diff --git a/mindnlp/injection.py b/mindnlp/injection.py
index b19b25cba..2ddafe1d2 100644
--- a/mindnlp/injection.py
+++ b/mindnlp/injection.py
@@ -628,17 +628,13 @@ def __init__(self,
             raise ValueError(f"The argument 'group' should be divisible by 'in_channels' " \
                              f"and 'out_channels', but got group:{group}, in_channels:{in_channels}, " \
                              f"out_channels:{out_channels}.")
-        kernel_size = (kernel_size,)
-        if mindspore.__version__ == '2.0.0':
-            stride = (1, stride,)
-        else:
-            stride = (stride,)
+        stride = (1, stride)
+        dilation = (1, dilation)
 
-        dilation = (dilation,)
         super().__init__(
             in_channels,
             out_channels,
-            kernel_size,
+            (kernel_size,),
             stride,
             pad_mode,
             padding,
@@ -647,12 +643,26 @@ def __init__(self,
             has_bias,
             None,
             None)
-        self.padding = padding
+        self.padding = (0, 0, padding, padding)
+        self.padding = (0, 0, padding, padding)
+        Validator.check_string(pad_mode, ['valid', 'same', 'pad'], 'pad_mode', self.cls_name)
+        self.conv2d = ops.Conv2D(out_channel=self.out_channels,
+                               kernel_size=(1, kernel_size),
+                               mode=1,
+                               pad_mode=self.pad_mode,
+                               pad=self.padding,
+                               stride=self.stride,
+                               dilation=self.dilation,
+                               group=self.group)
 
     def construct(self, x):
-        return ops.conv1d(x, self.weight, self.bias, stride=self.stride, pad_mode=self.pad_mode,
-                          padding=self.padding, dilation=self.dilation, groups=self.group)
+        x = x.expand_dims(2)
+        output = self.conv2d(x, self.weight.expand_dims(2))
+        if self.has_bias:
+            output = ops.bias_add(output, self.bias)
 
+        output = output.squeeze(2)
+        return output
 
 class Conv1dTranspose(_Conv):
     """patched Conv1dTranspose"""
diff --git a/mindnlp/transformers/generation/logits_process.py b/mindnlp/transformers/generation/logits_process.py
index 279b9fbf3..9a51961d5 100644
--- a/mindnlp/transformers/generation/logits_process.py
+++ b/mindnlp/transformers/generation/logits_process.py
@@ -1302,7 +1302,7 @@ def __call__(self, input_ids: mindspore.Tensor, scores: mindspore.Tensor) -> min
         for k in range(input_ids.shape[0]):
             timestamp_logprob = logprobs[k, self.timestamp_begin :].logsumexp(axis=-1)
             max_text_token_logprob = logprobs[k, : self.timestamp_begin].max()
-            if timestamp_logprob > max_text_token_logprob:
+            if not ops.isnan(timestamp_logprob) and timestamp_logprob > max_text_token_logprob:
                 scores[k, : self.timestamp_begin] = -float("inf")
 
         return scores