", "seconds": 0.0, "rank_score": np.nan})
 
     results = pd.DataFrame(rows)
     metric_cols = [c for c in ["accuracy", "f1_weighted", "roc_auc", "rmse", "mae", "r2", "rank_score", "seconds"] if c in results.columns]
@@ -913,6 +961,9 @@ def metric_chart(results: pd.DataFrame, selected_metrics: list[str] | None = Non
         return go.Figure()
 
     clean = results.sort_values("rank") if "rank" in results.columns else results.copy()
+    clean = clean.dropna(subset=metric_cols, how="all")
+    if clean.empty:
+        return go.Figure()
     x_labels = clean["model"].astype(str).tolist()
     if chart_style == "Radar":
         normalized = clean[["model", *metric_cols]].copy()
@@ -985,6 +1036,7 @@ def metric_chart(results: pd.DataFrame, selected_metrics: list[str] | None = Non
 def bar_chart(results: pd.DataFrame, selected_metrics: list[str] | None = None) -> go.Figure:
     if results is None or results.empty:
         return go.Figure()
+    results = results.copy()
     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:
@@ -992,6 +1044,9 @@ def bar_chart(results: pd.DataFrame, selected_metrics: list[str] | None = None)
     if not metric_cols:
         return go.Figure()
     clean = results.sort_values("rank") if "rank" in results.columns else results.copy()
+    clean = clean.dropna(subset=metric_cols, how="all")
+    if clean.empty:
+        return go.Figure()
     long = clean.melt(id_vars=["model"], value_vars=metric_cols, var_name="metric", value_name="score")
     fig = px.bar(
         long,
@@ -1015,6 +1070,11 @@ def bar_chart(results: pd.DataFrame, selected_metrics: list[str] | None = None)
 def time_chart(results: pd.DataFrame) -> go.Figure:
     if results is None or results.empty or "seconds" not in results:
         return go.Figure()
+    results = results.copy()
+    if "status" in results.columns:
+        results = results[~results["status"].astype(str).str.startswith("skipped")]
+    if results.empty:
+        return go.Figure()
     fig = px.scatter(
         results,
         x="seconds",
@@ -1111,7 +1171,21 @@ def catalog_table() -> pd.DataFrame:
     )
 
 
-DEFAULT_MODELS = ["Logistic", "Ridge", "RandomForest", "HistGradientBoosting", "XGBoost", "Dummy"]
+DEFAULT_MODELS = [
+    "Logistic",
+    "Ridge",
+    "RandomForest",
+    "ExtraTrees",
+    "GradientBoosting",
+    "HistGradientBoosting",
+    "AdaBoost",
+    "KNN",
+    "SVM",
+    "XGBoost",
+    "LightGBM",
+    "Dummy",
+]
+DEFAULT_SELECTED_MODELS = ["HistGradientBoosting", "XGBoost", "LightGBM", "Dummy"]
 
 
 def build_app() -> gr.Blocks:
@@ -1151,8 +1225,8 @@ def build_app() -> gr.Blocks:
                             sample = gr.Slider(100, 50000, value=1000, step=100, label="Sample size")
                             test_pct = gr.Slider(10, 40, value=25, step=5, label="Test split (%)")
                             seed = gr.Number(value=42, precision=0, label="Random seed")
-                            models = gr.CheckboxGroup(DEFAULT_MODELS, value=["HistGradientBoosting", "Dummy"], label="Baselines")
-                            include_tabfm = gr.Checkbox(value=False, label="Run TabFM live")
+                            models = gr.CheckboxGroup(DEFAULT_MODELS, value=DEFAULT_SELECTED_MODELS, label="Baselines")
+                            include_tabfm = gr.Checkbox(value=False, label="Run TabFM live (adds TabFM row)")
                             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")
                             with gr.Accordion("TabFM tuning", open=False):
@@ -1169,10 +1243,10 @@ def build_app() -> gr.Blocks:
                         with gr.Column(scale=3):
                             summary = gr.Markdown()
                             leaderboard = gr.Dataframe(label="Leaderboard", interactive=False)
+                            bars = gr.Plot(label="Main grouped comparison")
                             with gr.Row():
                                 chart = gr.Plot(label="Metric comparison")
                                 speed = gr.Plot(label="Speed")
-                            bars = gr.Plot(label="Grouped comparison")
                             preview = gr.Dataframe(label="Held-out preview", interactive=False)
                     run_inputs = [
                         dataset,
@@ -1208,8 +1282,8 @@ def build_app() -> gr.Blocks:
                             upload_sample = gr.Slider(100, 50000, value=1000, step=100, label="Sample size")
                             upload_test_pct = gr.Slider(10, 40, value=25, step=5, label="Test split (%)")
                             upload_seed = gr.Number(value=42, precision=0, label="Random seed")
-                            upload_models = gr.CheckboxGroup(DEFAULT_MODELS, value=["HistGradientBoosting", "Dummy"], label="Baselines")
-                            upload_tabfm = gr.Checkbox(value=False, label="Run TabFM live")
+                            upload_models = gr.CheckboxGroup(DEFAULT_MODELS, value=DEFAULT_SELECTED_MODELS, label="Baselines")
+                            upload_tabfm = gr.Checkbox(value=False, label="Run TabFM live (adds TabFM row)")
                             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")
                             with gr.Accordion("TabFM tuning", open=False):
@@ -1226,10 +1300,10 @@ def build_app() -> gr.Blocks:
                         with gr.Column(scale=3):
                             upload_summary = gr.Markdown()
                             upload_leaderboard = gr.Dataframe(label="Upload leaderboard", interactive=False)
+                            upload_bars = gr.Plot(label="Main grouped comparison")
                             with gr.Row():
                                 upload_chart = gr.Plot(label="Metric comparison")
                                 upload_speed = gr.Plot(label="Speed")
-                            upload_bars = gr.Plot(label="Grouped comparison")
                             upload_preview = gr.Dataframe(label="Held-out preview", interactive=False)
                     upload_btn.click(
                         run_upload,

packages.txt

@@ -0,0 +1 @@
+libgomp1

requirements.txt

@@ -5,5 +5,6 @@ scikit-learn>=1.5.0,<1.9
 plotly>=5.22.0,<7
 scipy>=1.13.0,<1.18
 xgboost>=2.1.0,<3
+lightgbm>=4.5.0,<5
 kagglehub>=0.3.6,<1
 tabfm[pytorch] @ git+https://github.com/google-research/tabfm.git