Fix quantile transform scaling and document

example/data_utils.py

@@ -288,6 +288,8 @@ def normalize_cont(
         if not quantile_probs or not quantile_values:
             raise ValueError("use_quantile_transform enabled but quantile stats missing")
         x = apply_quantile_transform(x, cont_cols, quantile_probs, quantile_values)
+        # quantile transform already targets N(0,1); skip extra standardization
+        return x
     mean_t = torch.tensor([mean[c] for c in cont_cols], dtype=x.dtype, device=x.device)
     std_t = torch.tensor([std[c] for c in cont_cols], dtype=x.dtype, device=x.device)
     return (x - mean_t) / std_t