nn.embeddinglookup note fix

pull/9373/head
yao_yf 4 years ago
parent 0c7ba7a7fa
commit ef2c859715

@ -145,7 +145,9 @@ class EmbeddingLookup(Cell):
Args:
vocab_size (int): Size of the dictionary of embeddings.
embedding_size (int): The size of each embedding vector.
param_init (str): The initialize way of embedding table. Default: 'normal'.
param_init (Union[Tensor, str, Initializer, numbers.Number]): Initializer for the embedding_table.
Refer to class `initializer` for the values of string when a string
is specified. Default: 'normal'.
target (str): Specifies the target where the op is executed. The value must in
['DEVICE', 'CPU']. Default: 'CPU'.
slice_mode (str): The slicing way in semi_auto_parallel/auto_parallel. The value must get through
@ -291,7 +293,9 @@ class MultiFieldEmbeddingLookup(EmbeddingLookup):
vocab_size (int): The size of the dictionary of embeddings.
embedding_size (int): The size of each embedding vector.
field_size (int): The field size of the final outputs.
param_init (str): The initialize way of embedding table. Default: 'normal'.
param_init (Union[Tensor, str, Initializer, numbers.Number]): Initializer for the embedding_table.
Refer to class `initializer` for the values of string when a string
is specified. Default: 'normal'.
target (str): Specifies the target where the op is executed. The value must in
['DEVICE', 'CPU']. Default: 'CPU'.
slice_mode (str): The slicing way in semi_auto_parallel/auto_parallel. The value must get through

Loading…
Cancel
Save