fix len bug for std calc.

Author: janfb

sbi/analysis/sensitivity_analysis.py

@@ -47,7 +47,9 @@ def destandardizing_net(batch_t: Tensor, min_std: float = 1e-7) -> nn.Module:
     is_valid_t, *_ = handle_invalid_x(batch_t, True)
 
     t_mean = torch.mean(batch_t[is_valid_t], dim=0)
-    if len(batch_t > 1):
+
+    # Use batch size to decide whether a reliable std can be computed.
+    if len(batch_t) > 1:
         t_std = torch.std(batch_t[is_valid_t], dim=0)
         t_std[t_std < min_std] = min_std
     else: