fix train autogluon

train_autogluon.py

@@ -1,39 +1,59 @@
-from autogluon.timeseries import TimeSeriesPredictor
+# train_autogluon.py
+from autogluon.timeseries import TimeSeriesPredictor, TimeSeriesDataFrame
 from utils_vol import fetch_close_series, realized_vol, rv_to_autogluon_df
 
-def train_bolt_small(ticker="AAPL", start="2015-01-01", interval="1d",
-                     prediction_length=30, time_limit=900):
+def train_bolt_small(
+    ticker="AAPL",
+    start="2015-01-01",
+    interval="1d",
+    prediction_length=30,
+    time_limit=900,              # seconds (15 min). Adjust if needed.
+):
     """
-    Trainiert Chronos-Bolt-Small mit AutoGluon auf CPU.
-    time_limit in Sekunden (Standard: 15 min).
+    Trains Chronos-Bolt-Small on CPU via AutoGluon with strict CPU-friendly limits.
+    Uses business-day ('B') frequency to handle weekends/holidays.
     """
-    print(f"[AutoFT] Lade {ticker}...")
+    print(f"[AutoFT] Lade {ticker} ...")
     close = fetch_close_series(ticker, start=start, interval=interval)
     rv = realized_vol(close)
-    df = rv_to_autogluon_df(rv)
+
+    # Make a tidy dataframe
+    df = rv_to_autogluon_df(rv)  # columns: item_id, timestamp, target
+
+    # Create a TimeSeriesDataFrame with explicit frequency
+    tsdf = TimeSeriesDataFrame.from_data_frame(
+        df,
+        id_column="item_id",
+        timestamp_column="timestamp",
+        target_column="target",
+        freq="B",                           # <- set business-day freq
+    )
+    # Ensure regular frequency (fills gaps with NaNs)
+    tsdf = tsdf.convert_frequency("B")
 
     predictor = TimeSeriesPredictor(
         path="/mnt/data/AutogluonChronosBoltSmall",
         prediction_length=prediction_length,
         eval_metric="WQL",
+        freq="B",                           # <- set freq on predictor too
         verbosity=2,
     )
 
     predictor.fit(
-        train_data=df,
+        train_data=tsdf,
         enable_ensemble=False,
         num_val_windows=1,
         hyperparameters={
             "Chronos": {
                 "model_path": "autogluon/chronos-bolt-small",
                 "fine_tune": True,
-                "fine_tune_steps": 200,
+                "fine_tune_steps": 200,     # keep small for CPU
                 "fine_tune_lr": 1e-4,
-                "context_length": 128,
+                "context_length": 128,      # keep small for CPU
                 "quantile_levels": [0.1, 0.5, 0.9],
             }
         },
-        time_limit=time_limit,
+        time_limit=time_limit,               # hard cap so HF Space won’t time out
     )
 
     print("✅ Training abgeschlossen. Modellpfad:", predictor.path)