update

tests/test_modeling_common.py

@@ -4004,6 +4004,8 @@ def get_mean_reldiff(failcase, x, ref, atol, rtol):
             config.rms_norm_eps = 1.0
             config.layer_norm_eps = 1.0
             config.norm_eps = 1.0
+            config.norm_epsilon = 1.0
+            config.layer_norm_epsilon = 1.0
             if hasattr(config, "text_config"):
                 config.text_config.rms_norm_eps = 1.0
                 config.text_config.layer_norm_eps = 1.0
@@ -4033,6 +4035,13 @@ def get_mean_reldiff(failcase, x, ref, atol, rtol):
                 )
                 model_eager = model_eager.eval().to(torch_device, dtype=torch_dtype)
 
+                for x in model_eager.modules():
+                    if isinstance(x, nn.LayerNorm):
+                        x.eps = 1.0
+                for x in model_sdpa.modules():
+                    if isinstance(x, nn.LayerNorm):
+                        x.eps = 1.0
+
                 # We use these for loops instead of parameterized.expand just for the interest of avoiding loading/saving 16 times the model,
                 # but it would be nicer to have an efficient way to use parameterized.expand
                 fail_cases = []