Chinese bert (#1429)

* add docs

* add docs

* Update tokenizer.py

try import pypinyin

* update pypinyin

Co-authored-by: yingyibiao <[email protected]>

Author: 27182812

paddlenlp/transformers/__init__.py

@@ -86,3 +86,5 @@
 from .reformer.tokenizer import *
 from .mobilebert.modeling import *
 from .mobilebert.tokenizer import *
+from .chinesebert.modeling import *
+from .chinesebert.tokenizer import *

paddlenlp/transformers/chinesebert/modeling.py

@@ -42,8 +42,6 @@
 from paddlenlp.transformers import PretrainedModel, register_base_model
 from paddlenlp.transformers.bert.modeling import BertPooler, BertPretrainingHeads
 
-# from .fusion_embedding import FusionBertEmbeddings
-
 __all__ = [
     "ChineseBertModel",
     "ChineseBertPretrainedModel",
@@ -55,7 +53,6 @@
 ]
 
 
-# fusion_embedding.py
 class PinyinEmbedding(nn.Layer):
     def __init__(self,
                  pinyin_map_len: int,
@@ -65,6 +62,7 @@ def __init__(self,
         Pinyin Embedding Layer.
 
         Args:
+            pinyin_map_len (int): the size of pinyin map, which about 26 Romanian characters and 6 numbers. 
             embedding_size (int): the size of each embedding vector.
             pinyin_out_dim (int): kernel number of conv.