fixed doc for merge

fixed Embedding

mindspore/nn/layer/embedding.py

@@ -44,10 +44,11 @@ class Embedding(Cell):
         dtype (:class:`mindspore.dtype`): Data type of input. Default: mindspore.float32.
 
     Inputs:
-        - **input** (Tensor) - Tensor of shape :math:`(\text{vocab_size})`.
-
+        - **input** (Tensor) - Tensor of shape :math:`(\text{batch_size}, \text{input_length})`. The element of
+        the Tensor should be integer and not larger than vocab_size. else the corresponding embedding vector is zero
+        if larger than vocab_size.
     Outputs:
-        Tensor of shape :math:`(\text{vocab_size}, \text{embedding_size})`.
+        Tensor of shape :math:`(\text{batch_size}, \text{input_length}, \text{embedding_size})`.
 
     Examples:
         >>> net = nn.Embedding(20000, 768,  True)
@@ -61,6 +62,7 @@ class Embedding(Cell):
     def __init__(self, vocab_size, embedding_size, use_one_hot=False, embedding_table='normal', dtype=mstype.float32):
         super(Embedding, self).__init__()
         validator.check_subclass("dtype", dtype, mstype.number_type, self.cls_name)
+        validator.check_value_type('use_one_hot', use_one_hot, [bool], self.cls_name)
         self.vocab_size = vocab_size
         self.embedding_size = embedding_size
         self.use_one_hot = use_one_hot

mindspore/ops/operations/control_ops.py

@@ -144,7 +144,7 @@ class Merge(PrimitiveWithInfer):
     One and only one of the inputs should be selected as the output
 
     Inputs:
-        - **inputs** (Tuple) - The data to be merged. All tuple elements should have same data type.
+        - **inputs** (Union(Tuple, List)) - The data to be merged. All tuple elements should have same data type.
 
     Outputs:
         tuple. Output is tuple(`data`, `output_index`). The `data` has the same shape of `inputs` element.