diff --git a/test_se_multi_gpu.py b/test_se_multi_gpu.py
index 794c8f55..edb40c43 100644
--- a/test_se_multi_gpu.py
+++ b/test_se_multi_gpu.py
@@ -56,13 +56,15 @@ for gpu in [0, 1]:
     
     # Run
     x = torch.randn(1, 7168, dtype=torch.bfloat16, device=dev)
-    out = se.run(x)
     
-    # Fix gsa
-    if hasattr(se, '_saved_l1_gsa'):
-        se._l1_activation_global_scale = se._saved_l1_gsa
-    if hasattr(se, '_saved_l2_gsa'):
-        se._l2_activation_global_scale = se._saved_l2_gsa
+    # Must set gsa AFTER _ensure_initialized but BEFORE run
+    # _ensure_initialized is called lazily in run(), so we need to call it first
+    se._ensure_initialized()
+    # Now fix the gsa
+    se._l1_activation_global_scale = gisc.float().item()
+    se._l2_activation_global_scale = disc.float().item()
+    
+    out = se.run(x)
     
     has_nan = torch.isnan(out).any().item()
     print(f"GPU {gpu}: |out|={out.abs().max().item() if not has_nan else 'NaN'} has_nan={has_nan} shape={out.shape}")