fix nn.dropout performance problem

mindspore/nn/layer/basic.py

@@ -112,7 +112,11 @@ class Dropout(Cell):
             return x
 
         shape = self.get_shape(x)
-        keep_prob = self.cast(self.keep_prob, mstype.float32)
+        dtype = P.DType()(x)
+        if _is_float_dtype(dtype):
+            keep_prob = self.cast(self.keep_prob, dtype)
+        else:
+            keep_prob = self.cast(self.keep_prob, mstype.float16)
         output = self.dropout_gen_mask(shape, keep_prob)
         return self.dropout_do_mask(x, output, keep_prob)
 
@@ -256,6 +260,12 @@ def _dtype_check(x_dtype):
     if x_dtype not in [mstype.float32, mstype.float16]:
         raise  TypeError("The input type must be float32 or float16.")
 
+@constexpr
+def _is_float_dtype(dtype):
+    if dtype in [mstype.float32, mstype.float16]:
+        return True
+    return False
+
 class ClipByNorm(Cell):
     r"""
     Clips tensor values to a maximum :math:`L_2`-norm.