diff --git a/paddlespeech/t2s/modules/diffnet.py b/paddlespeech/t2s/modules/diffnet.py
index 2f433ad68..98be9a9a3 100644
--- a/paddlespeech/t2s/modules/diffnet.py
+++ b/paddlespeech/t2s/modules/diffnet.py
@@ -121,6 +121,7 @@ class SinusoidalPosEmb(nn.Layer):
 
     def forward(self, x: paddle.Tensor):
         x = paddle.cast(x, 'float32')
+        x = x.unsqueeze(0)
         half_dim = self.dim // 2
         emb = math.log(10000) / (half_dim - 1)
         emb = paddle.exp(paddle.arange(half_dim) * -emb)
diff --git a/paddlespeech/t2s/modules/nets_utils.py b/paddlespeech/t2s/modules/nets_utils.py
index 57c46e3a8..d34c579b3 100644
--- a/paddlespeech/t2s/modules/nets_utils.py
+++ b/paddlespeech/t2s/modules/nets_utils.py
@@ -180,7 +180,7 @@ def make_pad_mask(lengths, xs=None, length_dim=-1):
     """
     if length_dim == 0:
         raise ValueError("length_dim cannot be 0: {}".format(length_dim))
-
+    lengths = lengths.unsqueeze(0)
     bs = paddle.shape(lengths)
     if xs is None:
         maxlen = paddle.cast(lengths.max(), dtype=bs.dtype)