Docs Update (#1432)

huhuiwen99 · web-flow · commit 21e61b5b078c · 2021-12-10T16:33:12.000+08:00
* modify transforner-rst

* modify roformer tokenizer

* delete modifications

* modify chunk

* delete changes

* init layoutlm model

* modify layoutlmmodel

* test

* fix errors

* add layoutlmformaskedlm

* modify tokenizer

* update

* update

* add layoutlm example

* add example

* update

* update

* delete data

* update

* modify preprocess

* update

* modify loss

* modify example

* find error

* fix errors

* fix errors

* update

* update

* update
diff --git a/docs/source/paddlenlp.transformers.chinesebert.modeling.rst b/docs/source/paddlenlp.transformers.chinesebert.modeling.rst
@@ -0,0 +1,7 @@
+modeling
+==================================================
+
+.. automodule:: paddlenlp.transformers.chinesebert.modeling
+   :members:
+   :no-undoc-members:
+   :show-inheritance:
diff --git a/docs/source/paddlenlp.transformers.chinesebert.rst b/docs/source/paddlenlp.transformers.chinesebert.rst
@@ -0,0 +1,14 @@
+chinesebert
+==========================================
+
+.. automodule:: paddlenlp.transformers.chinesebert
+   :members:
+   :no-undoc-members:
+   :show-inheritance:
+
+
+.. toctree::
+   :maxdepth: 4
+
+   paddlenlp.transformers.chinesebert.modeling
+   paddlenlp.transformers.chinesebert.tokenizer
diff --git a/docs/source/paddlenlp.transformers.chinesebert.tokenizer.rst b/docs/source/paddlenlp.transformers.chinesebert.tokenizer.rst
@@ -0,0 +1,7 @@
+tokenizer
+===================================================
+
+.. automodule:: paddlenlp.transformers.chinesebert.tokenizer
+   :members:
+   :no-undoc-members:
+   :show-inheritance:
diff --git a/docs/source/paddlenlp.transformers.layoutlm.modeling.rst b/docs/source/paddlenlp.transformers.layoutlm.modeling.rst
@@ -0,0 +1,7 @@
+modeling
+===============================================
+
+.. automodule:: paddlenlp.transformers.layoutlm.modeling
+   :members:
+   :no-undoc-members:
+   :show-inheritance:
diff --git a/docs/source/paddlenlp.transformers.layoutlm.rst b/docs/source/paddlenlp.transformers.layoutlm.rst
@@ -0,0 +1,14 @@
+layoutlm
+=======================================
+
+.. automodule:: paddlenlp.transformers.layoutlm
+   :members:
+   :no-undoc-members:
+   :show-inheritance:
+
+
+.. toctree::
+   :maxdepth: 4
+
+   paddlenlp.transformers.layoutlm.modeling
+   paddlenlp.transformers.layoutlm.tokenizer
diff --git a/docs/source/paddlenlp.transformers.layoutlm.tokenizer.rst b/docs/source/paddlenlp.transformers.layoutlm.tokenizer.rst
@@ -0,0 +1,7 @@
+tokenizer
+================================================
+
+.. automodule:: paddlenlp.transformers.layoutlm.tokenizer
+   :members:
+   :no-undoc-members:
+   :show-inheritance:
diff --git a/docs/source/paddlenlp.transformers.layoutlmv2.modeling.rst b/docs/source/paddlenlp.transformers.layoutlmv2.modeling.rst
@@ -0,0 +1,7 @@
+modeling
+=================================================
+
+.. automodule:: paddlenlp.transformers.layoutlmv2.modeling
+   :members:
+   :no-undoc-members:
+   :show-inheritance:
diff --git a/docs/source/paddlenlp.transformers.layoutlmv2.rst b/docs/source/paddlenlp.transformers.layoutlmv2.rst
@@ -0,0 +1,14 @@
+layoutlmv2
+=========================================
+
+.. automodule:: paddlenlp.transformers.layoutlmv2
+   :members:
+   :no-undoc-members:
+   :show-inheritance:
+
+
+.. toctree::
+   :maxdepth: 4
+
+   paddlenlp.transformers.layoutlmv2.modeling
+   paddlenlp.transformers.layoutlmv2.tokenizer
diff --git a/docs/source/paddlenlp.transformers.layoutlmv2.tokenizer.rst b/docs/source/paddlenlp.transformers.layoutlmv2.tokenizer.rst
@@ -0,0 +1,7 @@
+tokenizer
+==================================================
+
+.. automodule:: paddlenlp.transformers.layoutlmv2.tokenizer
+   :members:
+   :no-undoc-members:
+   :show-inheritance:
diff --git a/docs/source/paddlenlp.transformers.mobilebert.modeling.rst b/docs/source/paddlenlp.transformers.mobilebert.modeling.rst
@@ -0,0 +1,7 @@
+modeling
+=================================================
+
+.. automodule:: paddlenlp.transformers.mobilebert.modeling
+   :members:
+   :no-undoc-members:
+   :show-inheritance:
diff --git a/docs/source/paddlenlp.transformers.mobilebert.rst b/docs/source/paddlenlp.transformers.mobilebert.rst
@@ -0,0 +1,14 @@
+mobilebert
+=========================================
+
+.. automodule:: paddlenlp.transformers.mobilebert
+   :members:
+   :no-undoc-members:
+   :show-inheritance:
+
+
+.. toctree::
+   :maxdepth: 4
+
+   paddlenlp.transformers.mobilebert.modeling
+   paddlenlp.transformers.mobilebert.tokenizer
diff --git a/docs/source/paddlenlp.transformers.mobilebert.tokenizer.rst b/docs/source/paddlenlp.transformers.mobilebert.tokenizer.rst
@@ -0,0 +1,7 @@
+tokenizer
+==================================================
+
+.. automodule:: paddlenlp.transformers.mobilebert.tokenizer
+   :members:
+   :no-undoc-members:
+   :show-inheritance:
diff --git a/docs/source/paddlenlp.transformers.reformer.modeling.rst b/docs/source/paddlenlp.transformers.reformer.modeling.rst
@@ -0,0 +1,7 @@
+modeling
+===============================================
+
+.. automodule:: paddlenlp.transformers.reformer.modeling
+   :members:
+   :no-undoc-members:
+   :show-inheritance:
diff --git a/docs/source/paddlenlp.transformers.reformer.rst b/docs/source/paddlenlp.transformers.reformer.rst
@@ -0,0 +1,14 @@
+reformer
+=======================================
+
+.. automodule:: paddlenlp.transformers.reformer
+   :members:
+   :no-undoc-members:
+   :show-inheritance:
+
+
+.. toctree::
+   :maxdepth: 4
+
+   paddlenlp.transformers.reformer.modeling
+   paddlenlp.transformers.reformer.tokenizer
diff --git a/docs/source/paddlenlp.transformers.reformer.tokenizer.rst b/docs/source/paddlenlp.transformers.reformer.tokenizer.rst
@@ -0,0 +1,7 @@
+tokenizer
+================================================
+
+.. automodule:: paddlenlp.transformers.reformer.tokenizer
+   :members:
+   :no-undoc-members:
+   :show-inheritance:
diff --git a/docs/source/paddlenlp.transformers.rst b/docs/source/paddlenlp.transformers.rst
@@ -18,6 +18,7 @@ paddlenlp.transformers
    paddlenlp.transformers.bigbird
    paddlenlp.transformers.blenderbot
    paddlenlp.transformers.blenderbot_small
+   paddlenlp.transformers.chinesebert
    paddlenlp.transformers.convbert
    paddlenlp.transformers.ctrl
    paddlenlp.transformers.distilbert
@@ -28,10 +29,14 @@ paddlenlp.transformers
    paddlenlp.transformers.ernie_gen
    paddlenlp.transformers.ernie_gram
    paddlenlp.transformers.gpt
+   paddlenlp.transformers.layoutlm
+   paddlenlp.transformers.layoutlmv2
    paddlenlp.transformers.layoutxlm
    paddlenlp.transformers.mbart
+   paddlenlp.transformers.mobilebert
    paddlenlp.transformers.mpnet
    paddlenlp.transformers.nezha
+   paddlenlp.transformers.reformer
    paddlenlp.transformers.roberta
    paddlenlp.transformers.roformer
    paddlenlp.transformers.skep
diff --git a/paddlenlp/transformers/auto/modeling.py b/paddlenlp/transformers/auto/modeling.py
@@ -38,6 +38,8 @@
     ("BlenderbotSmall", "blenderbot_small"),
     ("Blenderbot", "blenderbot"),
     ("ConvBert", "convbert"),
+    ("MobileBert", "mobilebert"),
+    ("ChineseBert", "chinesebert"),
     ("CTRL", "ctrl"),
     ("DistilBert", "distilbert"),
     ("Electra", "electra"),
@@ -49,11 +51,14 @@
     ("Ernie", "ernie"),
     ("GPT", "gpt"),
     ("LayoutXLM", "layoutxlm"),
+    ("LayoutLMv2", "layoutlmv2"),
+    ("LayoutLM", "layoutlm"),
     ("MBart", "mbart"),
     ("MPNet", "mpnet"),
     ("NeZha", "nezha"),
     ("Roberta", "roberta"),
     ("RoFormer", "roformer"),
+    ("Reformer", "reformer"),
     ("SqueezeBert", "squeezebert"),
     ("T5", "t5"),
     ("TinyBert", "tinybert"),
diff --git a/paddlenlp/transformers/auto/tokenizer.py b/paddlenlp/transformers/auto/tokenizer.py
@@ -30,6 +30,8 @@
     ("BlenderbotSmallTokenizer", "blenderbot_small"),
     ("BlenderbotTokenizer", "blenderbot"),
     ("ConvBertTokenizer", "convbert"),
+    ("MobileBertTokenizer", "mobilebert"),
+    ("ChineseBertTokenizer", "chinesebert"),
     ("CTRLTokenizer", "ctrl"),
     ("DistilBertTokenizer", "distilbert"),
     ("ElectraTokenizer", "electra"),
@@ -40,11 +42,14 @@
     ("ErnieTokenizer", "ernie"),
     ("GPTTokenizer", "gpt"),
     ("LayoutXLMTokenizer", "layoutxlm"),
+    ("LayoutLMv2Tokenizer", "layoutlmv2"),
+    ("LayoutLMTokenizer", "layoutlm"),
     ("MBartTokenizer", "mbart"),
     ("MPNetTokenizer", "mpnet"),
     ("NeZhaTokenizer", "nezha"),
     ("RobertaTokenizer", "roberta"),
     ("RoFormerTokenizer", "roformer"),
+    ("ReformerTokenizer", "reformer"),
     ("SqueezeBertTokenizer", "squeezebert"),
     ("T5Tokenizer", 't5'),
     ("TinyBertTokenizer", "tinybert"),
diff --git a/paddlenlp/transformers/layoutlm/modeling.py b/paddlenlp/transformers/layoutlm/modeling.py
@@ -158,27 +158,22 @@ class LayoutLMPretrainedModel(PretrainedModel):
             "layer_norm_eps": 1e-12,
             "pad_token_id": 0,
             "type_vocab_size": 2,
-            "use_cache": True,
         },
         "layoutlm-large-uncased": {
-            "attention_probs_dropout_prob": 0.1,
-            "hidden_act": "gelu",
-            "hidden_dropout_prob": 0.1,
+            "vocab_size": 30522,
             "hidden_size": 1024,
-            "initializer_range": 0.02,
+            "num_attention_heads": 16,
+            "num_hidden_layers": 24,
             "intermediate_size": 4096,
-            "layer_norm_eps": 1e-12,
+            "hidden_act": "gelu",
+            "hidden_dropout_prob": 0.1,
+            "attention_probs_dropout_prob": 0.1,
             "max_2d_position_embeddings": 1024,
             "max_position_embeddings": 512,
-            "num_attention_heads": 16,
-            "num_hidden_layers": 24,
+            "initializer_range": 0.02,
+            "layer_norm_eps": 1e-12,
             "pad_token_id": 0,
             "type_vocab_size": 2,
-            "output_attentions": False,
-            "output_hidden_states": False,
-            "num_labels": 2,
-            "use_cache": True,
-            "vocab_size": 30522
         }
     }
     resource_files_names = {"model_state": "model_state.pdparams"}
@@ -242,8 +237,23 @@ class LayoutLMModel(LayoutLMPretrainedModel):
             The dropout probability for all fully connected layers in the embeddings and encoder.
         attention_probs_dropout_prob (float):
             The dropout probability for all fully connected layers in the pooler.
+        type_vocab_size (int, optional):
+            The vocabulary size of `token_type_ids`.
+            Defaults to `16`.
         initializer_range (float):
-            The standard deviation of the truncated_normal_initializer for initializing all weight matrices.
+            The standard deviation of the normal initializer.
+            Defaults to 0.02.
+
+            .. note::
+                A normal_initializer initializes weight matrices as normal distributions.
+                See :meth:`LayoutLMPretrainedModel.init_weights()` for how weights are initialized in `LayoutLMModel`.
+
+        pad_token_id (int, optional):
+            The index of padding token in the token vocabulary.
+            Defaults to `0`.
+        pooled_act (str, optional):
+            The non-linear activation function in the pooling layer.
+            Defaults to `"tanh"`.
     """
 
     def __init__(
