[GitHub] szha commented on a change in pull request #9986: gluon language modeling dataset and text token reader
szha commented on a change in pull request #9986: gluon language modeling dataset and text token reader URL: https://github.com/apache/incubator-mxnet/pull/9986#discussion_r173085151 ## File path: python/mxnet/gluon/data/text/base.py ## @@ -0,0 +1,99 @@ +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license agreements. See the NOTICE file +# distributed with this work for additional information +# regarding copyright ownership. The ASF licenses this file +# to you 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. + +# coding: utf-8 +# pylint: disable= + +"""Base classes for text datasets and readers.""" + +__all__ = ['WordLanguageReader'] + +import io +import os + +from ..dataset import SimpleDataset +from ..datareader import DataReader +from .utils import flatten_samples, collate, pair + +class WordLanguageReader(DataReader): +"""Text reader that reads a whole corpus and produces samples based on provided sample splitter +and word tokenizer. + +Parameters +-- +filename : str +Path to the input text file. +encoding : str, default 'utf8' +File encoding format. +sample_splitter : function, default str.splitlines +A function that splits the dataset string into samples. +tokenizer : function, default str.split +A function that splits each sample string into list of tokens. +seq_len : int or None +The length of each of the samples. If None, samples are divided according to +`sample_splitter` only, and may have variable lengths. +bos : str or None, default None +The token to add at the begining of each sentence. If None, nothing is added. +eos : str or None, default None +The token to add at the end of each sentence. If None, nothing is added. +pad : str or None, default None +The padding token to add at the end of dataset if `seq_len` is specified and the total +number of tokens in the corpus don't evenly divide `seq_len`. If pad is None or seq_len +is None, no padding is added. Otherwise, padding token is added to the last sample if +its length is less than `seq_len`. If `pad` is None and `seq_len` is specified, the last +sample is discarded if it's shorter than `seq_len`. Review comment: To illustrate, I'd need to add equal number of examples. I'm not sure if further verbosity really helps. This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] szha commented on a change in pull request #9986: gluon language modeling dataset and text token reader
szha commented on a change in pull request #9986: gluon language modeling dataset and text token reader URL: https://github.com/apache/incubator-mxnet/pull/9986#discussion_r173084963 ## File path: python/mxnet/gluon/data/text/base.py ## @@ -0,0 +1,99 @@ +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license agreements. See the NOTICE file +# distributed with this work for additional information +# regarding copyright ownership. The ASF licenses this file +# to you 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. + +# coding: utf-8 +# pylint: disable= + +"""Base classes for text datasets and readers.""" + +__all__ = ['WordLanguageReader'] + +import io +import os + +from ..dataset import SimpleDataset +from ..datareader import DataReader +from .utils import flatten_samples, collate, pair + +class WordLanguageReader(DataReader): +"""Text reader that reads a whole corpus and produces samples based on provided sample splitter +and word tokenizer. + +Parameters +-- +filename : str +Path to the input text file. +encoding : str, default 'utf8' +File encoding format. +sample_splitter : function, default str.splitlines +A function that splits the dataset string into samples. +tokenizer : function, default str.split +A function that splits each sample string into list of tokens. +seq_len : int or None +The length of each of the samples. If None, samples are divided according to +`sample_splitter` only, and may have variable lengths. +bos : str or None, default None +The token to add at the begining of each sentence. If None, nothing is added. +eos : str or None, default None +The token to add at the end of each sentence. If None, nothing is added. +pad : str or None, default None +The padding token to add at the end of dataset if `seq_len` is specified and the total +number of tokens in the corpus don't evenly divide `seq_len`. If pad is None or seq_len Review comment: Yes if specified. This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] szha commented on a change in pull request #9986: gluon language modeling dataset and text token reader
szha commented on a change in pull request #9986: gluon language modeling dataset and text token reader URL: https://github.com/apache/incubator-mxnet/pull/9986#discussion_r172323412 ## File path: python/mxnet/gluon/data/datareader.py ## @@ -0,0 +1,33 @@ +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license agreements. See the NOTICE file +# distributed with this work for additional information +# regarding copyright ownership. The ASF licenses this file +# to you 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. + +# coding: utf-8 +# pylint: disable= +"""Dataset reader.""" +__all__ = ['DataReader'] + +class DataReader(object): Review comment: Shall we make a data reader a generator? (i.e. define `__iter__`) This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] szha commented on a change in pull request #9986: gluon language modeling dataset and text token reader
szha commented on a change in pull request #9986: gluon language modeling dataset and text token reader URL: https://github.com/apache/incubator-mxnet/pull/9986#discussion_r172323412 ## File path: python/mxnet/gluon/data/datareader.py ## @@ -0,0 +1,33 @@ +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license agreements. See the NOTICE file +# distributed with this work for additional information +# regarding copyright ownership. The ASF licenses this file +# to you 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. + +# coding: utf-8 +# pylint: disable= +"""Dataset reader.""" +__all__ = ['DataReader'] + +class DataReader(object): Review comment: Shall we make a data reader a generator? (i.e. `define __iter__`) This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] szha commented on a change in pull request #9986: gluon language modeling dataset and text token reader
szha commented on a change in pull request #9986: gluon language modeling dataset and text token reader URL: https://github.com/apache/incubator-mxnet/pull/9986#discussion_r172315558 ## File path: python/mxnet/gluon/data/text/_constants.py ## @@ -0,0 +1,26 @@ +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license agreements. See the NOTICE file +# distributed with this work for additional information +# regarding copyright ownership. The ASF licenses this file +# to you 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. + +# coding: utf-8 + +"""Constants relevant to text processing.""" + +UNK_TOKEN = '' + +BOS_TOKEN = '' Review comment: > Isn't this usually ``? I see more use of ``, though honestly it would just be a preference, as long as we can agree on one. This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] szha commented on a change in pull request #9986: gluon language modeling dataset and text token reader
szha commented on a change in pull request #9986: gluon language modeling dataset and text token reader URL: https://github.com/apache/incubator-mxnet/pull/9986#discussion_r172317063 ## File path: python/mxnet/gluon/data/text/lm.py ## @@ -0,0 +1,157 @@ +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license agreements. See the NOTICE file +# distributed with this work for additional information +# regarding copyright ownership. The ASF licenses this file +# to you 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. + +# coding: utf-8 +# pylint: disable= +"""Language model datasets.""" + +__all__ = ['LanguageModelDataset', 'WikiText2', 'WikiText103'] + +import os +import zipfile +import shutil + +from .base import TextDataset, TextTokenReader +from . import _constants as C +from ...utils import download, check_sha1, get_repo_file_url + + +class LanguageModelDataset(TextDataset): Review comment: It's common for language modeling to preprocess arbitrary corpus into sequences of fixed number of tokens (which I call `collate` here). I put this preprocessing step in LangaugeModelDataset as it's not common for other datasets. An alternative is to treat such preprocessing function as util function. Would that be preferred? This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] szha commented on a change in pull request #9986: gluon language modeling dataset and text token reader
szha commented on a change in pull request #9986: gluon language modeling dataset and text token reader URL: https://github.com/apache/incubator-mxnet/pull/9986#discussion_r172316324 ## File path: python/mxnet/gluon/data/text/base.py ## @@ -0,0 +1,89 @@ +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license agreements. See the NOTICE file +# distributed with this work for additional information +# regarding copyright ownership. The ASF licenses this file +# to you 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. + +# coding: utf-8 +# pylint: disable= + +"""Base classes for text datasets and readers.""" + +__all__ = ['TextTokenReader'] + +import io +import os + +from ..dataset import Dataset +from ..datareader import DataReader + +class TextTokenReader(DataReader): +"""Text reader that produces lists of tokens. + +Parameters +-- +filename : str +Path to the input text file. +encoding : str, default 'utf8' +File encoding format. +sample_splitter : function, default str.splitlines +A function that splits the dataset string into samples. +tokenizer : function, default str.split +A function that splits each sample string into list of tokens. +bos : str or None, default None +The token to add at the begining of each sentence. If None, nothing is added. +eos : str or None, default None +The token to add at the end of each sentence. If None, nothing is added. +""" +def __init__(self, filename, encoding='utf8', sample_splitter=lambda s: s.splitlines(), + tokenizer=lambda s: s.split(), bos=None, eos=None): +filename = os.path.expanduser(filename) +with io.open(filename, 'r', encoding=encoding) as fin: +content = fin.read() + +samples = [s for s in [tokenizer(x) for x in sample_splitter(content)] if s] +if bos or eos: +for tokens in samples: +if bos: +tokens.insert(0, bos) +if eos: +tokens.append(eos) + +self._samples = samples + +def __len__(self): +return self._samples.__len__ + +def __getitem__(self, idx): +return self._samples[idx] + +class TextDataset(Dataset): +"""Abstract dataset class for text data. + +Subclasses need to override `__getitem__`, which returns the i-th +element, and `__len__`, which returns the total number elements.""" + +def __getitem__(self, idx): +raise NotImplementedError + +def __len__(self): +raise NotImplementedError + +def _flatten(self, samples): Review comment: I will organize these as utils and get rid of TextDataset then. This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] szha commented on a change in pull request #9986: gluon language modeling dataset and text token reader
szha commented on a change in pull request #9986: gluon language modeling dataset and text token reader URL: https://github.com/apache/incubator-mxnet/pull/9986#discussion_r172316070 ## File path: python/mxnet/gluon/data/text/base.py ## @@ -0,0 +1,89 @@ +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license agreements. See the NOTICE file +# distributed with this work for additional information +# regarding copyright ownership. The ASF licenses this file +# to you 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. + +# coding: utf-8 +# pylint: disable= + +"""Base classes for text datasets and readers.""" + +__all__ = ['TextTokenReader'] + +import io +import os + +from ..dataset import Dataset +from ..datareader import DataReader + +class TextTokenReader(DataReader): +"""Text reader that produces lists of tokens. Review comment: For this dataset, word tokenization happens in the data reader because the result of tokenization could decide the sample boundary. It's also possible to defer the chunking-by-length to the dataset class. I will change it to the later and see how it is. This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] szha commented on a change in pull request #9986: gluon language modeling dataset and text token reader
szha commented on a change in pull request #9986: gluon language modeling dataset and text token reader URL: https://github.com/apache/incubator-mxnet/pull/9986#discussion_r172315558 ## File path: python/mxnet/gluon/data/text/_constants.py ## @@ -0,0 +1,26 @@ +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license agreements. See the NOTICE file +# distributed with this work for additional information +# regarding copyright ownership. The ASF licenses this file +# to you 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. + +# coding: utf-8 + +"""Constants relevant to text processing.""" + +UNK_TOKEN = '' + +BOS_TOKEN = '' Review comment: > Isn't this usually ``? I see more use of ``, though honestly it would just be a preference, as long as we can agree on one. This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] szha commented on a change in pull request #9986: gluon language modeling dataset and text token reader
szha commented on a change in pull request #9986: gluon language modeling dataset and text token reader URL: https://github.com/apache/incubator-mxnet/pull/9986#discussion_r172314994 ## File path: python/mxnet/gluon/data/text/__init__.py ## @@ -0,0 +1,24 @@ +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license agreements. See the NOTICE file +# distributed with this work for additional information +# regarding copyright ownership. The ASF licenses this file +# to you 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. + +# coding: utf-8 +# pylint: disable=wildcard-import +"""Text utilities.""" + +from .base import * + +from . import lm Review comment: I thought it would be a good idea to separate the datasets by application. The reason is that some dataset names are common among more than one tasks, such as semeval datasets (e.g. semeval 2017 dataset consists of three tasks on the same set of data). This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] szha commented on a change in pull request #9986: gluon language modeling dataset and text token reader
szha commented on a change in pull request #9986: gluon language modeling dataset and text token reader URL: https://github.com/apache/incubator-mxnet/pull/9986#discussion_r172314482 ## File path: python/mxnet/gluon/data/datareader.py ## @@ -0,0 +1,33 @@ +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license agreements. See the NOTICE file +# distributed with this work for additional information +# regarding copyright ownership. The ASF licenses this file +# to you 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. + +# coding: utf-8 +# pylint: disable= +"""Dataset reader.""" +__all__ = ['DataReader'] + +class DataReader(object): Review comment: (e.g. SemEval2017 reader can read the dataset, while different dataset classes can treat them as datasets for different tasks, such as POS tagging and keyphrase extraction) This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] szha commented on a change in pull request #9986: gluon language modeling dataset and text token reader
szha commented on a change in pull request #9986: gluon language modeling dataset and text token reader URL: https://github.com/apache/incubator-mxnet/pull/9986#discussion_r172314053 ## File path: python/mxnet/gluon/data/datareader.py ## @@ -0,0 +1,33 @@ +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license agreements. See the NOTICE file +# distributed with this work for additional information +# regarding copyright ownership. The ASF licenses this file +# to you 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. + +# coding: utf-8 +# pylint: disable= +"""Dataset reader.""" +__all__ = ['DataReader'] + +class DataReader(object): Review comment: OK. At first I separated those classes by division of concern instead of interface. I agree that it would be better if the interfaces differ too. This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services