reconsitution reader and dataset(#279)

* reconsitution reader and dataset

demo/reading-comprehension/predict.py

@@ -20,20 +20,7 @@ from __future__ import print_function
 
 import argparse
 import ast
-import collections
-import json
-import io
-import math
-import numpy as np
-import os
-import six
-import sys
-import time
-
-import paddle
-import paddle.fluid as fluid
 import paddlehub as hub
-from paddlehub.finetune.task.reading_comprehension_task import write_predictions
 
 hub.common.logger.logger.setLevel("INFO")
 

paddlehub/dataset/__init__.py

@@ -14,7 +14,7 @@
 # limitations under the License.
 
 # NLP Dataset
-from .dataset import InputExample, HubDataset
+from .dataset import InputExample, BaseDataset
 from .chnsenticorp import ChnSentiCorp
 from .msra_ner import MSRA_NER
 from .nlpcc_dbqa import NLPCC_DBQA

paddlehub/dataset/base_cv_dataset.py

@@ -18,15 +18,61 @@ from __future__ import division
 from __future__ import print_function
 
 import os
-
 import numpy as np
 
+from paddlehub.dataset import BaseDataset
 import paddlehub as hub
 from paddlehub.common.downloader import default_downloader
+from paddlehub.common.logger import logger
+
+
+class BaseCVDatast(BaseDataset):
+    def __init__(self,
+                 base_path,
+                 train_list_file=None,
+                 validate_list_file=None,
+                 test_list_file=None,
+                 predict_list_file=None,
+                 label_list_file=None,
+                 label_list=None):
+        super(BaseCVDatast, self).__init__(
+            base_path=base_path,
+            train_file=train_list_file,
+            dev_file=validate_list_file,
+            test_file=test_list_file,
+            predict_file=predict_list_file,
+            label_file=label_list_file,
+            label_list=label_list)
+
+    def _read_file(self, data_path, phase=None):
+        data = []
+        with open(data_path, "r") as file:
+            while True:
+                line = file.readline()
+                if not line:
+                    break
+                line = line.strip()
+                items = line.split(" ")
+                if len(items) > 2:
+                    image_path = " ".join(items[0:-1])
+                else:
+                    image_path = items[0]
+                if not os.path.isabs(image_path):
+                    if self.base_path is not None:
+                        image_path = os.path.join(self.base_path, image_path)
+                label = items[-1]
+                data.append((image_path, label))
+        return data
 
 
+# discarded. please use BaseCVDatast
 class ImageClassificationDataset(object):
     def __init__(self):
+        logger.warning(
+            "ImageClassificationDataset is no longer recommended from PaddleHub v1.5.0, "
+            "please use BaseCVDataset instead of ImageClassificationDataset. "
+            "It's more easy-to-use with more functions and support evaluating test set "
+            "in the end of finetune automatically.")
         self.base_path = None
         self.train_list_file = None
         self.test_list_file = None
@@ -99,12 +145,12 @@ class ImageClassificationDataset(object):
 
     def test_data(self, shuffle=False):
         test_data_path = os.path.join(self.base_path, self.test_list_file)
-        return self._parse_data(test_data_path, shuffle, phase='dev')
+        return self._parse_data(test_data_path, shuffle, phase='test')
 
     def validate_data(self, shuffle=False):
         validate_data_path = os.path.join(self.base_path,
                                           self.validate_list_file)
-        return self._parse_data(validate_data_path, shuffle, phase='test')
+        return self._parse_data(validate_data_path, shuffle, phase='dev')
 
     def get_train_examples(self):
         return self.train_examples

paddlehub/dataset/base_nlp_dataset.py

+# coding:utf-8
+# Copyright (c) 2019  PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License"
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from __future__ import absolute_import
+from __future__ import division
+from __future__ import print_function
+
+import io
+import csv
+
+from paddlehub.dataset import InputExample, BaseDataset
+
+
+class BaseNLPDatast(BaseDataset):
+    def __init__(self,
+                 base_path,
+                 train_file=None,
+                 dev_file=None,
+                 test_file=None,
+                 predict_file=None,
+                 label_file=None,
+                 label_list=None,
+                 train_file_with_head=False,
+                 dev_file_with_head=False,
+                 test_file_with_head=False,
+                 predict_file_with_head=False):
+        super(BaseNLPDatast, self).__init__(
+            base_path=base_path,
+            train_file=train_file,
+            dev_file=dev_file,
+            test_file=test_file,
+            predict_file=predict_file,
+            label_file=label_file,
+            label_list=label_list,
+            train_file_with_head=train_file_with_head,
+            dev_file_with_head=dev_file_with_head,
+            test_file_with_head=test_file_with_head,
+            predict_file_with_head=predict_file_with_head)
+
+    def _read_file(self, input_file, phase=None):
+        """Reads a tab separated value file."""
+        with io.open(input_file, "r", encoding="UTF-8") as file:
+            reader = csv.reader(file, delimiter="\t", quotechar=None)
+            examples = []
+            for (i, line) in enumerate(reader):
+                if i == 0:
+                    ncol = len(line)
+                    if self.if_file_with_head[phase]:
+                        continue
+                if ncol == 1:
+                    if phase != "predict":
+                        example = InputExample(guid=i, text_a=line[0])
+                    else:
+                        raise Exception(
+                            "the %s file: %s only has one column but it is not a predict file"
+                            % (phase, input_file))
+                elif ncol == 2:
+                    example = InputExample(
+                        guid=i, text_a=line[0], label=line[1])
+                elif ncol == 3:
+                    example = InputExample(
+                        guid=i, text_a=line[0], text_b=line[1], label=line[2])
+                else:
+                    raise Exception(
+                        "the %s file: %s has too many columns (should <=3)" %
+                        (phase, input_file))
+                examples.append(example)
+            return examples

paddlehub/dataset/bq.py

@@ -17,76 +17,37 @@ from __future__ import absolute_import
 from __future__ import division
 from __future__ import print_function
 
-from collections import namedtuple
-import io
 import os
-import csv
 
-from paddlehub.dataset import InputExample, HubDataset
-from paddlehub.common.downloader import default_downloader
 from paddlehub.common.dir import DATA_HOME
-from paddlehub.common.logger import logger
+from paddlehub.dataset.base_nlp_dataset import BaseNLPDatast
 
-_DATA_URL = "https://bj.bcebos.com/paddlehub-dataset/bq.tar.gz"
 
-
-class BQ(HubDataset):
+class BQ(BaseNLPDatast):
     def __init__(self):
-        self.dataset_dir = os.path.join(DATA_HOME, "bq")
-        if not os.path.exists(self.dataset_dir):
-            ret, tips, self.dataset_dir = default_downloader.download_file_and_uncompress(
-                url=_DATA_URL, save_path=DATA_HOME, print_progress=True)
-        else:
-            logger.info("Dataset {} already cached.".format(self.dataset_dir))
-
-        self._load_train_examples()
-        self._load_test_examples()
-        self._load_dev_examples()
-
-    def _load_train_examples(self):
-        self.train_file = os.path.join(self.dataset_dir, "train.txt")
-        self.train_examples = self._read_file(self.train_file)
-
-    def _load_dev_examples(self):
-        self.dev_file = os.path.join(self.dataset_dir, "dev.txt")
-        self.dev_examples = self._read_file(self.dev_file)
-
-    def _load_test_examples(self):
-        self.test_file = os.path.join(self.dataset_dir, "test.txt")
-        self.test_examples = self._read_file(self.test_file)
-
-    def get_train_examples(self):
-        return self.train_examples
-
-    def get_dev_examples(self):
-        return self.dev_examples
-
-    def get_test_examples(self):
-        return self.test_examples
-
-    def get_labels(self):
-        return ["0", "1"]
-
-    @property
-    def num_labels(self):
-        """
-        Return the number of labels in the dataset.
-        """
-        return len(self.get_labels())
-
-    def _read_file(self, input_file):
-        """Reads a tab separated value file."""
-        with io.open(input_file, "r", encoding="UTF-8") as file:
-            examples = []
-            for (i, line) in enumerate(file):
-                data = line.strip().split("\t")
-                example = InputExample(
-                    guid=i, label=data[2], text_a=data[0], text_b=data[1])
-                examples.append(example)
-            return examples
+        dataset_dir = os.path.join(DATA_HOME, "bq")
+        base_path = self._download_dataset(
+            dataset_dir,
+            url="https://bj.bcebos.com/paddlehub-dataset/bq.tar.gz")
+        super(BQ, self).__init__(
+            base_path=base_path,
+            train_file="train.txt",
+            dev_file="dev.txt",
+            test_file="test.txt",
+            label_file=None,
+            label_list=["0", "1"],
+        )
 
 
 if __name__ == "__main__":
     ds = BQ()
+    print("first 10 dev")
     for e in ds.get_dev_examples()[:10]:
         print("{}\t{}\t{}\t{}".format(e.guid, e.text_a, e.text_b, e.label))
+    print("first 10 train")
+    for e in ds.get_train_examples()[:10]:
+        print("{}\t{}\t{}\t{}".format(e.guid, e.text_a, e.text_b, e.label))
+    print("first 10 test")
+    for e in ds.get_test_examples()[:10]:
+        print("{}\t{}\t{}\t{}".format(e.guid, e.text_a, e.text_b, e.label))
+    print(ds)

paddlehub/dataset/chnsenticorp.py

@@ -17,72 +17,39 @@ from __future__ import absolute_import
 from __future__ import division
 from __future__ import print_function
 
-from collections import namedtuple
 import codecs
 import os
 import csv
 
-from paddlehub.dataset import InputExample, HubDataset
-from paddlehub.common.downloader import default_downloader
+from paddlehub.dataset import InputExample
 from paddlehub.common.dir import DATA_HOME
-from paddlehub.common.logger import logger
+from paddlehub.dataset.base_nlp_dataset import BaseNLPDatast
 
-_DATA_URL = "https://bj.bcebos.com/paddlehub-dataset/chnsenticorp.tar.gz"
 
-
-class ChnSentiCorp(HubDataset):
+class ChnSentiCorp(BaseNLPDatast):
     """
     ChnSentiCorp (by Tan Songbo at ICT of Chinese Academy of Sciences, and for
     opinion mining)
     """
 
     def __init__(self):
-        self.dataset_dir = os.path.join(DATA_HOME, "chnsenticorp")
-        if not os.path.exists(self.dataset_dir):
-            ret, tips, self.dataset_dir = default_downloader.download_file_and_uncompress(
-                url=_DATA_URL, save_path=DATA_HOME, print_progress=True)
-        else:
-            logger.info("Dataset {} already cached.".format(self.dataset_dir))
-
-        self._load_train_examples()
-        self._load_test_examples()
-        self._load_dev_examples()
-
-    def _load_train_examples(self):
-        self.train_file = os.path.join(self.dataset_dir, "train.tsv")
-        self.train_examples = self._read_tsv(self.train_file)
-
-    def _load_dev_examples(self):
-        self.dev_file = os.path.join(self.dataset_dir, "dev.tsv")
-        self.dev_examples = self._read_tsv(self.dev_file)
-
-    def _load_test_examples(self):
-        self.test_file = os.path.join(self.dataset_dir, "test.tsv")
-        self.test_examples = self._read_tsv(self.test_file)
-
-    def get_train_examples(self):
-        return self.train_examples
-
-    def get_dev_examples(self):
-        return self.dev_examples
-
-    def get_test_examples(self):
-        return self.test_examples
-
-    def get_labels(self):
-        return ["0", "1"]
-
-    @property
-    def num_labels(self):
-        """
-        Return the number of labels in the dataset.
-        """
-        return len(self.get_labels())
-
-    def _read_tsv(self, input_file, quotechar=None):
+        dataset_dir = os.path.join(DATA_HOME, "chnsenticorp")
+        base_path = self._download_dataset(
+            dataset_dir,
+            url="https://bj.bcebos.com/paddlehub-dataset/chnsenticorp.tar.gz")
+        super(ChnSentiCorp, self).__init__(
+            base_path=base_path,
+            train_file="train.tsv",
+            dev_file="dev.tsv",
+            test_file="test.tsv",
+            label_file=None,
+            label_list=["0", "1"],
+        )
+
+    def _read_file(self, input_file, phase=None):
         """Reads a tab separated value file."""
         with codecs.open(input_file, "r", encoding="UTF-8") as f:
-            reader = csv.reader(f, delimiter="\t", quotechar=quotechar)
+            reader = csv.reader(f, delimiter="\t", quotechar=None)
             examples = []
             seq_id = 0
             header = next(reader)  # skip header
@@ -97,5 +64,5 @@ class ChnSentiCorp(HubDataset):
 
 if __name__ == "__main__":
     ds = ChnSentiCorp()
-    for e in ds.get_train_examples():
+    for e in ds.get_train_examples()[:10]:
         print("{}\t{}\t{}\t{}".format(e.guid, e.text_a, e.text_b, e.label))

paddlehub/dataset/cmrc2018.py

@@ -16,12 +16,11 @@
 
 import json
 import os
-import sys
 
 from paddlehub.reader import tokenization
-from paddlehub.common.downloader import default_downloader
 from paddlehub.common.dir import DATA_HOME
 from paddlehub.common.logger import logger
+from paddlehub.dataset.base_nlp_dataset import BaseNLPDatast
 
 _DATA_URL = "https://bj.bcebos.com/paddlehub-dataset/cmrc2018.tar.gz"
 SPIECE_UNDERLINE = '▁'
@@ -63,42 +62,22 @@ class CMRC2018Example(object):
         return s
 
 
-class CMRC2018(object):
+class CMRC2018(BaseNLPDatast):
     """A single set of features of data."""
 
     def __init__(self):
-        self.dataset_dir = os.path.join(DATA_HOME, "cmrc2018")
-        if not os.path.exists(self.dataset_dir):
-            ret, tips, self.dataset_dir = default_downloader.download_file_and_uncompress(
-                url=_DATA_URL, save_path=DATA_HOME, print_progress=True)
-        else:
-            logger.info("Dataset {} already cached.".format(self.dataset_dir))
-
-        self._load_train_examples()
-        self._load_dev_examples()
-        self._load_test_examples()
-
-    def _load_train_examples(self):
-        self.train_file = os.path.join(self.dataset_dir, "cmrc2018_train.json")
-        self.train_examples = self._read_json(self.train_file, is_training=True)
-
-    def _load_dev_examples(self):
-        self.dev_file = os.path.join(self.dataset_dir, "cmrc2018_dev.json")
-        self.dev_examples = self._read_json(self.dev_file, is_training=False)
-
-    def _load_test_examples(self):
-        pass
-
-    def get_train_examples(self):
-        return self.train_examples
-
-    def get_dev_examples(self):
-        return self.dev_examples
-
-    def get_test_examples(self):
-        return []
-
-    def _read_json(self, input_file, is_training=False):
+        dataset_dir = os.path.join(DATA_HOME, "cmrc2018")
+        base_path = self._download_dataset(dataset_dir, url=_DATA_URL)
+        super(CMRC2018, self).__init__(
+            base_path=base_path,
+            train_file="cmrc2018_train.json",
+            dev_file="cmrc2018_dev.json",
+            test_file=None,
+            label_file=None,
+            label_list=None,
+        )
+
+    def _read_file(self, input_file, phase=False):
         """Read a cmrc2018 json file into a list of CRCDExample."""
 
         def _is_chinese_char(cp):
@@ -197,7 +176,7 @@ class CMRC2018(object):
                     #
                     # Note that this means for training mode, every example is NOT
                     # guaranteed to be preserved.
-                    if is_training:
+                    if phase == "train":
                         actual_text = "".join(
                             doc_tokens[start_position:(end_position + 1)])
                         cleaned_answer_text = "".join(

paddlehub/dataset/dataset.py

@@ -17,6 +17,12 @@ from __future__ import absolute_import
 from __future__ import division
 from __future__ import print_function
 
+import os
+
+import paddlehub as hub
+from paddlehub.common.downloader import default_downloader
+from paddlehub.common.logger import logger
+
 
 class InputExample(object):
     """
@@ -49,21 +55,124 @@ class InputExample(object):
                 self.text_a, self.text_b, self.label)
 
 
-class HubDataset(object):
+class BaseDataset(object):
+    def __init__(self,
+                 base_path,
+                 train_file=None,
+                 dev_file=None,
+                 test_file=None,
+                 predict_file=None,
+                 label_file=None,
+                 label_list=None,
+                 train_file_with_head=False,
+                 dev_file_with_head=False,
+                 test_file_with_head=False,
+                 predict_file_with_head=False):
+        if not (train_file or dev_file or test_file):
+            raise ValueError("At least one file should be assigned")
+        self.base_path = base_path
+        self.train_file = train_file
+        self.dev_file = dev_file
+        self.test_file = test_file
+        self.predict_file = predict_file
+        self.label_file = label_file
+        self.label_list = label_list
+
+        self.train_examples = []
+        self.dev_examples = []
+        self.test_examples = []
+        self.predict_examples = []
+
+        self.if_file_with_head = {
+            "train": train_file_with_head,
+            "dev": dev_file_with_head,
+            "test": test_file_with_head,
+            "predict": predict_file_with_head
+        }
+
+        if train_file:
+            self._load_train_examples()
+        if dev_file:
+            self._load_dev_examples()
+        if test_file:
+            self._load_test_examples()
+        if predict_file:
+            self._load_predict_examples()
+        if self.label_file:
+            if not self.label_list:
+                self.label_list = self._load_label_data()
+            else:
+                logger.warning(
+                    "As label_list has been assigned, label_file is noneffective"
+                )
+
     def get_train_examples(self):
-        raise NotImplementedError()
+        return self.train_examples
 
     def get_dev_examples(self):
-        raise NotImplementedError()
+        return self.dev_examples
 
     def get_test_examples(self):
-        raise NotImplementedError()
+        return self.test_examples
 
     def get_val_examples(self):
         return self.get_dev_examples()
 
+    def get_predict_examples(self):
+        return self.predict_examples
+
     def get_labels(self):
-        raise NotImplementedError()
+        return self.label_list
 
+    @property
     def num_labels(self):
-        raise NotImplementedError()
+        return len(self.label_list)
+
+    def label_dict(self):
+        return {index: key for index, key in enumerate(self.label_list)}
+
+    def _download_dataset(self, dataset_path, url):
+        if not os.path.exists(dataset_path):
+            result, tips, dataset_path = default_downloader.download_file_and_uncompress(
+                url=url,
+                save_path=hub.common.dir.DATA_HOME,
+                print_progress=True,
+                replace=True)
+            if not result:
+                raise Exception(tips)
+        else:
+            logger.info("Dataset {} already cached.".format(dataset_path))
+        return dataset_path
+
+    def _load_train_examples(self):
+        self.train_path = os.path.join(self.base_path, self.train_file)
+        self.train_examples = self._read_file(self.train_path, phase="train")
+
+    def _load_dev_examples(self):
+        self.dev_path = os.path.join(self.base_path, self.dev_file)
+        self.dev_examples = self._read_file(self.dev_path, phase="dev")
+
+    def _load_test_examples(self):
+        self.test_path = os.path.join(self.base_path, self.test_file)
+        self.test_examples = self._read_file(self.test_path, phase="test")
+
+    def _load_predict_examples(self):
+        self.predict_path = os.path.join(self.base_path, self.predict_file)
+        self.predict_examples = self._read_file(
+            self.predict_path, phase="predict")
+
+    def _read_file(self, path, phase=None):
+        raise NotImplementedError
+
+    def _load_label_data(self):
+        with open(os.path.join(self.base_path, self.label_file), "r") as file:
+            return file.read().split("\n")
+
+    def __str__(self):
+        return "Dataset: %s with %i train examples, %i dev examples and %i test examples" % (
+            self.__class__.__name__, len(self.train_examples),
+            len(self.dev_examples), len(self.test_examples))
+
+
+# add alias, compatible with old version
+HubDataset = BaseDataset

paddlehub/dataset/dogcat.py

@@ -20,18 +20,33 @@ from __future__ import print_function
 import os
 
 import paddlehub as hub
-from paddlehub.dataset.base_cv_dataset import ImageClassificationDataset
+from paddlehub.dataset.base_cv_dataset import BaseCVDatast
 
 
-class DogCatDataset(ImageClassificationDataset):
+class DogCatDataset(BaseCVDatast):
     def __init__(self):
-        super(DogCatDataset, self).__init__()
         dataset_path = os.path.join(hub.common.dir.DATA_HOME, "dog-cat")
-        self.base_path = self._download_dataset(
+        base_path = self._download_dataset(
             dataset_path=dataset_path,
             url="https://bj.bcebos.com/paddlehub-dataset/dog-cat.tar.gz")
-        self.train_list_file = "train_list.txt"
-        self.test_list_file = "test_list.txt"
-        self.validate_list_file = "validate_list.txt"
-        self.label_list_file = "label_list.txt"
-        self.num_labels = 2
+        super(DogCatDataset, self).__init__(
+            base_path=base_path,
+            train_list_file="train_list.txt",
+            validate_list_file="validate_list.txt",
+            test_list_file="test_list.txt",
+            label_list_file="label_list.txt",
+            label_list=None)
+
+
+if __name__ == "__main__":
+    ds = DogCatDataset()
+    print("first 10 dev")
+    for e in ds.get_dev_examples()[:10]:
+        print(e)
+    print("first 10 train")
+    for e in ds.get_train_examples()[:10]:
+        print(e)
+    print("first 10 test")
+    for e in ds.get_test_examples()[:10]:
+        print(e)
+    print(ds)

paddlehub/dataset/drcd.py

@@ -16,12 +16,11 @@
 
 import json
 import os
-import sys
 
 from paddlehub.reader import tokenization
-from paddlehub.common.downloader import default_downloader
 from paddlehub.common.dir import DATA_HOME
 from paddlehub.common.logger import logger
+from paddlehub.dataset.base_nlp_dataset import BaseNLPDatast
 
 _DATA_URL = "https://bj.bcebos.com/paddlehub-dataset/drcd.tar.gz"
 SPIECE_UNDERLINE = '▁'
@@ -39,8 +38,7 @@ class DRCDExample(object):
                  doc_tokens,
                  orig_answer_text=None,
                  start_position=None,
-                 end_position=None,
-                 is_impossible=False):
+                 end_position=None):
         self.qas_id = qas_id
         self.question_text = question_text
         self.doc_tokens = doc_tokens
@@ -64,43 +62,22 @@ class DRCDExample(object):
         return s
 
 
-class DRCD(object):
+class DRCD(BaseNLPDatast):
     """A single set of features of data."""
 
     def __init__(self):
-        self.dataset_dir = os.path.join(DATA_HOME, "drcd")
-        if not os.path.exists(self.dataset_dir):
-            ret, tips, self.dataset_dir = default_downloader.download_file_and_uncompress(
-                url=_DATA_URL, save_path=DATA_HOME, print_progress=True)
-        else:
-            logger.info("Dataset {} already cached.".format(self.dataset_dir))
-
-        self._load_train_examples()
-        self._load_dev_examples()
-        self._load_test_examples()
-
-    def _load_train_examples(self):
-        self.train_file = os.path.join(self.dataset_dir, "DRCD_training.json")
-        self.train_examples = self._read_json(self.train_file)
-
-    def _load_dev_examples(self):
-        self.dev_file = os.path.join(self.dataset_dir, "DRCD_dev.json")
-        self.dev_examples = self._read_json(self.dev_file)
-
-    def _load_test_examples(self):
-        self.test_file = os.path.join(self.dataset_dir, "DRCD_test.json")
-        self.test_examples = self._read_json(self.test_file)
-
-    def get_train_examples(self):
-        return self.train_examples
-
-    def get_dev_examples(self):
-        return self.dev_examples
-
-    def get_test_examples(self):
-        return self.test_examples
-
-    def _read_json(self, input_file):
+        dataset_dir = os.path.join(DATA_HOME, "drcd")
+        base_path = self._download_dataset(dataset_dir, url=_DATA_URL)
+        super(DRCD, self).__init__(
+            base_path=base_path,
+            train_file="DRCD_training.json",
+            dev_file="DRCD_dev.json",
+            test_file="DRCD_test.json",
+            label_file=None,
+            label_list=None,
+        )
+
+    def _read_file(self, input_file, phase=None):
         """Read a DRCD json file into a list of CRCDExample."""
 
         def _is_chinese_char(cp):

paddlehub/dataset/flowers.py

@@ -20,18 +20,33 @@ from __future__ import print_function
 import os
 
 import paddlehub as hub
-from paddlehub.dataset.base_cv_dataset import ImageClassificationDataset
+from paddlehub.dataset.base_cv_dataset import BaseCVDatast
 
 
-class FlowersDataset(ImageClassificationDataset):
+class FlowersDataset(BaseCVDatast):
     def __init__(self):
-        super(FlowersDataset, self).__init__()
         dataset_path = os.path.join(hub.common.dir.DATA_HOME, "flower_photos")
-        self.base_path = self._download_dataset(
+        base_path = self._download_dataset(
             dataset_path=dataset_path,
             url="https://bj.bcebos.com/paddlehub-dataset/flower_photos.tar.gz")
-        self.train_list_file = "train_list.txt"
-        self.test_list_file = "test_list.txt"
-        self.validate_list_file = "validate_list.txt"
-        self.label_list_file = "label_list.txt"
-        self.num_labels = 5
+        super(FlowersDataset, self).__init__(
+            base_path=base_path,
+            train_list_file="train_list.txt",
+            validate_list_file="validate_list.txt",
+            test_list_file="test_list.txt",
+            label_list_file="label_list.txt",
+            label_list=None)
+
+
+if __name__ == "__main__":
+    ds = FlowersDataset()
+    print("first 10 dev")
+    for e in ds.get_dev_examples()[:10]:
+        print(e)
+    print("first 10 train")
+    for e in ds.get_train_examples()[:10]:
+        print(e)
+    print("first 10 test")
+    for e in ds.get_test_examples()[:10]:
+        print(e)
+    print(ds)

paddlehub/dataset/food101.py

@@ -20,19 +20,33 @@ from __future__ import print_function
 import os
 
 import paddlehub as hub
-from paddlehub.dataset.base_cv_dataset import ImageClassificationDataset
+from paddlehub.dataset.base_cv_dataset import BaseCVDatast
 
 
-class Food101Dataset(ImageClassificationDataset):
+class Food101Dataset(BaseCVDatast):
     def __init__(self):
-        super(Food101Dataset, self).__init__()
         dataset_path = os.path.join(hub.common.dir.DATA_HOME, "food-101",
                                     "images")
-        self.base_path = self._download_dataset(
+        base_path = self._download_dataset(
             dataset_path=dataset_path,
             url="https://bj.bcebos.com/paddlehub-dataset/Food101.tar.gz")
-        self.train_list_file = "train_list.txt"
-        self.test_list_file = "test_list.txt"
-        self.validate_list_file = "validate_list.txt"
-        self.label_list_file = "label_list.txt"
-        self.num_labels = 101
+        super(Food101Dataset, self).__init__(
+            base_path=base_path,
+            train_list_file="train_list.txt",
+            test_list_file="test_list.txt",
+            validate_list_file="validate_list.txt",
+            label_list_file="label_list.txt")
+
+
+if __name__ == "__main__":
+    ds = Food101Dataset()
+    print("first 10 dev")
+    for e in ds.get_dev_examples()[:10]:
+        print(e)
+    print("first 10 train")
+    for e in ds.get_train_examples()[:10]:
+        print(e)
+    print("first 10 test")
+    for e in ds.get_test_examples()[:10]:
+        print(e)
+    print(ds)

paddlehub/dataset/glue.py

@@ -21,15 +21,15 @@ import os
 import csv
 import io
 
-from paddlehub.dataset import InputExample, HubDataset
-from paddlehub.common.downloader import default_downloader
-from paddlehub.common.dir import DATA_HOME
+from paddlehub.dataset import InputExample
 from paddlehub.common.logger import logger
+from paddlehub.common.dir import DATA_HOME
+from paddlehub.dataset.base_nlp_dataset import BaseNLPDatast
 
 _DATA_URL = "https://bj.bcebos.com/paddlehub-dataset/glue_data.tar.gz"
 
 
-class GLUE(HubDataset):
+class GLUE(BaseNLPDatast):
     """
     Please refer to
     https://gluebenchmark.com
@@ -43,147 +43,107 @@ class GLUE(HubDataset):
                 'RTE', 'SST-2', 'STS-B'
         ]:
             raise Exception(
-                sub_dataset +
-                " is not in GLUE benchmark. Please confirm the data set")
-        self.mismatch = False
+                "%s is not in GLUE benchmark. Please confirm the data set" %
+                sub_dataset)
+
+        mismatch = False
         if sub_dataset == 'MNLI_mm':
             sub_dataset = 'MNLI'
-            self.mismatch = True
+            mismatch = True
         elif sub_dataset == 'MNLI_m':
             sub_dataset = 'MNLI'
         self.sub_dataset = sub_dataset
-        self.dataset_dir = os.path.join(DATA_HOME, "glue_data")
-
-        if not os.path.exists(self.dataset_dir):
-            ret, tips, self.dataset_dir = default_downloader.download_file_and_uncompress(
-                url=_DATA_URL, save_path=DATA_HOME, print_progress=True)
-        else:
-            logger.info("Dataset {} already cached.".format(self.dataset_dir))
-
-        self._load_train_examples()
-        self._load_dev_examples()
-        self._load_test_examples()
-        self._load_predict_examples()
-
-    def _load_train_examples(self):
-        self.train_file = os.path.join(self.dataset_dir, self.sub_dataset,
-                                       "train.tsv")
-        self.train_examples = self._read_tsv(self.train_file)
-
-    def _load_dev_examples(self):
-        if self.sub_dataset == 'MNLI' and not self.mismatch:
-            self.dev_file = os.path.join(self.dataset_dir, self.sub_dataset,
-                                         "dev_matched.tsv")
-        elif self.sub_dataset == 'MNLI' and self.mismatch:
-            self.dev_file = os.path.join(self.dataset_dir, self.sub_dataset,
-                                         "dev_mismatched.tsv")
-        else:
-            self.dev_file = os.path.join(self.dataset_dir, self.sub_dataset,
-                                         "dev.tsv")
-        self.dev_examples = self._read_tsv(self.dev_file)
-
-    def _load_test_examples(self):
-        self.test_examples = []
-
-    def _load_predict_examples(self):
-        if self.sub_dataset == 'MNLI' and not self.mismatch:
-            self.predict_file = os.path.join(self.dataset_dir, self.sub_dataset,
-                                             "test_matched.tsv")
-        elif self.sub_dataset == 'MNLI' and self.mismatch:
-            self.predict_file = os.path.join(self.dataset_dir, self.sub_dataset,
-                                             "test_mismatched.tsv")
-        else:
-            self.predict_file = os.path.join(self.dataset_dir, self.sub_dataset,
-                                             "test.tsv")
-        self.predict_examples = self._read_tsv(self.predict_file, wo_label=True)
-
-    def get_train_examples(self):
-        return self.train_examples
-
-    def get_dev_examples(self):
-        return self.dev_examples
-
-    def get_test_examples(self):
-        return self.test_examples
-
-    def get_predict_examples(self):
-        return self.predict_examples
-
-    def get_labels(self):
-        """See base class."""
-        if self.sub_dataset in ['MRPC', 'QQP', 'SST-2', 'CoLA']:
-            return ["0", "1"]
-        elif self.sub_dataset in ['QNLI', 'RTE']:
-            return ['not_entailment', 'entailment']
-        elif self.sub_dataset in ['MNLI']:
-            return ["neutral", "contradiction", "entailment"]
-        elif self.sub_dataset in ['STS-B']:
-            return Exception("No category labels for regreesion tasks")
-
-    @property
-    def num_labels(self):
-        """
-        Return the number of labels in the dataset.
-        """
-        return len(self.get_labels())
-
-    def _read_tsv(self, input_file, quotechar=None, wo_label=False):
+
+        # test.tsv has not label,so it is a predict file
+        dev_file = "dev.tsv"
+        predict_file = "test.tsv"
+        if sub_dataset == 'MNLI' and not mismatch:
+            dev_file = 'dev_matched.tsv'
+            predict_file = "test_matched.tsv"
+        elif sub_dataset == 'MNLI' and mismatch:
+            dev_file = 'dev_mismatched.tsv'
+            predict_file = "test_mismatched.tsv"
+
+        dataset_dir = os.path.join(DATA_HOME, "glue_data")
+        dataset_dir = self._download_dataset(dataset_dir, url=_DATA_URL)
+        base_path = os.path.join(dataset_dir, self.sub_dataset)
+
+        label_list = None
+        if sub_dataset in ['MRPC', 'QQP', 'SST-2', 'CoLA']:
+            label_list = ["0", "1"]
+        elif sub_dataset in ['QNLI', 'RTE']:
+            label_list = ['not_entailment', 'entailment']
+        elif sub_dataset in ['MNLI']:
+            label_list = ["neutral", "contradiction", "entailment"]
+        elif sub_dataset in ['STS-B']:
+            label_list = None
+
+        super(GLUE, self).__init__(
+            base_path=base_path,
+            train_file="train.tsv",
+            dev_file=dev_file,
+            predict_file=predict_file,
+            label_file=None,
+            label_list=label_list,
+        )
+
+    def _read_file(self, input_file, phase=None):
         """Reads a tab separated value file."""
         with io.open(input_file, "r", encoding="UTF-8") as f:
-            reader = csv.reader(f, delimiter="\t", quotechar=quotechar)
+            reader = csv.reader(f, delimiter="\t", quotechar=None)
             examples = []
             seq_id = 0
-            if self.sub_dataset != 'CoLA' or wo_label:
+            if self.sub_dataset != 'CoLA' or phase == "predict":
                 header = next(reader)  # skip header
             if self.sub_dataset in [
                     'MRPC',
             ]:
-                if wo_label:
+                if phase == "predict":
                     label_index, text_a_index, text_b_index = [None, -2, -1]
                 else:
                     label_index, text_a_index, text_b_index = [0, -2, -1]
             elif self.sub_dataset in [
                     'QNLI',
             ]:
-                if wo_label:
+                if phase == "predict":
                     label_index, text_a_index, text_b_index = [None, 1, 2]
                 else:
                     label_index, text_a_index, text_b_index = [3, 1, 2]
             elif self.sub_dataset in [
                     'QQP',
             ]:
-                if wo_label:
+                if phase == "predict":
                     label_index, text_a_index, text_b_index = [None, 1, 2]
                 else:
                     label_index, text_a_index, text_b_index = [5, 3, 4]
             elif self.sub_dataset in [
                     'RTE',
             ]:
-                if wo_label:
+                if phase == "predict":
                     label_index, text_a_index, text_b_index = [None, 1, 2]
                 else:
                     label_index, text_a_index, text_b_index = [3, 1, 2]
             elif self.sub_dataset in [
                     'SST-2',
             ]:
-                if wo_label:
+                if phase == "predict":
                     label_index, text_a_index, text_b_index = [None, 1, None]
                 else:
                     label_index, text_a_index, text_b_index = [1, 0, None]
             elif self.sub_dataset in [
                     'MNLI',
             ]:
-                if wo_label:
+                if phase == "predict":
                     label_index, text_a_index, text_b_index = [None, 8, 9]
                 else:
                     label_index, text_a_index, text_b_index = [-1, 8, 9]
             elif self.sub_dataset in ['CoLA']:
-                if wo_label:
+                if phase == "predict":
                     label_index, text_a_index, text_b_index = [None, 1, None]
                 else:
                     label_index, text_a_index, text_b_index = [1, 3, None]
             elif self.sub_dataset in ['STS-B']:
-                if wo_label:
+                if phase == "predict":
                     label_index, text_a_index, text_b_index = [None, -2, -1]
                 else:
                     label_index, text_a_index, text_b_index = [-1, -3, -2]

paddlehub/dataset/iflytek.py

@@ -17,64 +17,30 @@ from __future__ import absolute_import
 from __future__ import division
 from __future__ import print_function
 
-from collections import namedtuple
 import io
 import os
-import csv
 
-from paddlehub.dataset import InputExample, HubDataset
-from paddlehub.common.downloader import default_downloader
+from paddlehub.dataset import InputExample
 from paddlehub.common.dir import DATA_HOME
-from paddlehub.common.logger import logger
+from paddlehub.dataset.base_nlp_dataset import BaseNLPDatast
 
 _DATA_URL = "https://bj.bcebos.com/paddlehub-dataset/iflytek.tar.gz"
 
 
-class IFLYTEK(HubDataset):
+class IFLYTEK(BaseNLPDatast):
     def __init__(self):
-        self.dataset_dir = os.path.join(DATA_HOME, "iflytek")
-        if not os.path.exists(self.dataset_dir):
-            ret, tips, self.dataset_dir = default_downloader.download_file_and_uncompress(
-                url=_DATA_URL, save_path=DATA_HOME, print_progress=True)
-        else:
-            logger.info("Dataset {} already cached.".format(self.dataset_dir))
-
-        self._load_train_examples()
-        self._load_test_examples()
-        self._load_dev_examples()
-
-    def _load_train_examples(self):
-        self.train_file = os.path.join(self.dataset_dir, "train.txt")
-        self.train_examples = self._read_file(self.train_file)
-
-    def _load_dev_examples(self):
-        self.dev_file = os.path.join(self.dataset_dir, "dev.txt")
-        self.dev_examples = self._read_file(self.dev_file)
-
-    def _load_test_examples(self):
-        self.test_file = os.path.join(self.dataset_dir, "test.txt")
-        self.test_examples = self._read_file(self.test_file)
-
-    def get_train_examples(self):
-        return self.train_examples
-
-    def get_dev_examples(self):
-        return self.dev_examples
-
-    def get_test_examples(self):
-        return self.test_examples
-
-    def get_labels(self):
-        return [str(i) for i in range(119)]
-
-    @property
-    def num_labels(self):
-        """
-        Return the number of labels in the dataset.
-        """
-        return len(self.get_labels())
-
-    def _read_file(self, input_file):
+        dataset_dir = os.path.join(DATA_HOME, "iflytek")
+        base_path = self._download_dataset(dataset_dir, url=_DATA_URL)
+        super(IFLYTEK, self).__init__(
+            base_path=base_path,
+            train_file="train.txt",
+            dev_file="dev.txt",
+            test_file="test.txt",
+            label_file=None,
+            label_list=[str(i) for i in range(119)],
+        )
+
+    def _read_file(self, input_file, phase=None):
         """Reads a tab separated value file."""
         with io.open(input_file, "r", encoding="UTF-8") as file:
             examples = []
@@ -91,5 +57,13 @@ class IFLYTEK(HubDataset):
 
 if __name__ == "__main__":
     ds = IFLYTEK()
+    print("first 10 dev")
+    for e in ds.get_dev_examples()[:10]:
+        print("{}\t{}\t{}\t{}".format(e.guid, e.text_a, e.text_b, e.label))
+    print("first 10 train")
     for e in ds.get_train_examples()[:10]:
         print("{}\t{}\t{}\t{}".format(e.guid, e.text_a, e.text_b, e.label))
+    print("first 10 test")
+    for e in ds.get_test_examples()[:10]:
+        print("{}\t{}\t{}\t{}".format(e.guid, e.text_a, e.text_b, e.label))
+    print(ds)

paddlehub/dataset/indoor67.py

@@ -20,18 +20,33 @@ from __future__ import print_function
 import os
 
 import paddlehub as hub
-from paddlehub.dataset.base_cv_dataset import ImageClassificationDataset
+from paddlehub.dataset.base_cv_dataset import BaseCVDatast
 
 
-class Indoor67Dataset(ImageClassificationDataset):
+class Indoor67Dataset(BaseCVDatast):
     def __init__(self):
-        super(Indoor67Dataset, self).__init__()
         dataset_path = os.path.join(hub.common.dir.DATA_HOME, "Indoor67")
-        self.base_path = self._download_dataset(
+        base_path = self._download_dataset(
             dataset_path=dataset_path,
             url="https://bj.bcebos.com/paddlehub-dataset/Indoor67.tar.gz")
-        self.train_list_file = "train_list.txt"
-        self.test_list_file = "test_list.txt"
-        self.validate_list_file = "validate_list.txt"
-        self.label_list_file = "label_list.txt"
-        self.num_labels = 67
+        super(Indoor67Dataset, self).__init__(
+            base_path=base_path,
+            train_list_file="train_list.txt",
+            validate_list_file="validate_list.txt",
+            test_list_file="test_list.txt",
+            label_list_file="label_list.txt",
+            label_list=None)
+
+
+if __name__ == "__main__":
+    ds = Indoor67Dataset()
+    print("first 10 dev")
+    for e in ds.get_dev_examples()[:10]:
+        print(e)
+    print("first 10 train")
+    for e in ds.get_train_examples()[:10]:
+        print(e)
+    print("first 10 test")
+    for e in ds.get_test_examples()[:10]:
+        print(e)
+    print(ds)

paddlehub/dataset/inews.py

@@ -17,73 +17,40 @@ from __future__ import absolute_import
 from __future__ import division
 from __future__ import print_function
 
-from collections import namedtuple
 import io
 import os
 import csv
 
-from paddlehub.dataset import InputExample, HubDataset
-from paddlehub.common.downloader import default_downloader
+from paddlehub.dataset import InputExample
 from paddlehub.common.dir import DATA_HOME
-from paddlehub.common.logger import logger
+from paddlehub.dataset.base_nlp_dataset import BaseNLPDatast
 
 _DATA_URL = "https://bj.bcebos.com/paddlehub-dataset/inews.tar.gz"
 
 
-class INews(HubDataset):
+class INews(BaseNLPDatast):
     """
     INews is a sentiment analysis dataset for Internet News
     """
 
     def __init__(self):
-        self.dataset_dir = os.path.join(DATA_HOME, "inews")
-        if not os.path.exists(self.dataset_dir):
-            ret, tips, self.dataset_dir = default_downloader.download_file_and_uncompress(
-                url=_DATA_URL, save_path=DATA_HOME, print_progress=True)
-        else:
-            logger.info("Dataset {} already cached.".format(self.dataset_dir))
-
-        self._load_train_examples()
-        self._load_test_examples()
-        self._load_dev_examples()
-
-    def _load_train_examples(self):
-        self.train_file = os.path.join(self.dataset_dir, "train.txt")
-        self.train_examples = self._read_file(self.train_file, is_training=True)
-
-    def _load_dev_examples(self):
-        self.dev_file = os.path.join(self.dataset_dir, "dev.txt")
-        self.dev_examples = self._read_file(self.dev_file, is_training=False)
-
-    def _load_test_examples(self):
-        self.test_file = os.path.join(self.dataset_dir, "test.txt")
-        self.test_examples = self._read_file(self.test_file, is_training=False)
-
-    def get_train_examples(self):
-        return self.train_examples
-
-    def get_dev_examples(self):
-        return self.dev_examples
-
-    def get_test_examples(self):
-        return self.test_examples
-
-    def get_labels(self):
-        return ["0", "1", "2"]
-
-    @property
-    def num_labels(self):
-        """
-        Return the number of labels in the dataset.
-        """
-        return len(self.get_labels())
-
-    def _read_file(self, input_file, is_training):
+        dataset_dir = os.path.join(DATA_HOME, "inews")
+        base_path = self._download_dataset(dataset_dir, url=_DATA_URL)
+        super(INews, self).__init__(
+            base_path=base_path,
+            train_file="train.txt",
+            dev_file="dev.txt",
+            test_file="test.txt",
+            label_file=None,
+            label_list=["0", "1", "2"],
+        )
+
+    def _read_file(self, input_file, phase=None):
         """Reads a tab separated value file."""
         with io.open(input_file, "r", encoding="UTF-8") as file:
             examples = []
             for (i, line) in enumerate(file):
-                if i == 0 and is_training:
+                if i == 0 and phase == 'train':
                     continue
                 data = line.strip().split("_!_")
                 example = InputExample(
@@ -94,5 +61,13 @@ class INews(HubDataset):
 
 if __name__ == "__main__":
     ds = INews()
+    print("first 10 dev")
+    for e in ds.get_dev_examples()[:10]:
+        print("{}\t{}\t{}\t{}".format(e.guid, e.text_a, e.text_b, e.label))
+    print("first 10 train")
     for e in ds.get_train_examples()[:10]:
         print("{}\t{}\t{}\t{}".format(e.guid, e.text_a, e.text_b, e.label))
+    print("first 10 test")
+    for e in ds.get_test_examples()[:10]:
+        print("{}\t{}\t{}\t{}".format(e.guid, e.text_a, e.text_b, e.label))
+    print(ds)

paddlehub/dataset/lcqmc.py

@@ -17,68 +17,34 @@ from __future__ import absolute_import
 from __future__ import division
 from __future__ import print_function
 
-from collections import namedtuple
 import codecs
 import os
 import csv
 
-from paddlehub.dataset import InputExample, HubDataset
-from paddlehub.common.downloader import default_downloader
+from paddlehub.dataset import InputExample
 from paddlehub.common.dir import DATA_HOME
-from paddlehub.common.logger import logger
+from paddlehub.dataset.base_nlp_dataset import BaseNLPDatast
 
 _DATA_URL = "https://bj.bcebos.com/paddlehub-dataset/lcqmc.tar.gz"
 
 
-class LCQMC(HubDataset):
+class LCQMC(BaseNLPDatast):
     def __init__(self):
-        self.dataset_dir = os.path.join(DATA_HOME, "lcqmc")
-        if not os.path.exists(self.dataset_dir):
-            ret, tips, self.dataset_dir = default_downloader.download_file_and_uncompress(
-                url=_DATA_URL, save_path=DATA_HOME, print_progress=True)
-        else:
-            logger.info("Dataset {} already cached.".format(self.dataset_dir))
-
-        self._load_train_examples()
-        self._load_test_examples()
-        self._load_dev_examples()
-
-    def _load_train_examples(self):
-        self.train_file = os.path.join(self.dataset_dir, "train.tsv")
-        self.train_examples = self._read_tsv(self.train_file)
-
-    def _load_dev_examples(self):
-        self.dev_file = os.path.join(self.dataset_dir, "dev.tsv")
-        self.dev_examples = self._read_tsv(self.dev_file)
-
-    def _load_test_examples(self):
-        self.test_file = os.path.join(self.dataset_dir, "test.tsv")
-        self.test_examples = self._read_tsv(self.test_file)
-
-    def get_train_examples(self):
-        return self.train_examples
-
-    def get_dev_examples(self):
-        return self.dev_examples
-
-    def get_test_examples(self):
-        return self.test_examples
-
-    def get_labels(self):
-        """See base class."""
-        return ["0", "1"]
-
-    @property
-    def num_labels(self):
-        """
-        Return the number of labels in the dataset.
-        """
-        return len(self.get_labels())
-
-    def _read_tsv(self, input_file, quotechar=None):
+        dataset_dir = os.path.join(DATA_HOME, "lcqmc")
+        base_path = self._download_dataset(dataset_dir, url=_DATA_URL)
+        super(LCQMC, self).__init__(
+            base_path=base_path,
+            train_file="train.tsv",
+            dev_file="dev.tsv",
+            test_file="test.tsv",
+            label_file=None,
+            label_list=["0", "1"],
+        )
+
+    def _read_file(self, input_file, phase=None):
         """Reads a tab separated value file."""
         with codecs.open(input_file, "r", encoding="UTF-8") as f:
-            reader = csv.reader(f, delimiter="\t", quotechar=quotechar)
+            reader = csv.reader(f, delimiter="\t", quotechar=None)
             examples = []
             seq_id = 0
             header = next(reader)  # skip header
@@ -93,5 +59,13 @@ class LCQMC(HubDataset):
 
 if __name__ == "__main__":
     ds = LCQMC()
-    for e in ds.get_train_examples():
+    print("first 10 dev")
+    for e in ds.get_dev_examples()[:10]:
+        print("{}\t{}\t{}\t{}".format(e.guid, e.text_a, e.text_b, e.label))
+    print("first 10 train")
+    for e in ds.get_train_examples()[:10]:
+        print("{}\t{}\t{}\t{}".format(e.guid, e.text_a, e.text_b, e.label))
+    print("first 10 test")
+    for e in ds.get_test_examples()[:10]:
         print("{}\t{}\t{}\t{}".format(e.guid, e.text_a, e.text_b, e.label))
+    print(ds)

paddlehub/dataset/msra_ner.py

@@ -20,18 +20,15 @@ from __future__ import print_function
 import os
 import codecs
 import csv
-import json
-from collections import namedtuple
 
-from paddlehub.dataset import InputExample, HubDataset
-from paddlehub.common.downloader import default_downloader
+from paddlehub.dataset import InputExample
 from paddlehub.common.dir import DATA_HOME
-from paddlehub.common.logger import logger
+from paddlehub.dataset.base_nlp_dataset import BaseNLPDatast
 
 _DATA_URL = "https://bj.bcebos.com/paddlehub-dataset/msra_ner.tar.gz"
 
 
-class MSRA_NER(HubDataset):
+class MSRA_NER(BaseNLPDatast):
     """
     A set of manually annotated Chinese word-segmentation data and
     specifications for training and testing a Chinese word-segmentation system
@@ -40,55 +37,23 @@ class MSRA_NER(HubDataset):
     """
 
     def __init__(self):
-        self.dataset_dir = os.path.join(DATA_HOME, "msra_ner")
-        if not os.path.exists(self.dataset_dir):
-            ret, tips, self.dataset_dir = default_downloader.download_file_and_uncompress(
-                url=_DATA_URL, save_path=DATA_HOME, print_progress=True)
-        else:
-            logger.info("Dataset {} already cached.".format(self.dataset_dir))
-
-        self._load_train_examples()
-        self._load_test_examples()
-        self._load_dev_examples()
-
-    def _load_train_examples(self):
-        train_file = os.path.join(self.dataset_dir, "train.tsv")
-        self.train_examples = self._read_tsv(train_file)
-
-    def _load_dev_examples(self):
-        self.dev_file = os.path.join(self.dataset_dir, "dev.tsv")
-        self.dev_examples = self._read_tsv(self.dev_file)
-
-    def _load_test_examples(self):
-        self.test_file = os.path.join(self.dataset_dir, "test.tsv")
-        self.test_examples = self._read_tsv(self.test_file)
-
-    def get_train_examples(self):
-        return self.train_examples
-
-    def get_dev_examples(self):
-        return self.dev_examples
-
-    def get_test_examples(self):
-        return self.test_examples
-
-    def get_labels(self):
-        return ["B-PER", "I-PER", "B-ORG", "I-ORG", "B-LOC", "I-LOC", "O"]
-
-    @property
-    def num_labels(self):
-        """
-        Return the number of labels in the dataset.
-        """
-        return len(self.get_labels())
-
-    def get_label_map(self):
-        return self.label_map
-
-    def _read_tsv(self, input_file, quotechar=None):
+        dataset_dir = os.path.join(DATA_HOME, "msra_ner")
+        base_path = self._download_dataset(dataset_dir, url=_DATA_URL)
+        super(MSRA_NER, self).__init__(
+            base_path=base_path,
+            train_file="train.tsv",
+            dev_file="dev.tsv",
+            test_file="test.tsv",
+            label_file=None,
+            label_list=[
+                "B-PER", "I-PER", "B-ORG", "I-ORG", "B-LOC", "I-LOC", "O"
+            ],
+        )
+
+    def _read_file(self, input_file, phase=None):
         """Reads a tab separated value file."""
         with codecs.open(input_file, "r", encoding="UTF-8") as f:
-            reader = csv.reader(f, delimiter="\t", quotechar=quotechar)
+            reader = csv.reader(f, delimiter="\t", quotechar=None)
             examples = []
             seq_id = 0
             header = next(reader)  # skip header
@@ -103,5 +68,13 @@ class MSRA_NER(HubDataset):
 
 if __name__ == "__main__":
     ds = MSRA_NER()
-    for e in ds.get_train_examples():
+    print("first 10 dev")
+    for e in ds.get_dev_examples()[:10]:
+        print("{}\t{}\t{}\t{}".format(e.guid, e.text_a, e.text_b, e.label))
+    print("first 10 train")
+    for e in ds.get_train_examples()[:10]:
+        print("{}\t{}\t{}\t{}".format(e.guid, e.text_a, e.text_b, e.label))
+    print("first 10 test")
+    for e in ds.get_test_examples()[:10]:
         print("{}\t{}\t{}\t{}".format(e.guid, e.text_a, e.text_b, e.label))
+    print(ds)

paddlehub/dataset/nlpcc_dbqa.py

@@ -17,20 +17,18 @@ from __future__ import absolute_import
 from __future__ import division
 from __future__ import print_function
 
-from collections import namedtuple
 import codecs
 import os
 import csv
 
-from paddlehub.dataset import InputExample, HubDataset
-from paddlehub.common.downloader import default_downloader
+from paddlehub.dataset import InputExample
 from paddlehub.common.dir import DATA_HOME
-from paddlehub.common.logger import logger
+from paddlehub.dataset.base_nlp_dataset import BaseNLPDatast
 
 _DATA_URL = "https://bj.bcebos.com/paddlehub-dataset/nlpcc-dbqa.tar.gz"
 
 
-class NLPCC_DBQA(HubDataset):
+class NLPCC_DBQA(BaseNLPDatast):
     """
     Please refer to
     http://tcci.ccf.org.cn/conference/2017/dldoc/taskgline05.pdf
@@ -38,53 +36,21 @@ class NLPCC_DBQA(HubDataset):
     """
 
     def __init__(self):
-        self.dataset_dir = os.path.join(DATA_HOME, "nlpcc-dbqa")
-        if not os.path.exists(self.dataset_dir):
-            ret, tips, self.dataset_dir = default_downloader.download_file_and_uncompress(
-                url=_DATA_URL, save_path=DATA_HOME, print_progress=True)
-        else:
-            logger.info("Dataset {} already cached.".format(self.dataset_dir))
-
-        self._load_train_examples()
-        self._load_test_examples()
-        self._load_dev_examples()
-
-    def _load_train_examples(self):
-        self.train_file = os.path.join(self.dataset_dir, "train.tsv")
-        self.train_examples = self._read_tsv(self.train_file)
-
-    def _load_dev_examples(self):
-        self.dev_file = os.path.join(self.dataset_dir, "dev.tsv")
-        self.dev_examples = self._read_tsv(self.dev_file)
-
-    def _load_test_examples(self):
-        self.test_file = os.path.join(self.dataset_dir, "test.tsv")
-        self.test_examples = self._read_tsv(self.test_file)
-
-    def get_train_examples(self):
-        return self.train_examples
-
-    def get_dev_examples(self):
-        return self.dev_examples
-
-    def get_test_examples(self):
-        return self.test_examples
-
-    def get_labels(self):
-        """See base class."""
-        return ["0", "1"]
-
-    @property
-    def num_labels(self):
-        """
-        Return the number of labels in the dataset.
-        """
-        return len(self.get_labels())
-
-    def _read_tsv(self, input_file, quotechar=None):
+        dataset_dir = os.path.join(DATA_HOME, "nlpcc-dbqa")
+        base_path = self._download_dataset(dataset_dir, url=_DATA_URL)
+        super(NLPCC_DBQA, self).__init__(
+            base_path=base_path,
+            train_file="train.tsv",
+            dev_file="dev.tsv",
+            test_file="test.tsv",
+            label_file=None,
+            label_list=["0", "1"],
+        )
+
+    def _read_file(self, input_file, phase=None):
         """Reads a tab separated value file."""
         with codecs.open(input_file, "r", encoding="UTF-8") as f:
-            reader = csv.reader(f, delimiter="\t", quotechar=quotechar)
+            reader = csv.reader(f, delimiter="\t", quotechar=None)
             examples = []
             seq_id = 0
             header = next(reader)  # skip header
@@ -99,5 +65,13 @@ class NLPCC_DBQA(HubDataset):
 
 if __name__ == "__main__":
     ds = NLPCC_DBQA()
-    for e in ds.get_train_examples():
+    print("first 10 dev")
+    for e in ds.get_dev_examples()[:10]:
+        print("{}\t{}\t{}\t{}".format(e.guid, e.text_a, e.text_b, e.label))
+    print("first 10 train")
+    for e in ds.get_train_examples()[:10]:
+        print("{}\t{}\t{}\t{}".format(e.guid, e.text_a, e.text_b, e.label))
+    print("first 10 test")
+    for e in ds.get_test_examples()[:10]:
         print("{}\t{}\t{}\t{}".format(e.guid, e.text_a, e.text_b, e.label))
+    print(ds)

paddlehub/dataset/squad.py

@@ -16,12 +16,11 @@
 
 import json
 import os
-import sys
 
 from paddlehub.reader import tokenization
-from paddlehub.common.downloader import default_downloader
 from paddlehub.common.dir import DATA_HOME
 from paddlehub.common.logger import logger
+from paddlehub.dataset.base_nlp_dataset import BaseNLPDatast
 
 _DATA_URL = "https://bj.bcebos.com/paddlehub-dataset/squad.tar.gz"
 
@@ -66,61 +65,31 @@ class SquadExample(object):
         return s
 
 
-class SQUAD(object):
+class SQUAD(BaseNLPDatast):
     """A single set of features of data."""
 
     def __init__(self, version_2_with_negative=False):
-        self.dataset_dir = os.path.join(DATA_HOME, "squad_data")
-        if not os.path.exists(self.dataset_dir):
-            ret, tips, self.dataset_dir = default_downloader.download_file_and_uncompress(
-                url=_DATA_URL, save_path=DATA_HOME, print_progress=True)
-        else:
-            logger.info("Dataset {} already cached.".format(self.dataset_dir))
         self.version_2_with_negative = version_2_with_negative
-        self._load_train_examples(version_2_with_negative, if_has_answer=True)
-        self._load_dev_examples(version_2_with_negative, if_has_answer=True)
-
-    def _load_train_examples(self,
-                             version_2_with_negative=False,
-                             if_has_answer=True):
-        if not version_2_with_negative:
-            self.train_file = os.path.join(self.dataset_dir, "train-v1.1.json")
-        else:
-            self.train_file = os.path.join(self.dataset_dir, "train-v2.0.json")
-
-        self.train_examples = self._read_json(self.train_file, if_has_answer,
-                                              version_2_with_negative)
-
-    def _load_dev_examples(self,
-                           version_2_with_negative=False,
-                           if_has_answer=True):
         if not version_2_with_negative:
-            self.dev_file = os.path.join(self.dataset_dir, "dev-v1.1.json")
+            train_file = "train-v1.1.json"
+            dev_file = "dev-v1.1.json"
         else:
-            self.dev_file = os.path.join(self.dataset_dir, "dev-v2.0.json")
-
-        self.dev_examples = self._read_json(self.dev_file, if_has_answer,
-                                            version_2_with_negative)
-
-    def _load_test_examples(self,
-                            version_2_with_negative=False,
-                            is_training=False):
-        self.test_file = None
-        logger.error("not test_file")
-
-    def get_train_examples(self):
-        return self.train_examples
-
-    def get_dev_examples(self):
-        return self.dev_examples
-
-    def get_test_examples(self):
-        return []
-
-    def _read_json(self,
-                   input_file,
-                   if_has_answer,
-                   version_2_with_negative=False):
+            train_file = "train-v2.0.json"
+            dev_file = "dev-v2.0.json"
+
+        dataset_dir = os.path.join(DATA_HOME, "squad_data")
+        base_path = self._download_dataset(dataset_dir, url=_DATA_URL)
+
+        super(SQUAD, self).__init__(
+            base_path=base_path,
+            train_file=train_file,
+            dev_file=dev_file,
+            test_file=None,
+            label_file=None,
+            label_list=None,
+        )
+
+    def _read_file(self, input_file, phase=None):
         """Read a SQuAD json file into a list of SquadExample."""
         with open(input_file, "r") as reader:
             input_data = json.load(reader)["data"]
@@ -156,13 +125,15 @@ class SQUAD(object):
                     end_position = None
                     orig_answer_text = None
                     is_impossible = False
-                    if if_has_answer:
-                        if version_2_with_negative:
+                    if phase in ["train", "dev"]:
+                        if self.version_2_with_negative:
                             is_impossible = qa["is_impossible"]
-                        # if (len(qa["answers"]) != 1) and (not is_impossible):
-                        #     raise ValueError(
-                        #         "For training, each question should have exactly 1 answer."
-                        #     )
+                        if phase == "train" and (len(qa["answers"]) !=
+                                                 1) and (not is_impossible):
+                            print(qa)
+                            raise ValueError(
+                                "For training, each question should have exactly 1 answer."
+                            )
                         if not is_impossible:
                             answer = qa["answers"][0]
                             orig_answer_text = answer["text"]
@@ -206,8 +177,14 @@ class SQUAD(object):
 
 
 if __name__ == "__main__":
-    ds = SQUAD(version_2_with_negative=False)
-    examples = ds.get_train_examples()
-    for index, e in enumerate(examples):
-        if index < 10:
-            print(e)
+    ds = SQUAD(version_2_with_negative=True)
+    print("first 10 dev")
+    for e in ds.get_dev_examples()[:2]:
+        print(e)
+    print("first 10 train")
+    for e in ds.get_train_examples()[:2]:
+        print(e)
+    print("first 10 test")
+    for e in ds.get_test_examples()[:2]:
+        print(e)
+    print(ds)

paddlehub/dataset/stanford_dogs.py

@@ -20,20 +20,35 @@ from __future__ import print_function
 import os
 
 import paddlehub as hub
-from paddlehub.dataset.base_cv_dataset import ImageClassificationDataset
+from paddlehub.dataset.base_cv_dataset import BaseCVDatast
 
 
-class StanfordDogsDataset(ImageClassificationDataset):
+class StanfordDogsDataset(BaseCVDatast):
     def __init__(self):
-        super(StanfordDogsDataset, self).__init__()
         dataset_path = os.path.join(hub.common.dir.DATA_HOME,
                                     "StanfordDogs-120")
-        self.base_path = self._download_dataset(
+        base_path = self._download_dataset(
             dataset_path=dataset_path,
             url="https://bj.bcebos.com/paddlehub-dataset/StanfordDogs-120.tar.gz"
         )
-        self.train_list_file = "train_list.txt"
-        self.test_list_file = "test_list.txt"
-        self.validate_list_file = "validate_list.txt"
-        self.label_list_file = "label_list.txt"
-        self.num_labels = 120
+        super(StanfordDogsDataset, self).__init__(
+            base_path=base_path,
+            train_list_file="train_list.txt",
+            validate_list_file="validate_list.txt",
+            test_list_file="test_list.txt",
+            label_list_file="label_list.txt",
+            label_list=None)
+
+
+if __name__ == "__main__":
+    ds = StanfordDogsDataset()
+    print("first 10 dev")
+    for e in ds.get_dev_examples()[:10]:
+        print(e)
+    print("first 10 train")
+    for e in ds.get_train_examples()[:10]:
+        print(e)
+    print("first 10 test")
+    for e in ds.get_test_examples()[:10]:
+        print(e)
+    print(ds)

paddlehub/dataset/thucnews.py

@@ -17,64 +17,30 @@ from __future__ import absolute_import
 from __future__ import division
 from __future__ import print_function
 
-from collections import namedtuple
 import io
 import os
-import csv
 
-from paddlehub.dataset import InputExample, HubDataset
-from paddlehub.common.downloader import default_downloader
+from paddlehub.dataset import InputExample
 from paddlehub.common.dir import DATA_HOME
-from paddlehub.common.logger import logger
+from paddlehub.dataset.base_nlp_dataset import BaseNLPDatast
 
 _DATA_URL = "https://bj.bcebos.com/paddlehub-dataset/thucnews.tar.gz"
 
 
-class THUCNEWS(HubDataset):
+class THUCNEWS(BaseNLPDatast):
     def __init__(self):
-        self.dataset_dir = os.path.join(DATA_HOME, "thucnews")
-        if not os.path.exists(self.dataset_dir):
-            ret, tips, self.dataset_dir = default_downloader.download_file_and_uncompress(
-                url=_DATA_URL, save_path=DATA_HOME, print_progress=True)
-        else:
-            logger.info("Dataset {} already cached.".format(self.dataset_dir))
-
-        self._load_train_examples()
-        self._load_test_examples()
-        self._load_dev_examples()
-
-    def _load_train_examples(self):
-        self.train_file = os.path.join(self.dataset_dir, "train.txt")
-        self.train_examples = self._read_file(self.train_file)
-
-    def _load_dev_examples(self):
-        self.dev_file = os.path.join(self.dataset_dir, "dev.txt")
-        self.dev_examples = self._read_file(self.dev_file)
-
-    def _load_test_examples(self):
-        self.test_file = os.path.join(self.dataset_dir, "test.txt")
-        self.test_examples = self._read_file(self.test_file)
-
-    def get_train_examples(self):
-        return self.train_examples
-
-    def get_dev_examples(self):
-        return self.dev_examples
-
-    def get_test_examples(self):
-        return self.test_examples
-
-    def get_labels(self):
-        return [str(i) for i in range(14)]
-
-    @property
-    def num_labels(self):
-        """
-        Return the number of labels in the dataset.
-        """
-        return len(self.get_labels())
-
-    def _read_file(self, input_file):
+        dataset_dir = os.path.join(DATA_HOME, "thucnews")
+        base_path = self._download_dataset(dataset_dir, url=_DATA_URL)
+        super(THUCNEWS, self).__init__(
+            base_path=base_path,
+            train_file="train.txt",
+            dev_file="dev.txt",
+            test_file="test.txt",
+            label_file=None,
+            label_list=[str(i) for i in range(14)],
+        )
+
+    def _read_file(self, input_file, phase=None):
         """Reads a tab separated value file."""
         with io.open(input_file, "r", encoding="UTF-8") as file:
             examples = []
@@ -91,5 +57,13 @@ class THUCNEWS(HubDataset):
 
 if __name__ == "__main__":
     ds = THUCNEWS()
+    print("first 10 dev")
+    for e in ds.get_dev_examples()[:10]:
+        print("{}\t{}\t{}\t{}".format(e.guid, e.text_a, e.text_b, e.label))
+    print("first 10 train")
     for e in ds.get_train_examples()[:10]:
         print("{}\t{}\t{}\t{}".format(e.guid, e.text_a, e.text_b, e.label))
+    print("first 10 test")
+    for e in ds.get_test_examples()[:10]:
+        print("{}\t{}\t{}\t{}".format(e.guid, e.text_a, e.text_b, e.label))
+    print(ds)

paddlehub/dataset/tnews.py

@@ -17,15 +17,11 @@ from __future__ import absolute_import
 from __future__ import division
 from __future__ import print_function
 
-from collections import namedtuple
 import io
 import os
-import csv
 
-from paddlehub.dataset import InputExample, HubDataset
-from paddlehub.common.downloader import default_downloader
+from paddlehub.dataset import InputExample, BaseDataset
 from paddlehub.common.dir import DATA_HOME
-from paddlehub.common.logger import logger
 
 _DATA_URL = "https://bj.bcebos.com/paddlehub-dataset/tnews.tar.gz"
 
@@ -48,64 +44,31 @@ LABEL_NAME = {
 }
 
 
-class TNews(HubDataset):
+class TNews(BaseDataset):
     """
     TNews is the chinese news classification dataset on Jinri Toutiao App.
     """
 
     def __init__(self):
-        self.dataset_dir = os.path.join(DATA_HOME, "tnews")
-        if not os.path.exists(self.dataset_dir):
-            ret, tips, self.dataset_dir = default_downloader.download_file_and_uncompress(
-                url=_DATA_URL, save_path=DATA_HOME, print_progress=True)
-        else:
-            logger.info("Dataset {} already cached.".format(self.dataset_dir))
-
-        self._load_train_examples()
-        self._load_test_examples()
-        self._load_dev_examples()
-
-    def _load_train_examples(self):
-        self.train_file = os.path.join(self.dataset_dir,
-                                       "toutiao_category_train.txt")
-        self.train_examples = self._read_file(self.train_file)
-
-    def _load_dev_examples(self):
-        self.dev_file = os.path.join(self.dataset_dir,
-                                     "toutiao_category_dev.txt")
-        self.dev_examples = self._read_file(self.dev_file)
-
-    def _load_test_examples(self):
-        self.test_file = os.path.join(self.dataset_dir,
-                                      "toutiao_category_test.txt")
-        self.test_examples = self._read_file(self.test_file)
-
-    def get_train_examples(self):
-        return self.train_examples
-
-    def get_dev_examples(self):
-        return self.dev_examples
-
-    def get_test_examples(self):
-        return self.test_examples
-
-    def get_labels(self):
-        return [
+        dataset_dir = os.path.join(DATA_HOME, "tnews")
+        base_path = self._download_dataset(dataset_dir, url=_DATA_URL)
+        label_list = [
             '100', '101', '102', '103', '104', '106', '107', '108', '109',
             '110', '112', '113', '114', '115', '116'
         ]
+        super(TNews, self).__init__(
+            base_path=base_path,
+            train_file="toutiao_category_train.txt",
+            dev_file="toutiao_category_dev.txt",
+            test_file="toutiao_category_test.txt",
+            label_file=None,
+            label_list=label_list,
+        )
 
     def get_label_name(self, id):
         return LABEL_NAME[id]
 
-    @property
-    def num_labels(self):
-        """
-        Return the number of labels in the dataset.
-        """
-        return len(self.get_labels())
-
-    def _read_file(self, input_file):
+    def _read_file(self, input_file, phase=None):
         """Reads a tab separated value file."""
         with io.open(input_file, "r", encoding="UTF-8") as file:
             examples = []
@@ -120,5 +83,13 @@ class TNews(HubDataset):
 
 if __name__ == "__main__":
     ds = TNews()
+    print("first 10 dev")
+    for e in ds.get_dev_examples()[:10]:
+        print("{}\t{}\t{}\t{}".format(e.guid, e.text_a, e.text_b, e.label))
+    print("first 10 train")
     for e in ds.get_train_examples()[:10]:
         print("{}\t{}\t{}\t{}".format(e.guid, e.text_a, e.text_b, e.label))
+    print("first 10 test")
+    for e in ds.get_test_examples()[:10]:
+        print("{}\t{}\t{}\t{}".format(e.guid, e.text_a, e.text_b, e.label))
+    print(ds)

paddlehub/dataset/toxic.py

@@ -17,73 +17,39 @@ from __future__ import absolute_import
 from __future__ import division
 from __future__ import print_function
 
-from collections import namedtuple
-import codecs
 import os
 import pandas as pd
-from numpy import nan
 
-from paddlehub.dataset import InputExample, HubDataset
-from paddlehub.common.downloader import default_downloader
+from paddlehub.dataset import InputExample
 from paddlehub.common.dir import DATA_HOME
-from paddlehub.common.logger import logger
+from paddlehub.dataset.base_nlp_dataset import BaseNLPDatast
 
 _DATA_URL = "https://bj.bcebos.com/paddlehub-dataset/toxic.tar.gz"
 
 
-class Toxic(HubDataset):
+class Toxic(BaseNLPDatast):
     """
     The kaggle Toxic dataset:
     https://www.kaggle.com/c/jigsaw-toxic-comment-classification-challenge
     """
 
     def __init__(self):
-        self.dataset_dir = os.path.join(DATA_HOME, "toxic")
-        if not os.path.exists(self.dataset_dir):
-            ret, tips, self.dataset_dir = default_downloader.download_file_and_uncompress(
-                url=_DATA_URL, save_path=DATA_HOME, print_progress=True)
-        else:
-            logger.info("Dataset {} already cached.".format(self.dataset_dir))
-
-        self._load_train_examples()
-        self._load_test_examples()
-        self._load_dev_examples()
-
-    def _load_train_examples(self):
-        self.train_file = os.path.join(self.dataset_dir, "train.csv")
-        self.train_examples = self._read_csv(self.train_file)
-
-    def _load_dev_examples(self):
-        self.dev_file = os.path.join(self.dataset_dir, "dev.csv")
-        self.dev_examples = self._read_csv(self.dev_file)
-
-    def _load_test_examples(self):
-        self.test_file = os.path.join(self.dataset_dir, "test.csv")
-        self.test_examples = self._read_csv(self.test_file)
-
-    def get_train_examples(self):
-        return self.train_examples
-
-    def get_dev_examples(self):
-        return self.dev_examples
-
-    def get_test_examples(self):
-        return self.test_examples
-
-    def get_labels(self):
-        return [
+        dataset_dir = os.path.join(DATA_HOME, "toxic")
+        base_path = self._download_dataset(dataset_dir, url=_DATA_URL)
+        label_list = [
             'toxic', 'severe_toxic', 'obscene', 'threat', 'insult',
             'identity_hate'
         ]
-
-    @property
-    def num_labels(self):
-        """
-        Return the number of labels in the dataset.
-        """
-        return len(self.get_labels())
-
-    def _read_csv(self, input_file, quotechar=None):
+        super(Toxic, self).__init__(
+            base_path=base_path,
+            train_file="train.csv",
+            dev_file="dev.csv",
+            test_file="test.csv",
+            label_file=None,
+            label_list=label_list,
+        )
+
+    def _read_file(self, input_file, phase=None):
         """Reads a tab separated value file."""
         data = pd.read_csv(input_file, encoding="UTF-8")
         examples = []
@@ -99,5 +65,13 @@ class Toxic(HubDataset):
 
 if __name__ == "__main__":
     ds = Toxic()
-    for e in ds.get_train_examples():
+    print("first 10 dev")
+    for e in ds.get_dev_examples()[:10]:
+        print("{}\t{}\t{}\t{}".format(e.guid, e.text_a, e.text_b, e.label))
+    print("first 10 train")
+    for e in ds.get_train_examples()[:10]:
+        print("{}\t{}\t{}\t{}".format(e.guid, e.text_a, e.text_b, e.label))
+    print("first 10 test")
+    for e in ds.get_test_examples()[:10]:
         print("{}\t{}\t{}\t{}".format(e.guid, e.text_a, e.text_b, e.label))
+    print(ds)

paddlehub/dataset/xnli.py

@@ -23,15 +23,14 @@ import io
 import os
 import csv
 
-from paddlehub.dataset import InputExample, HubDataset
-from paddlehub.common.downloader import default_downloader
+from paddlehub.dataset import InputExample
 from paddlehub.common.dir import DATA_HOME
-from paddlehub.common.logger import logger
+from paddlehub.dataset.base_nlp_dataset import BaseNLPDatast
 
 _DATA_URL = "https://bj.bcebos.com/paddlehub-dataset/XNLI-lan.tar.gz"
 
 
-class XNLI(HubDataset):
+class XNLI(BaseNLPDatast):
     """
     Please refer to
     https://arxiv.org/pdf/1809.05053.pdf
@@ -43,61 +42,25 @@ class XNLI(HubDataset):
                 "ar", "bg", "de", "el", "en", "es", "fr", "hi", "ru", "sw",
                 "th", "tr", "ur", "vi", "zh"
         ]:
-
-            raise Exception(language +
-                            "is not in XNLI. Please confirm the language")
+            raise Exception(
+                "%s is not in XNLI. Please confirm the language" % language)
         self.language = language
-        self.dataset_dir = os.path.join(DATA_HOME, "XNLI-lan")
-
-        if not os.path.exists(self.dataset_dir):
-            ret, tips, self.dataset_dir = default_downloader.download_file_and_uncompress(
-                url=_DATA_URL, save_path=DATA_HOME, print_progress=True)
-        else:
-            logger.info("Dataset {} already cached.".format(self.dataset_dir))
-
-        self._load_train_examples()
-        self._load_test_examples()
-        self._load_dev_examples()
-
-    def _load_train_examples(self):
-        self.train_file = os.path.join(self.dataset_dir, self.language,
-                                       self.language + "_train.tsv")
-        self.train_examples = self._read_tsv(self.train_file)
-
-    def _load_dev_examples(self):
-        self.dev_file = os.path.join(self.dataset_dir, self.language,
-                                     self.language + "_dev.tsv")
-        self.dev_examples = self._read_tsv(self.dev_file)
-
-    def _load_test_examples(self):
-        self.test_file = os.path.join(self.dataset_dir, self.language,
-                                      self.language + "_test.tsv")
-        self.test_examples = self._read_tsv(self.test_file)
-
-    def get_train_examples(self):
-        return self.train_examples
-
-    def get_dev_examples(self):
-        return self.dev_examples
-
-    def get_test_examples(self):
-        return self.test_examples
-
-    def get_labels(self):
-        """See base class."""
-        return ["neutral", "contradiction", "entailment"]
-
-    @property
-    def num_labels(self):
-        """
-        Return the number of labels in the dataset.
-        """
-        return len(self.get_labels())
-
-    def _read_tsv(self, input_file, quotechar=None):
+        dataset_dir = os.path.join(DATA_HOME, "XNLI-lan")
+        dataset_dir = self._download_dataset(dataset_dir, url=_DATA_URL)
+        base_path = os.path.join(dataset_dir, language)
+        super(XNLI, self).__init__(
+            base_path=base_path,
+            train_file="%s_train.tsv" % language,
+            dev_file="%s_dev.tsv" % language,
+            test_file="%s_test.tsv" % language,
+            label_file=None,
+            label_list=["neutral", "contradiction", "entailment"],
+        )
+
+    def _read_file(self, input_file, phase=None):
         """Reads a tab separated value file."""
         with io.open(input_file, "r", encoding="UTF-8") as f:
-            reader = csv.reader(f, delimiter="\t", quotechar=quotechar)
+            reader = csv.reader(f, delimiter="\t", quotechar=None)
             examples = []
             seq_id = 0
             header = next(reader)  # skip header
@@ -112,5 +75,13 @@ class XNLI(HubDataset):
 
 if __name__ == "__main__":
     ds = XNLI()
-    for e in ds.get_train_examples()[:3]:
+    print("first 10 dev")
+    for e in ds.get_dev_examples()[:10]:
+        print("{}\t{}\t{}\t{}".format(e.guid, e.text_a, e.text_b, e.label))
+    print("first 10 train")
+    for e in ds.get_train_examples()[:10]:
+        print("{}\t{}\t{}\t{}".format(e.guid, e.text_a, e.text_b, e.label))
+    print("first 10 test")
+    for e in ds.get_test_examples()[:10]:
         print("{}\t{}\t{}\t{}".format(e.guid, e.text_a, e.text_b, e.label))
+    print(ds)

paddlehub/finetune/task/__init__.py

@@ -13,7 +13,7 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-from .basic_task import BasicTask, RunEnv, RunState
+from .base_task import BaseTask, RunEnv, RunState
 from .classifier_task import ClassifierTask, ImageClassifierTask, TextClassifierTask, MultiLabelClassifierTask
 from .reading_comprehension_task import ReadingComprehensionTask
 from .regression_task import RegressionTask

paddlehub/finetune/task/basic_task.py → paddlehub/finetune/task/base_task.py

@@ -192,7 +192,7 @@ class TaskHooks():
         return self.info(only_customized=False)
 
 
-class BasicTask(object):
+class BaseTask(object):
     def __init__(self,
                  feed_list,
                  data_reader,
@@ -265,7 +265,7 @@ class BasicTask(object):
         for hook_type, event_hooks in self._hooks._registered_hooks.items():
             self._hooks.add(hook_type, "default",
                             eval("self._default_%s_event" % hook_type))
-            setattr(BasicTask, "_%s_event" % hook_type,
+            setattr(BaseTask, "_%s_event" % hook_type,
                     self.create_event_function(hook_type))
 
         # accelerate predict

paddlehub/finetune/task/classifier_task.py

@@ -23,10 +23,10 @@ import numpy as np
 import paddle.fluid as fluid
 
 from paddlehub.finetune.evaluate import calculate_f1_np, matthews_corrcoef
-from .basic_task import BasicTask
+from .base_task import BaseTask
 
 
-class ClassifierTask(BasicTask):
+class ClassifierTask(BaseTask):
     def __init__(self,
                  feature,
                  num_classes,

paddlehub/finetune/task/reading_comprehension_task.py

@@ -28,7 +28,7 @@ from collections import OrderedDict
 
 import numpy as np
 import paddle.fluid as fluid
-from .basic_task import BasicTask
+from .base_task import BaseTask
 from paddlehub.common.logger import logger
 from paddlehub.reader import tokenization
 from paddlehub.finetune.evaluator import squad1_evaluate
@@ -176,6 +176,13 @@ def write_predictions(all_examples, all_features, all_results, n_best_size,
                       output_nbest_file, output_null_log_odds_file,
                       version_2_with_negative, null_score_diff_threshold,
                       is_english):
+    _PrelimPrediction = collections.namedtuple("PrelimPrediction", [
+        "feature_index", "start_index", "end_index", "start_logit", "end_logit"
+    ])
+
+    _NbestPrediction = collections.namedtuple(
+        "NbestPrediction", ["text", "start_logit", "end_logit"])
+
     example_index_to_features = collections.defaultdict(list)
     for feature in all_features:
         example_index_to_features[feature.example_index].append(feature)
@@ -184,10 +191,6 @@ def write_predictions(all_examples, all_features, all_results, n_best_size,
     for result in all_results:
         unique_id_to_result[result.unique_id] = result
 
-    _PrelimPrediction = collections.namedtuple("PrelimPrediction", [
-        "feature_index", "start_index", "end_index", "start_logit", "end_logit"
-    ])
-
     all_predictions = collections.OrderedDict()
     all_nbest_json = collections.OrderedDict()
     scores_diff_json = collections.OrderedDict()
@@ -262,9 +265,6 @@ def write_predictions(all_examples, all_features, all_results, n_best_size,
             key=lambda x: (x.start_logit + x.end_logit),
             reverse=True)
 
-        _NbestPrediction = collections.namedtuple(  # pylint: disable=invalid-name
-            "NbestPrediction", ["text", "start_logit", "end_logit"])
-
         seen_predictions = {}
         nbest = []
         if not prelim_predictions:
@@ -384,7 +384,7 @@ def write_predictions(all_examples, all_features, all_results, n_best_size,
                 + "\n")
 
 
-class ReadingComprehensionTask(BasicTask):
+class ReadingComprehensionTask(BaseTask):
     def __init__(self,
                  feature,
                  feed_list,
@@ -420,6 +420,9 @@ class ReadingComprehensionTask(BasicTask):
         self.n_best_size = n_best_size
         self.max_answer_length = max_answer_length
 
+        self.RawResult = collections.namedtuple(
+            "RawResult", ["unique_id", "start_logits", "end_logits"])
+
     def _build_net(self):
         self.unique_ids = fluid.layers.data(
             name="unique_ids", shape=[-1, 1], lod_level=0, dtype="int64")
@@ -493,8 +496,6 @@ class ReadingComprehensionTask(BasicTask):
     def _calculate_metrics(self, run_states):
         total_cost, total_num_seqs, all_results = [], [], []
         run_step = 0
-        RawResult = collections.namedtuple(
-            "RawResult", ["unique_id", "start_logits", "end_logits"])
         for run_state in run_states:
             np_loss = run_state.run_results[0]
             np_num_seqs = run_state.run_results[1]
@@ -510,7 +511,7 @@ class ReadingComprehensionTask(BasicTask):
                     start_logits = [float(x) for x in np_start_logits[idx].flat]
                     end_logits = [float(x) for x in np_end_logits[idx].flat]
                     all_results.append(
-                        RawResult(
+                        self.RawResult(
                             unique_id=unique_id,
                             start_logits=start_logits,
                             end_logits=end_logits))
@@ -544,13 +545,13 @@ class ReadingComprehensionTask(BasicTask):
                 is_english=self.is_english)
             if self.phase == 'val' or self.phase == 'dev':
                 with open(
-                        self.data_reader.dataset.dev_file, 'r',
+                        self.data_reader.dataset.dev_path, 'r',
                         encoding="utf8") as dataset_file:
                     dataset_json = json.load(dataset_file)
                     dataset = dataset_json['data']
             elif self.phase == 'test':
                 with open(
-                        self.data_reader.dataset.test_file, 'r',
+                        self.data_reader.dataset.test_path, 'r',
                         encoding="utf8") as dataset_file:
                     dataset_json = json.load(dataset_file)
                     dataset = dataset_json['data']
@@ -577,8 +578,6 @@ class ReadingComprehensionTask(BasicTask):
 
     def _default_predict_end_event(self, run_states):
         all_results = []
-        RawResult = collections.namedtuple(
-            "RawResult", ["unique_id", "start_logits", "end_logits"])
         for run_state in run_states:
             np_unique_ids = run_state.run_results[0]
             np_start_logits = run_state.run_results[1]
@@ -588,7 +587,7 @@ class ReadingComprehensionTask(BasicTask):
                 start_logits = [float(x) for x in np_start_logits[idx].flat]
                 end_logits = [float(x) for x in np_end_logits[idx].flat]
                 all_results.append(
-                    RawResult(
+                    self.RawResult(
                         unique_id=unique_id,
                         start_logits=start_logits,
                         end_logits=end_logits))

paddlehub/finetune/task/regression_task.py

@@ -23,10 +23,10 @@ from collections import OrderedDict
 import numpy as np
 import paddle.fluid as fluid
 from scipy.stats import spearmanr
-from .basic_task import BasicTask
+from .base_task import BaseTask
 
 
-class RegressionTask(BasicTask):
+class RegressionTask(BaseTask):
     def __init__(self,
                  feature,
                  feed_list,

paddlehub/finetune/task/sequence_task.py

@@ -25,10 +25,10 @@ import paddle
 import paddle.fluid as fluid
 from paddlehub.finetune.evaluate import chunk_eval, calculate_f1
 from paddlehub.common.utils import version_compare
-from .basic_task import BasicTask
+from .base_task import BaseTask
 
 
-class SequenceLabelTask(BasicTask):
+class SequenceLabelTask(BaseTask):
     def __init__(self,
                  feature,
                  max_seq_len,

paddlehub/reader/base_reader.py

+import numpy as np
+
+
+class BaseReader(object):
+    def __init__(self, dataset, random_seed=None):
+        self.dataset = dataset
+        self.num_examples = {'train': -1, 'dev': -1, 'test': -1}
+        np.random.seed(random_seed)
+
+    def get_train_examples(self):
+        return self.dataset.get_train_examples()
+
+    def get_dev_examples(self):
+        return self.dataset.get_dev_examples()
+
+    def get_test_examples(self):
+        return self.dataset.get_test_examples()
+
+    def data_generator(self):
+        raise NotImplementedError

paddlehub/reader/cv_reader.py

@@ -22,6 +22,7 @@ import numpy as np
 from PIL import Image
 
 import paddlehub.io.augmentation as image_augmentation
+from .base_reader import BaseReader
 
 channel_order_dict = {
     "RGB": [0, 1, 2],
@@ -33,7 +34,7 @@ channel_order_dict = {
 }
 
 
-class ImageClassificationReader(object):
+class ImageClassificationReader(BaseReader):
     def __init__(self,
                  image_width,
                  image_height,
@@ -41,15 +42,15 @@ class ImageClassificationReader(object):
                  channel_order="RGB",
                  images_mean=None,
                  images_std=None,
-                 data_augmentation=False):
+                 data_augmentation=False,
+                 random_seed=None):
+        super(ImageClassificationReader, self).__init__(dataset, random_seed)
         self.image_width = image_width
         self.image_height = image_height
         self.channel_order = channel_order
-        self.dataset = dataset
         self.data_augmentation = data_augmentation
         self.images_std = images_std
         self.images_mean = images_mean
-        self.num_examples = {'train': -1, 'dev': -1, 'test': -1}
 
         if self.images_mean is None:
             try:
@@ -73,24 +74,38 @@ class ImageClassificationReader(object):
             raise ValueError("Image width and height should not be negative.")
 
     def data_generator(self,
-                       batch_size,
+                       batch_size=1,
                        phase="train",
                        shuffle=False,
                        data=None):
         if phase != 'predict' and not self.dataset:
             raise ValueError("The dataset is none and it's not allowed!")
         if phase == "train":
-            data = self.dataset.train_data(shuffle)
-            self.num_examples['train'] = len(self.get_train_examples())
-        elif phase == "test":
-            shuffle = False
-            data = self.dataset.test_data(shuffle)
-            self.num_examples['test'] = len(self.get_test_examples())
+            shuffle = True
+            if hasattr(self.dataset, "train_data"):
+                # Compatible with ImageClassificationDataset which has done shuffle
+                self.dataset.train_data()
+                shuffle = False
+            data = self.get_train_examples()
+            self.num_examples['train'] = len(data)
         elif phase == "val" or phase == "dev":
             shuffle = False
-            data = self.dataset.validate_data(shuffle)
-            self.num_examples['dev'] = len(self.get_dev_examples())
+            if hasattr(self.dataset, "validate_data"):
+                # Compatible with ImageClassificationDataset
+                self.dataset.validate_data()
+                shuffle = False
+            data = self.get_dev_examples()
+            self.num_examples['dev'] = len(data)
+        elif phase == "test":
+            shuffle = False
+            if hasattr(self.dataset, "test_data"):
+                # Compatible with ImageClassificationDataset
+                data = self.dataset.test_data()
+                shuffle = False
+            data = self.get_test_examples()
+            self.num_examples['test'] = len(data)
         elif phase == "predict":
+            shuffle = False
             data = data
 
         def preprocess(image_path):
@@ -118,6 +133,9 @@ class ImageClassificationReader(object):
             return image
 
         def _data_reader():
+            if shuffle:
+                np.random.shuffle(data)
+
             if phase == "predict":
                 for image_path in data:
                     image = preprocess(image_path)
@@ -128,12 +146,3 @@ class ImageClassificationReader(object):
                     yield (image, label)
 
         return paddle.batch(_data_reader, batch_size=batch_size)
-
-    def get_train_examples(self):
-        return self.dataset.train_examples
-
-    def get_dev_examples(self):
-        return self.dataset.dev_examples
-
-    def get_test_examples(self):
-        return self.dataset.test_examples

paddlehub/reader/nlp_reader.py

@@ -18,11 +18,8 @@ from __future__ import division
 from __future__ import print_function
 
 import collections
-import json
 import numpy as np
-import platform
 import six
-import sys
 from collections import namedtuple
 
 import paddle
@@ -31,11 +28,12 @@ from paddlehub.reader import tokenization
 from paddlehub.common.logger import logger
 from paddlehub.common.utils import sys_stdout_encoding
 from paddlehub.dataset.dataset import InputExample
-from .batching import pad_batch_data, prepare_batch_data
+from .batching import pad_batch_data
 import paddlehub as hub
+from .base_reader import BaseReader
 
 
-class BaseReader(object):
+class BaseNLPReader(BaseReader):
     def __init__(self,
                  vocab_path,
                  dataset=None,
@@ -47,6 +45,7 @@ class BaseReader(object):
                  sp_model_path=None,
                  word_dict_path=None,
                  in_tokens=False):
+        super(BaseNLPReader, self).__init__(dataset, random_seed)
         self.max_seq_len = max_seq_len
         if sp_model_path and word_dict_path:
             self.tokenizer = tokenization.WSSPTokenizer(
@@ -55,52 +54,35 @@ class BaseReader(object):
             self.tokenizer = tokenization.FullTokenizer(
                 vocab_file=vocab_path, do_lower_case=do_lower_case)
         self.vocab = self.tokenizer.vocab
-        self.dataset = dataset
         self.pad_id = self.vocab["[PAD]"]
         self.cls_id = self.vocab["[CLS]"]
         self.sep_id = self.vocab["[SEP]"]
+        self.mask_id = self.vocab["[MASK]"]
         self.in_tokens = in_tokens
         self.use_task_id = use_task_id
 
         if self.use_task_id:
+            logger.warning(
+                "use_task_id has been de discarded since PaddleHub v1.4.0")
             self.task_id = 0
 
-        np.random.seed(random_seed)
-
         # generate label map
         self.label_map = {}
-        if self.dataset:
+        try:
             for index, label in enumerate(self.dataset.get_labels()):
                 self.label_map[label] = index
             logger.info("Dataset label map = {}".format(self.label_map))
-        else:
-            logger.info("Dataset is None! label map = {}".format(
-                self.label_map))
-
-        self.current_example = 0
-        self.current_epoch = 0
-
-        self.num_examples = {'train': -1, 'dev': -1, 'test': -1}
-
-    def get_train_examples(self):
-        """Gets a collection of `InputExample`s for the train set."""
-        return self.dataset.get_train_examples()
-
-    def get_dev_examples(self):
-        """Gets a collection of `InputExample`s for the dev set."""
-        return self.dataset.get_dev_examples()
+        except:
+            # some dataset like squad, its label_list=None
+            logger.info(
+                "Dataset is None or it has not any labels, label map = {}".
+                format(self.label_map))
 
-    def get_val_examples(self):
-        """Gets a collection of `InputExample`s for the val set."""
-        return self.dataset.get_val_examples()
-
-    def get_test_examples(self):
-        """Gets a collection of `InputExample`s for prediction."""
-        return self.dataset.get_test_examples()
-
-    def get_train_progress(self):
-        """Gets progress for training phase."""
-        return self.current_example, self.current_epoch
+        self.Record_With_Label_Id = namedtuple(
+            'Record',
+            ['token_ids', 'text_type_ids', 'position_ids', 'label_id'])
+        self.Record_Wo_Label_Id = namedtuple(
+            'Record', ['token_ids', 'text_type_ids', 'position_ids'])
 
     def _truncate_seq_pair(self, tokens_a, tokens_b, max_length):
         """Truncates a sequence pair in place to the maximum length."""
@@ -189,24 +171,14 @@ class BaseReader(object):
         else:
             label_id = example.label
 
-        Record = namedtuple(
-            'Record',
-            ['token_ids', 'text_type_ids', 'position_ids', 'label_id'])
-
         if phase != "predict":
-            Record = namedtuple(
-                'Record',
-                ['token_ids', 'text_type_ids', 'position_ids', 'label_id'])
-
-            record = Record(
+            record = self.Record_With_Label_Id(
                 token_ids=token_ids,
                 text_type_ids=text_type_ids,
                 position_ids=position_ids,
                 label_id=label_id)
         else:
-            Record = namedtuple('Record',
-                                ['token_ids', 'text_type_ids', 'position_ids'])
-            record = Record(
+            record = self.Record_Wo_Label_Id(
                 token_ids=token_ids,
                 text_type_ids=text_type_ids,
                 position_ids=position_ids)
@@ -238,14 +210,6 @@ class BaseReader(object):
         if batch_records:
             yield self._pad_batch_records(batch_records, phase)
 
-    def get_num_examples(self, phase):
-        """Get number of examples for train, dev or test."""
-        if phase not in ['train', 'val', 'dev', 'test']:
-            raise ValueError(
-                "Unknown phase, which should be in ['train', 'val'/'dev', 'test']."
-            )
-        return self.num_examples[phase]
-
     def data_generator(self,
                        batch_size=1,
                        phase='train',
@@ -307,7 +271,7 @@ class BaseReader(object):
         return wrapper
 
 
-class ClassifyReader(BaseReader):
+class ClassifyReader(BaseNLPReader):
     def _pad_batch_records(self, batch_records, phase=None):
         batch_token_ids = [record.token_ids for record in batch_records]
         batch_text_type_ids = [record.text_type_ids for record in batch_records]
@@ -360,7 +324,7 @@ class ClassifyReader(BaseReader):
         return return_list
 
 
-class SequenceLabelReader(BaseReader):
+class SequenceLabelReader(BaseNLPReader):
     def __init__(self,
                  vocab_path,
                  dataset=None,
@@ -508,15 +472,11 @@ class SequenceLabelReader(BaseReader):
             label_ids = [no_entity_id
                          ] + [self.label_map[label]
                               for label in labels] + [no_entity_id]
-
-            Record = namedtuple(
-                'Record',
-                ['token_ids', 'text_type_ids', 'position_ids', 'label_ids'])
-            record = Record(
+            record = self.Record_With_Label_Id(
                 token_ids=token_ids,
                 text_type_ids=text_type_ids,
                 position_ids=position_ids,
-                label_ids=label_ids)
+                label_id=label_ids)
         else:
             tokens = self._reseg_token_label(
                 tokens=tokens, tokenizer=tokenizer, phase=phase)
@@ -529,9 +489,7 @@ class SequenceLabelReader(BaseReader):
             position_ids = list(range(len(token_ids)))
             text_type_ids = [0] * len(token_ids)
 
-            Record = namedtuple('Record',
-                                ['token_ids', 'text_type_ids', 'position_ids'])
-            record = Record(
+            record = self.Record_Wo_Label_Id(
                 token_ids=token_ids,
                 text_type_ids=text_type_ids,
                 position_ids=position_ids,
@@ -540,109 +498,7 @@ class SequenceLabelReader(BaseReader):
         return record
 
 
-class LACClassifyReader(object):
-    def __init__(self, vocab_path, dataset=None, in_tokens=False):
-        self.dataset = dataset
-        self.lac = hub.Module(name="lac")
-        self.tokenizer = tokenization.FullTokenizer(
-            vocab_file=vocab_path, do_lower_case=False)
-        self.vocab = self.tokenizer.vocab
-        self.feed_key = list(
-            self.lac.processor.data_format(
-                sign_name="lexical_analysis").keys())[0]
-
-        self.num_examples = {'train': -1, 'dev': -1, 'test': -1}
-        self.in_tokens = in_tokens
-
-    def get_num_examples(self, phase):
-        """Get number of examples for train, dev or test."""
-        if phase not in ['train', 'val', 'dev', 'test']:
-            raise ValueError(
-                "Unknown phase, which should be in ['train', 'val'/'dev', 'test']."
-            )
-        return self.num_examples[phase]
-
-    def get_train_examples(self):
-        """Gets a collection of `InputExample`s for the train set."""
-        return self.dataset.get_train_examples()
-
-    def get_dev_examples(self):
-        """Gets a collection of `InputExample`s for the dev set."""
-        return self.dataset.get_dev_examples()
-
-    def get_val_examples(self):
-        """Gets a collection of `InputExample`s for the val set."""
-        return self.dataset.get_val_examples()
-
-    def get_test_examples(self):
-        """Gets a collection of `InputExample`s for prediction."""
-        return self.dataset.get_test_examples()
-
-    def get_train_progress(self):
-        """Gets progress for training phase."""
-        return self.current_example, self.current_epoch
-
-    def data_generator(self,
-                       batch_size=1,
-                       phase="train",
-                       shuffle=False,
-                       data=None):
-        if phase != "predict" and not self.dataset:
-            raise ValueError("The dataset is None and it isn't allowed.")
-        if phase == "train":
-            shuffle = True
-            data = self.dataset.get_train_examples()
-            self.num_examples['train'] = len(data)
-        elif phase == "test":
-            shuffle = False
-            data = self.dataset.get_test_examples()
-            self.num_examples['test'] = len(data)
-        elif phase == "val" or phase == "dev":
-            shuffle = False
-            data = self.dataset.get_dev_examples()
-            self.num_examples['dev'] = len(data)
-        elif phase == "predict":
-            data = data
-        else:
-            raise ValueError(
-                "Unknown phase, which should be in ['train', 'dev', 'test'].")
-
-        def preprocess(text):
-            data_dict = {self.feed_key: [text]}
-            processed = self.lac.lexical_analysis(data=data_dict)
-            processed = [
-                self.vocab[word] for word in processed[0]['word']
-                if word in self.vocab
-            ]
-            if len(processed) == 0:
-                if six.PY2:
-                    text = text.encode(sys_stdout_encoding())
-                logger.warning(
-                    "The words in text %s can't be found in the vocabulary." %
-                    (text))
-            return processed
-
-        def _data_reader():
-            if shuffle:
-                np.random.shuffle(data)
-
-            if phase == "predict":
-                for text in data:
-                    text = preprocess(text)
-                    if not text:
-                        continue
-                    yield (text, )
-            else:
-                for item in data:
-                    text = preprocess(item.text_a)
-                    if not text:
-                        continue
-                    yield (text, item.label)
-
-        return paddle.batch(_data_reader, batch_size=batch_size)
-
-
-class MultiLabelClassifyReader(BaseReader):
+class MultiLabelClassifyReader(BaseNLPReader):
     def _pad_batch_records(self, batch_records, phase=None):
         batch_token_ids = [record.token_ids for record in batch_records]
         batch_text_type_ids = [record.text_type_ids for record in batch_records]
@@ -749,19 +605,13 @@ class MultiLabelClassifyReader(BaseReader):
                 label_ids.append(int(label))
 
         if phase != "predict":
-            Record = namedtuple(
-                'Record',
-                ['token_ids', 'text_type_ids', 'position_ids', 'label_ids'])
-
-            record = Record(
+            record = self.Record_With_Label_Id(
                 token_ids=token_ids,
                 text_type_ids=text_type_ids,
                 position_ids=position_ids,
-                label_ids=label_ids)
+                label_id=label_ids)
         else:
-            Record = namedtuple('Record',
-                                ['token_ids', 'text_type_ids', 'position_ids'])
-            record = Record(
+            record = self.Record_Wo_Label_Id(
                 token_ids=token_ids,
                 text_type_ids=text_type_ids,
                 position_ids=position_ids)
@@ -769,39 +619,7 @@ class MultiLabelClassifyReader(BaseReader):
         return record
 
 
-class RegressionReader(BaseReader):
-    def __init__(self,
-                 vocab_path,
-                 dataset=None,
-                 label_map_config=None,
-                 max_seq_len=128,
-                 do_lower_case=True,
-                 random_seed=None,
-                 use_task_id=False):
-        self.max_seq_len = max_seq_len
-        self.tokenizer = tokenization.FullTokenizer(
-            vocab_file=vocab_path, do_lower_case=do_lower_case)
-        self.vocab = self.tokenizer.vocab
-        self.dataset = dataset
-        self.pad_id = self.vocab["[PAD]"]
-        self.cls_id = self.vocab["[CLS]"]
-        self.sep_id = self.vocab["[SEP]"]
-        self.in_tokens = False
-        self.use_task_id = use_task_id
-
-        if self.use_task_id:
-            self.task_id = 0
-
-        np.random.seed(random_seed)
-
-        # generate label map
-        self.label_map = {}  # Unlike BaseReader, it's not filled
-
-        self.current_example = 0
-        self.current_epoch = 0
-
-        self.num_examples = {'train': -1, 'dev': -1, 'test': -1}
-
+class RegressionReader(BaseNLPReader):
     def _pad_batch_records(self, batch_records, phase=None):
         batch_token_ids = [record.token_ids for record in batch_records]
         batch_text_type_ids = [record.text_type_ids for record in batch_records]
@@ -881,7 +699,7 @@ class RegressionReader(BaseReader):
 
             for item in data:
                 # set label in order to run the program
-                label = -1  # different from BaseReader
+                label = -1  # different from BaseNLPReader
                 if len(item) == 1:
                     item_i = InputExample(
                         guid=seq_id, text_a=item[0], label=label)
@@ -956,7 +774,7 @@ class Features(object):
         return s
 
 
-class ReadingComprehensionReader(BaseReader):
+class ReadingComprehensionReader(BaseNLPReader):
     def __init__(self,
                  dataset,
                  vocab_path,
@@ -965,33 +783,30 @@ class ReadingComprehensionReader(BaseReader):
                  doc_stride=128,
                  max_query_length=64,
                  random_seed=None,
-                 use_task_id=False):
-        self.dataset = dataset
-        self.tokenizer = tokenization.FullTokenizer(
-            vocab_file=vocab_path, do_lower_case=do_lower_case)
-        self.max_seq_len = max_seq_len
+                 use_task_id=False,
+                 sp_model_path=None,
+                 word_dict_path=None,
+                 in_tokens=False):
+        super(ReadingComprehensionReader, self).__init__(
+            vocab_path=vocab_path,
+            dataset=dataset,
+            label_map_config=None,
+            max_seq_len=max_seq_len,
+            do_lower_case=do_lower_case,
+            random_seed=random_seed,
+            use_task_id=use_task_id,
+            sp_model_path=sp_model_path,
+            word_dict_path=word_dict_path,
+            in_tokens=in_tokens)
+
         self.doc_stride = doc_stride
         self.max_query_length = max_query_length
-        self.use_task_id = use_task_id
-        self.in_tokens = False
+        self._DocSpan = collections.namedtuple("DocSpan", ["start", "length"])
         # self.all_examples[phase] and self.all_features[phase] will be used
         # in write_prediction in reading_comprehension_task
         self.all_features = {"train": [], "dev": [], "test": [], "predict": []}
         self.all_examples = {"train": [], "dev": [], "test": [], "predict": []}
 
-        np.random.seed(random_seed)
-
-        self.vocab = self.tokenizer.vocab
-        self.vocab_size = len(self.vocab)
-        self.pad_id = self.vocab["[PAD]"]
-        self.cls_id = self.vocab["[CLS]"]
-        self.sep_id = self.vocab["[SEP]"]
-        self.mask_id = self.vocab["[MASK]"]
-
-        self.current_train_example = 0
-
-        self.num_examples = {'train': -1, 'dev': -1, 'test': -1}
-
     def _pad_batch_records(self, batch_records, phase):
         batch_token_ids = [record.token_ids for record in batch_records]
         batch_text_type_ids = [record.text_type_ids for record in batch_records]
@@ -1175,14 +990,14 @@ class ReadingComprehensionReader(BaseReader):
             # We can have documents that are longer than the maximum sequence length.
             # To deal with this we do a sliding window approach, where we take chunks
             # of the up to our max length with a stride of `doc_stride`.
-            _DocSpan = collections.namedtuple("DocSpan", ["start", "length"])
             doc_spans = []
             start_offset = 0
             while start_offset < len(all_doc_tokens):
                 length = len(all_doc_tokens) - start_offset
                 if length > max_tokens_for_doc:
                     length = max_tokens_for_doc
-                doc_spans.append(_DocSpan(start=start_offset, length=length))
+                doc_spans.append(
+                    self._DocSpan(start=start_offset, length=length))
                 if start_offset + length == len(all_doc_tokens):
                     break
                 start_offset += min(length, self.doc_stride)
@@ -1331,5 +1146,78 @@ class ReadingComprehensionReader(BaseReader):
         return cur_span_index == best_span_index
 
 
+class LACClassifyReader(BaseReader):
+    def __init__(self, vocab_path, dataset=None, in_tokens=False):
+        super(LACClassifyReader, self).__init__(dataset)
+        self.in_tokens = in_tokens
+
+        self.lac = hub.Module(name="lac")
+        self.tokenizer = tokenization.FullTokenizer(
+            vocab_file=vocab_path, do_lower_case=False)
+        self.vocab = self.tokenizer.vocab
+        self.feed_key = list(
+            self.lac.processor.data_format(
+                sign_name="lexical_analysis").keys())[0]
+
+    def data_generator(self,
+                       batch_size=1,
+                       phase="train",
+                       shuffle=False,
+                       data=None):
+        if phase != "predict" and not self.dataset:
+            raise ValueError("The dataset is None and it isn't allowed.")
+        if phase == "train":
+            shuffle = True
+            data = self.dataset.get_train_examples()
+            self.num_examples['train'] = len(data)
+        elif phase == "test":
+            shuffle = False
+            data = self.dataset.get_test_examples()
+            self.num_examples['test'] = len(data)
+        elif phase == "val" or phase == "dev":
+            shuffle = False
+            data = self.dataset.get_dev_examples()
+            self.num_examples['dev'] = len(data)
+        elif phase == "predict":
+            data = data
+        else:
+            raise ValueError(
+                "Unknown phase, which should be in ['train', 'dev', 'test'].")
+
+        def preprocess(text):
+            data_dict = {self.feed_key: [text]}
+            processed = self.lac.lexical_analysis(data=data_dict)
+            processed = [
+                self.vocab[word] for word in processed[0]['word']
+                if word in self.vocab
+            ]
+            if len(processed) == 0:
+                if six.PY2:
+                    text = text.encode(sys_stdout_encoding())
+                logger.warning(
+                    "The words in text %s can't be found in the vocabulary." %
+                    (text))
+            return processed
+
+        def _data_reader():
+            if shuffle:
+                np.random.shuffle(data)
+
+            if phase == "predict":
+                for text in data:
+                    text = preprocess(text)
+                    if not text:
+                        continue
+                    yield (text, )
+            else:
+                for item in data:
+                    text = preprocess(item.text_a)
+                    if not text:
+                        continue
+                    yield (text, item.label)
+
+        return paddle.batch(_data_reader, batch_size=batch_size)
+
+
 if __name__ == '__main__':
     pass