diff --git a/Paddle b/Paddle
index 935f0a6601f..fd66eaf3f0e 160000
--- a/Paddle
+++ b/Paddle
@@ -1 +1 @@
-Subproject commit 935f0a6601f90f4eb1d091682b31b243da50851b
+Subproject commit fd66eaf3f0e415f50f38bbff24fa84b9cc9f796a
diff --git a/backends/iluvatar_gpu/kernels/ernie_core/layer_norm_cuda_kernel.cu b/backends/iluvatar_gpu/kernels/ernie_core/layer_norm_cuda_kernel.cu
index 5f4288d97af..75417d058e6 100644
--- a/backends/iluvatar_gpu/kernels/ernie_core/layer_norm_cuda_kernel.cu
+++ b/backends/iluvatar_gpu/kernels/ernie_core/layer_norm_cuda_kernel.cu
@@ -81,8 +81,16 @@ void RMSLnBwd(const Context &ctx,
 
 }  // namespace phi
 
-PD_REGISTER_PLUGIN_KERNEL(
-    fused_rms_norm_ext, iluvatar_gpu, ALL_LAYOUT, phi::RMSLnFwd, float) {}
+PD_REGISTER_PLUGIN_KERNEL(fused_rms_norm_ext,
+                          iluvatar_gpu,
+                          ALL_LAYOUT,
+                          phi::RMSLnFwd,
+                          float,
+                          phi::bfloat16) {}
 
-PD_REGISTER_PLUGIN_KERNEL(
-    fused_rms_norm_ext_grad, iluvatar_gpu, ALL_LAYOUT, phi::RMSLnBwd, float) {}
+PD_REGISTER_PLUGIN_KERNEL(fused_rms_norm_ext_grad,
+                          iluvatar_gpu,
+                          ALL_LAYOUT,
+                          phi::RMSLnBwd,
+                          float,
+                          phi::bfloat16) {}
diff --git a/backends/iluvatar_gpu/tests/disabled_test.txt b/backends/iluvatar_gpu/tests/disabled_test.txt
index 277765189ed..57a868125e8 100644
--- a/backends/iluvatar_gpu/tests/disabled_test.txt
+++ b/backends/iluvatar_gpu/tests/disabled_test.txt
@@ -556,3 +556,4 @@ test_rms_norm_op.py
 test_batched_gemm.py
 test_match_matrix_tensor_op.py
 test_tensor.py
+test_rms_norm.py