Fix typos (PaddlePaddle#4959)

Author: co63oc

docs/get_started/installation.rst

@@ -49,7 +49,7 @@ Anaconda是一个开源的Python发行版本，其包含了conda、Python等180
 
 按如上方式配置后，即可在环境中使用PaddleNLP了，命令行输入python回车后，import paddlenlp试试吧，之后再次使用都可以通过打开'所有程序->Anaconda3/2（64-bit）->Anaconda Prompt'，再执行conda activate my_paddlenlp进入环境后，即可再次使用PaddleNLP。
 
-2、Linux/Mac安装安装Anaconda
+2、Linux/Mac安装Anaconda
 >>>>>>>>>
 
 第一步 下载

docs/locale/en/LC_MESSAGES/get_started/installation.po

@@ -83,7 +83,7 @@ msgid ""
 msgstr ""
 
 #: ../get_started/installation.rst:53
-msgid "2、Linux/Mac安装安装Anaconda"
+msgid "2、Linux/Mac安装Anaconda"
 msgstr ""
 
 #: ../get_started/installation.rst:57

docs/model_zoo/embeddings.md

@@ -23,7 +23,7 @@
 
 ## 介绍
 
-PaddleNLP提供多个开源的预训练词向量模型，用户仅需在使用`paddlenlp.embeddings.TokenEmbedding`时，指定预训练模型的名称，即可加载相对应的预训练模型。以下将介绍`TokenEmbeddign`详细用法，并列出PaddleNLP所支持的预训练Embedding模型。
+PaddleNLP提供多个开源的预训练词向量模型，用户仅需在使用`paddlenlp.embeddings.TokenEmbedding`时，指定预训练模型的名称，即可加载相对应的预训练模型。以下将介绍`TokenEmbedding`详细用法，并列出PaddleNLP所支持的预训练Embedding模型。
 
 ## 用法
 

examples/text_to_sql/IGSQL/README.md

@@ -80,7 +80,7 @@ python run.py --raw_train_filename="data/sparc_data_removefrom/train.pkl" \
 * embedding_filename: GLOVE 词向量文件路径。
 * data_directory: 预处理得到的文件夹路径。
 * logdir: 输出日志文件夹路径。
-* train，evaluate: 是否执行trian，evaluate。
+* train，evaluate: 是否执行train，evaluate。
 
 
 ### 训练阶段的输出日志

examples/text_to_sql/RAT-SQL/evaluation/README.md

@@ -4,7 +4,7 @@
 # 评估
 输入文件格式：
 1. 文件以.sql结尾
-2. 文件每行的格式："qid\tsql_query\tdb_id",其中predcit文件db_id是可选字段，gold文件db_id是必选字段
+2. 文件每行的格式："qid\tsql_query\tdb_id",其中predict文件db_id是可选字段，gold文件db_id是必选字段
 3. 评估指标：exact matching score
 
 # 使用

examples/text_to_sql/RAT-SQL/text2sql/dataproc/ernie_input_encoder_v2.py

@@ -183,7 +183,7 @@ def tokenize(self, question, db, column_match_cells=None, candi_nums=None, candi
             if match_cells is not None and len(match_cells) > 0:
                 if column.dtype in ("text", "time"):
                     if not self.config.predict_value:
-                        match_cells = match_cells[:1]  # the first cell used to complement senmantics
+                        match_cells = match_cells[:1]  # the first cell used to complement semantics
                     for mcell in match_cells:
                         value_list.append(mcell)
                         toks = [self.special_token_dict["value"]] + self.tokenizer.tokenize(mcell)

examples/text_to_sql/RAT-SQL/text2sql/dataproc/sql_preproc_v2.py

@@ -267,7 +267,7 @@ def add_item(self, section, sql_json, value_list):
         self.format_sql_value(sql_json, value_dict)
 
         parsed = self.grammar.parse(sql_json, section)
-        self.ast_wrapper.verify_ast(parsed)  # will raise AssertionError, if varify failed
+        self.ast_wrapper.verify_ast(parsed)  # will raise AssertionError, if verify failed
 
         root = parsed
         if section == "train":

examples/text_to_sql/RAT-SQL/text2sql/io.py

@@ -32,12 +32,12 @@ def init_ernie_model(model_class, model_dir):
     return ernie, config["hidden_size"]
 
 
-def save(model, optimzer, save_path):
+def save(model, optimizer, save_path):
     try:
         paddle.save(model.state_dict(), save_path + ".pdparams")
-        paddle.save(optimzer.state_dict(), save_path + ".pdopt")
+        paddle.save(optimizer.state_dict(), save_path + ".pdopt")
     except Exception as e:
-        logging.error("save model and optimzer failed. save path: %s", save_path)
+        logging.error("save model and optimizer failed. save path: %s", save_path)
         logging.error(traceback.format_exc())
 
 

examples/text_to_sql/RAT-SQL/text2sql/models/sql_beam_search.py

@@ -31,7 +31,7 @@ class Hypothesis4Filtering(Hypothesis):
 
 def beam_search_with_heuristics(model, inputs, beam_size, max_steps, from_cond=True):
     """
-    Find the valid FROM clasue with beam search
+    Find the valid FROM clause with beam search
     """
     orig_inputs = inputs["orig_inputs"][0]
     # inference_state, next_choices = model.inference(inputs, orig_inputs.db)
@@ -93,7 +93,7 @@ def beam_search_with_heuristics(model, inputs, beam_size, max_steps, from_cond=T
         prefixes2fill_from.sort(key=operator.attrgetter("score"), reverse=True)
         # assert len(prefixes) == beam_size
 
-        # emuerating
+        # enumerating
         beam_from = prefixes2fill_from
         max_size = 6
         unfiltered_finished = []

examples/text_to_sql/RAT-SQL/text2sql/optim.py

@@ -21,7 +21,7 @@
 
 import paddle
 
-param_name_to_exclue_from_weight_decay = re.compile(r".*layer_norm_scale|.*layer_norm_bias|.*b_0")
+param_name_to_exclude_from_weight_decay = re.compile(r".*layer_norm_scale|.*layer_norm_bias|.*b_0")
 
 
 def get_warmup_and_linear_decay(max_steps, warmup_steps):
@@ -43,7 +43,7 @@ def init_optimizer(model, config, train_steps, scale_params_lr=None):
         lr_scheduler,
         parameters=model.parameters(),
         weight_decay=config.weight_decay,
-        apply_decay_param_fun=lambda n: not param_name_to_exclue_from_weight_decay.match(n),
+        apply_decay_param_fun=lambda n: not param_name_to_exclude_from_weight_decay.match(n),
         grad_clip=paddle.nn.ClipGradByGlobalNorm(config.grad_clip),
     )
     return optimizer

examples/text_to_sql/RAT-SQL/text2sql/utils/dusql_evaluation.py

@@ -423,11 +423,11 @@ def parse_from(toks, start_idx, tables_with_alias, schema):
             assert toks[idx] == ")"
             idx += 1
         if idx < len_ and toks[idx] == "a":
-            assert last_table is not None, "last_table should be a table name strin, not None"
+            assert last_table is not None, "last_table should be a table name string, not None"
             tables_with_alias["a"] = last_table
             idx += 2
         elif idx < len_ and toks[idx] == "b":
-            assert last_table is not None, "last_table should be a table name strin, not None"
+            assert last_table is not None, "last_table should be a table name string, not None"
             tables_with_alias["b"] = last_table
             idx += 1
         if idx < len_ and (toks[idx] in CLAUSE_KEYWORDS or toks[idx] in (")", ";")):
@@ -675,7 +675,7 @@ def __eval_from_sql(pred_tables, gold_tables):
         # return 1
 
     def eval_exact_match(self, pred, gold):
-        """wrapper of evaluate examct match, to process
+        """wrapper of evaluate exact match, to process
         `SQL1 intersect/union SQL2` vs `SQL2 intersect/union SQL1`
         """
         score = self._eval_exact_match(pred, gold)

examples/text_to_sql/RAT-SQL/text2sql/utils/nn_utils.py

@@ -53,7 +53,7 @@ def lstm_init(num_layers, hidden_size, *batch_sizes):
 
 
 def batch_gather_2d(var, indices):
-    """Gather slices from var in each batch, according to corrensponding
+    """Gather slices from var in each batch, according to corresponding
     index in indices. Currently, it only support 2d Tensor.
 
     Args:
@@ -156,7 +156,7 @@ def pad_sequences_for_3d(seqs, max_col, max_num, dtype=np.int64):
 
 
 def pad_index_sequences(seqs, max_col, max_row, dtype=np.int64):
-    """padding squences for column token indexs"""
+    """padding sequences for column token indexes"""
     padded = []
     for query in seqs:
         new_cols = []

examples/text_to_sql/RAT-SQL/text2sql/utils/text_utils.py

@@ -191,7 +191,7 @@ def search_values(cls, question, table):
         all_candidate = []
         for col_id in range(len(table.header)):
             header = table.header[col_id]
-            # 提取col出现在quesiton中的cell
+            # 提取col出现在question中的cell
             # TODO 这里存在一个问题，一个text类型cell必须完全在question中出现才会被当做候选cell
             value_in_column = cls.extract_values_from_column(question, table, col_id, header.type)
             if header.type == "text":

examples/text_to_sql/RAT-SQL/text2sql/utils/utils.py

@@ -81,10 +81,10 @@ def count_file_lines(filename):
     return cnt
 
 
-def print_tensors(tag="*", **kwrags):
-    """print tensors for debuging"""
+def print_tensors(tag="*", **kwargs):
+    """print tensors for debugging"""
     print(tag * 50)
-    for key, value in kwrags.items():
+    for key, value in kwargs.items():
         print(key, ":", value)
 
 

examples/word_embedding/train.py

@@ -142,7 +142,7 @@ def forward(self, text, seq_len=None):
     # Loads dataset.
     train_ds, dev_ds = load_dataset("chnsenticorp", splits=["train", "dev"])
 
-    # Constructs the newtork.
+    # Constructs the network.
     model = BoWModel(
         vocab_size=len(vocab),
         num_classes=len(train_ds.label_list),

paddlenlp/ops/einsum.py

@@ -28,7 +28,7 @@ def einsum(equation, *operands):
             Uses uncased letters to specify the dimension of the operands and result. The input
             equation is on the left hand before `->` while the output equation is on the right side.
             Einsum can infer the result shape so that the `->` and the result label letters can be omitted.
-            Operands in the input equation are splited by commas (','), e.g. 'abc,cde' describes two 3D
+            Operands in the input equation are splitted by commas (','), e.g. 'abc,cde' describes two 3D
             operands. The dimensions labeled with same letter should be same or be 1. Ellipsis ('...') can
             be used to specify the broadcast dimensions.
 
@@ -129,14 +129,14 @@ def _mul_sum(left, right, sum_dims):
             is_right_summed_dim = right.shape[i] > 1
             if i in sum_dims_set:
                 if is_left_summed_dim and is_right_summed_dim:
-                    assert left.shape[i] == right.shape[i], "Non-brocast dim should be equal."
+                    assert left.shape[i] == right.shape[i], "Non-broadcast dim should be equal."
                     summed_size *= left.shape[i]
                 elif is_left_summed_dim:
                     left = left.sum(axis=i, keepdim=True)
                 elif is_right_summed_dim:
                     right = right.sum(axis=i, keepdim=True)
             elif is_left_summed_dim and is_right_summed_dim:
-                assert left.shape[i] == right.shape[i], "Non-brocast dim should be equal."
+                assert left.shape[i] == right.shape[i], "Non-broadcast dim should be equal."
                 batch_dims.append(i)
                 batch_size *= left.shape[i]
             elif is_left_summed_dim:
@@ -204,7 +204,7 @@ def _mul_sum(left, right, sum_dims):
         for ch in term:
             if ch == ".":
                 ell_char_count += 1
-                assert ell_char_count <= 3, "The '.' should only exist in one ellispis '...' in term {}".format(term)
+                assert ell_char_count <= 3, "The '.' should only exist in one ellipsis '...' in term {}".format(term)
                 if ell_char_count == 3:
                     if num_ell_idxes == -1:
                         num_ell_idxes = curr_num_ell_idxes
@@ -213,7 +213,7 @@ def _mul_sum(left, right, sum_dims):
                     else:
                         assert (
                             curr_num_ell_idxes == num_ell_idxes
-                        ), "Ellispis in all terms should represent same dimensions ({}).".format(num_ell_idxes)
+                        ), "Ellipsis in all terms should represent same dimensions ({}).".format(num_ell_idxes)
 
                     for j in range(num_ell_idxes):
                         curr_operand_idxes.append(j + first_ell_idx)
@@ -247,11 +247,11 @@ def _mul_sum(left, right, sum_dims):
         for ch in output_eqn:
             if ch == ".":
                 ell_char_count += 1
-                assert ell_char_count <= 3, "The '.' should only exist in one ellispis '...' in term {}".format(
+                assert ell_char_count <= 3, "The '.' should only exist in one ellipsis '...' in term {}".format(
                     output_eqn
                 )
                 if ell_char_count == 3:
-                    assert num_ell_idxes > -1, "Input equation '{}' don't have ellispis.".format(input_eqn)
+                    assert num_ell_idxes > -1, "Input equation '{}' don't have ellipsis.".format(input_eqn)
                     for j in range(num_ell_idxes):
                         idxes_to_output_dims[first_ell_idx + j] = num_output_dims
                         num_output_dims += 1

paddlenlp/prompt/prompt_utils.py

@@ -44,7 +44,7 @@ class PromptDataCollatorWithPadding:
     pad the inputs to the longest sequence in the batch.
 
     Args:
-        tokenizer (`paddlennlp.transformers.PretrainedTokenizer`):
+        tokenizer (`paddlenlp.transformers.PretrainedTokenizer`):
             The tokenizer used for encoding the data from PromptTokenizer.
     """
 

paddlenlp/prompt/template.py

@@ -507,7 +507,7 @@ def parse_soft_prompt(self):
             `Tuple[Dict[int, int], List[List[int]], int]`:
                 - Mapping from continuous ids to word ids for initialization.
                 - Continuous ids for each part. Id 0 denotes none-continuous part.
-                - Number of unique coutinuous tokens.
+                - Number of unique continuous tokens.
         """
         prompt = self._prompt.copy()
         num_soft_token = 1
@@ -560,15 +560,15 @@ def parse_soft_prompt(self):
                     else:
                         soft_id_reindex[part["soft_id"]] = soft_id_list
 
-            # Deal with continous prompt defined by `soft_id`.
+            # Deal with continuous prompt defined by `soft_id`.
             elif "soft_id" in part and part["soft_id"] in soft_id_reindex:
                 soft_id_list = soft_id_reindex[part["soft_id"]]
                 if "length" in part:
                     logger.warning("Ignore `length` because it is incompatible with existing `soft_id`.")
                 soft_token_ids.append(soft_id_list)
                 part_prompt = {"soft": [self.tokenizer.unk_token] * len(soft_id_list)}
 
-            # Deal with continous prompt with random initialization.
+            # Deal with continuous prompt with random initialization.
             else:
                 if "length" not in part:
                     part["length"] = 1

paddlenlp/seq2vec/encoder.py

@@ -260,7 +260,7 @@ def forward(self, inputs, mask=None):
                 Shape as `(batch_size, num_tokens, emb_dim)` and dtype as `float32` or `float64`.
                 Tensor containing the features of the input sequence.
             mask (Tensor, optional):
-                Shape shoule be same as `inputs` and dtype as `int32`, `int64`, `float32` or `float64`.
+                Shape should be same as `inputs` and dtype as `int32`, `int64`, `float32` or `float64`.
                 Its each elements identify whether the corresponding input token is padding or not.
                 If True, not padding token. If False, padding token.
                 Defaults to `None`.
@@ -319,7 +319,7 @@ class GRUEncoder(nn.Layer):
             Defaults to 1.
         direction (str, optional):
             The direction of the network. It can be "forward" and "bidirect"
-            (it means bidirection network). If "bidirect", it is a birectional GRU,
+            (it means bidirection network). If "bidirect", it is a bidirectional GRU,
             and returns the concat output from both directions.
             Defaults to "forward".
         dropout (float, optional):
@@ -504,7 +504,7 @@ class LSTMEncoder(nn.Layer):
             Defaults to 1.
         direction (str, optional):
             The direction of the network. It can be "forward" or "bidirect" (it means bidirection network).
-            If "bidirect", it is a birectional LSTM, and returns the concat output from both directions.
+            If "bidirect", it is a bidirectional LSTM, and returns the concat output from both directions.
             Defaults to "forward".
         dropout (float, optional):
             If non-zero, introduces a Dropout layer on the outputs of each LSTM layer
@@ -688,7 +688,7 @@ class RNNEncoder(nn.Layer):
             Defaults to 1.
         direction (str, optional):
             The direction of the network. It can be "forward" and "bidirect"
-            (it means bidirection network). If "biderect", it is a birectional RNN,
+            (it means bidirection network). If "bidirect", it is a bidirectional RNN,
             and returns the concat output from both directions. Defaults to "forward"
         dropout (float, optional):
             If non-zero, introduces a Dropout layer on the outputs of each RNN layer

paddlenlp/taskflow/dependency_parsing.py

@@ -163,7 +163,7 @@ def __init__(
         else:
             raise ValueError(
                 "The encoding model should be one of \
-                ddparser, ddparser-ernie-1.0 and ddoarser-ernie-gram-zh"
+                ddparser, ddparser-ernie-1.0 and ddparser-ernie-gram-zh"
             )
         self._check_task_files()
         self._construct_vocabs()
@@ -528,9 +528,9 @@ def eisner(scores, mask):
     s_i = np.full_like(scores, float("-inf"))
     # Score for complete span
     s_c = np.full_like(scores, float("-inf"))
-    # Incompelte span position for backtrack
+    # Incomplete span position for backtrack
     p_i = np.zeros((seq_len, seq_len, batch_size), dtype=np.int64)
-    # Compelte span position for backtrack
+    # Complete span position for backtrack
     p_c = np.zeros((seq_len, seq_len, batch_size), dtype=np.int64)
     # Set 0 to s_c.diagonal
     s_c = fill_diagonal(s_c, 0)

paddlenlp/taskflow/models/dependency_parsing_model.py

@@ -217,7 +217,7 @@ def index_sample(x, index):
     arr_index = paddle.arange(start=0, end=len(index), dtype=index.dtype)
     arr_index = paddle.unsqueeze(arr_index, axis=[1, 2])
     arr_index = paddle.expand(arr_index, index.shape)
-    # Genrate new index
+    # Generate new index
     new_index = paddle.concat((arr_index, index), -1)
     new_index = paddle.reshape(new_index, (-1, 2))
     # Get output

paddlenlp/taskflow/models/sentiment_analysis_model.py

@@ -34,7 +34,7 @@ class BoWModel(nn.Layer):
         padding_idx(int, optional): The padding value in the embedding, the padding_idx of embedding value will
             not be updated, the default value is 0.
         hidden_size(int, optional): The output size of linear that after the bow, default value is 128.
-        fc_hidden_size(int, optional): The output size of linear that after the fisrt linear, default value is 96.
+        fc_hidden_size(int, optional): The output size of linear that after the first linear, default value is 96.
     """
 
     def __init__(self, vocab_size, num_classes, emb_dim=128, padding_idx=0, hidden_size=128, fc_hidden_size=96):
@@ -71,15 +71,15 @@ class LSTMModel(nn.Layer):
     which is passed through some feed-forward layers to output a logits (`output_layer`).
     Args:
         vocab_size(int): The vocab size that used to create the embedding.
-        num_class(int):  The num clas of the classifier.
+        num_class(int):  The num class of the classifier.
         emb_dim(int. optional): The size of the embedding, default value is 128.
         padding_idx(int, optional): The padding value in the embedding, the padding_idx of embedding value will
             not be updated, the default value is 0.
-        lstm_hidden_size(int, optional): The output size of the lstm, defalut value 198.
+        lstm_hidden_size(int, optional): The output size of the lstm, default value 198.
         direction(string, optional): The direction of lstm, default value is `forward`.
         lstm_layers(string, optional): The num of lstm layer.
         dropout(float, optional): The dropout rate of lstm.
-        pooling_type(float, optional): The pooling type of lstm. Defalut value is None,
+        pooling_type(float, optional): The pooling type of lstm. Default value is None,
             if `pooling_type` is None, then the LSTMEncoder will return the hidden state of the last time step at last layer as a single vector.
     """