Upload train_v38_2.py with huggingface_hub

train_v38_2.py

@@ -54,7 +54,7 @@ def safe_num(v, default=-1):
 # 1. LOAD INTEGRATED FEATURE MATRIX
 # ============================================================
 print("=" * 70)
-print("  V38.2 BARE MODEL: MULTI-SEED ENSEMBLE WITH 116 FEATURES (v3 cleaned)")
+print("  V38.2 BARE MODEL v4: MULTI-SEED ENSEMBLE WITH 148 FEATURES (v4 + PS BERT + HS×Univ + Summer)")
 print("=" * 70)
 
 df_raw = pd.read_csv(os.path.join(DATA_DIR, 'v38_2_integrated_features.csv'))
@@ -573,6 +573,9 @@ feature_groups = {
     'Taste': [c for c in feature_cols if c.startswith('taste_')],
     'Cuilu': [c for c in feature_cols if c.startswith('cuilu_')],
     'Major': [c for c in feature_cols if c.startswith('major_') or c in ['has_secondary_major', 'has_college_division', 'is_undecided', 'has_major', 'school_major_count']],
+    'PS BERT PCA': [c for c in feature_cols if c.startswith('ps_bert_')],
+    'HS×Univ History': [c for c in feature_cols if c.startswith('hs_') and any(k in c for k in ['hist', 'overall', 'smooth', 'apps', 'admits'])],
+    'Summer Programs': [c for c in feature_cols if c.startswith('summer_')],
 }
 
 for group_name, group_cols in feature_groups.items():