initial commit

------------- LICENSE FOR huggingface(transformer) repository --------------

                                 Apache License

                           Version 2.0, January 2004

                        http://www.apache.org/licenses/

   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION

   1. Definitions.

      "License" shall mean the terms and conditions for use, reproduction,

      and distribution as defined by Sections 1 through 9 of this document.

      "Licensor" shall mean the copyright owner or entity authorized by

      the copyright owner that is granting the License.

      "Legal Entity" shall mean the union of the acting entity and all

      other entities that control, are controlled by, or are under common

      control with that entity. For the purposes of this definition,

      "control" means (i) the power, direct or indirect, to cause the

      direction or management of such entity, whether by contract or

      otherwise, or (ii) ownership of fifty percent (50%) or more of the

      outstanding shares, or (iii) beneficial ownership of such entity.

      "You" (or "Your") shall mean an individual or Legal Entity

      exercising permissions granted by this License.

      "Source" form shall mean the preferred form for making modifications,

      including but not limited to software source code, documentation

      source, and configuration files.

      "Object" form shall mean any form resulting from mechanical

      transformation or translation of a Source form, including but

      not limited to compiled object code, generated documentation,

      and conversions to other media types.

      "Work" shall mean the work of authorship, whether in Source or

      Object form, made available under the License, as indicated by a

      copyright notice that is included in or attached to the work

      (an example is provided in the Appendix below).

      "Derivative Works" shall mean any work, whether in Source or Object

      form, that is based on (or derived from) the Work and for which the

      editorial revisions, annotations, elaborations, or other modifications

      represent, as a whole, an original work of authorship. For the purposes

      of this License, Derivative Works shall not include works that remain

      separable from, or merely link (or bind by name) to the interfaces of,

      the Work and Derivative Works thereof.

      "Contribution" shall mean any work of authorship, including

      the original version of the Work and any modifications or additions

      to that Work or Derivative Works thereof, that is intentionally

      submitted to Licensor for inclusion in the Work by the copyright owner

      or by an individual or Legal Entity authorized to submit on behalf of

      the copyright owner. For the purposes of this definition, "submitted"

      means any form of electronic, verbal, or written communication sent

      to the Licensor or its representatives, including but not limited to

      communication on electronic mailing lists, source code control systems,

      and issue tracking systems that are managed by, or on behalf of, the

      Licensor for the purpose of discussing and improving the Work, but

      excluding communication that is conspicuously marked or otherwise

      designated in writing by the copyright owner as "Not a Contribution."

      "Contributor" shall mean Licensor and any individual or Legal Entity

      on behalf of whom a Contribution has been received by Licensor and

      subsequently incorporated within the Work.

   2. Grant of Copyright License. Subject to the terms and conditions of

      this License, each Contributor hereby grants to You a perpetual,

      worldwide, non-exclusive, no-charge, royalty-free, irrevocable

      copyright license to reproduce, prepare Derivative Works of,

      publicly display, publicly perform, sublicense, and distribute the

      Work and such Derivative Works in Source or Object form.

   3. Grant of Patent License. Subject to the terms and conditions of

      this License, each Contributor hereby grants to You a perpetual,

      worldwide, non-exclusive, no-charge, royalty-free, irrevocable

      (except as stated in this section) patent license to make, have made,

      use, offer to sell, sell, import, and otherwise transfer the Work,

      where such license applies only to those patent claims licensable

      by such Contributor that are necessarily infringed by their

      Contribution(s) alone or by combination of their Contribution(s)

      with the Work to which such Contribution(s) was submitted. If You

      institute patent litigation against any entity (including a

      cross-claim or counterclaim in a lawsuit) alleging that the Work

      or a Contribution incorporated within the Work constitutes direct

      or contributory patent infringement, then any patent licenses

      granted to You under this License for that Work shall terminate

      as of the date such litigation is filed.

   4. Redistribution. You may reproduce and distribute copies of the

      Work or Derivative Works thereof in any medium, with or without

      modifications, and in Source or Object form, provided that You

      meet the following conditions:

      (a) You must give any other recipients of the Work or

          Derivative Works a copy of this License; and

      (b) You must cause any modified files to carry prominent notices

          stating that You changed the files; and

      (c) You must retain, in the Source form of any Derivative Works

          that You distribute, all copyright, patent, trademark, and

          attribution notices from the Source form of the Work,

          excluding those notices that do not pertain to any part of

          the Derivative Works; and

      (d) If the Work includes a "NOTICE" text file as part of its

          distribution, then any Derivative Works that You distribute must

          include a readable copy of the attribution notices contained

          within such NOTICE file, excluding those notices that do not

          pertain to any part of the Derivative Works, in at least one

          of the following places: within a NOTICE text file distributed

          as part of the Derivative Works; within the Source form or

          documentation, if provided along with the Derivative Works; or,

          within a display generated by the Derivative Works, if and

          wherever such third-party notices normally appear. The contents

          of the NOTICE file are for informational purposes only and

          do not modify the License. You may add Your own attribution

          notices within Derivative Works that You distribute, alongside

          or as an addendum to the NOTICE text from the Work, provided

          that such additional attribution notices cannot be construed

          as modifying the License.

      You may add Your own copyright statement to Your modifications and

      may provide additional or different license terms and conditions

      for use, reproduction, or distribution of Your modifications, or

      for any such Derivative Works as a whole, provided Your use,

      reproduction, and distribution of the Work otherwise complies with

      the conditions stated in this License.

   5. Submission of Contributions. Unless You explicitly state otherwise,

      any Contribution intentionally submitted for inclusion in the Work

      by You to the Licensor shall be under the terms and conditions of

      this License, without any additional terms or conditions.

      Notwithstanding the above, nothing herein shall supersede or modify

      the terms of any separate license agreement you may have executed

      with Licensor regarding such Contributions.

   6. Trademarks. This License does not grant permission to use the trade

      names, trademarks, service marks, or product names of the Licensor,

      except as required for reasonable and customary use in describing the

      origin of the Work and reproducing the content of the NOTICE file.

   7. Disclaimer of Warranty. Unless required by applicable law or

      agreed to in writing, Licensor provides the Work (and each

      Contributor provides its Contributions) on an "AS IS" BASIS,

      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or

      implied, including, without limitation, any warranties or conditions

      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A

      PARTICULAR PURPOSE. You are solely responsible for determining the

      appropriateness of using or redistributing the Work and assume any

      risks associated with Your exercise of permissions under this License.

   8. Limitation of Liability. In no event and under no legal theory,

      whether in tort (including negligence), contract, or otherwise,

      unless required by applicable law (such as deliberate and grossly

      negligent acts) or agreed to in writing, shall any Contributor be

      liable to You for damages, including any direct, indirect, special,

      incidental, or consequential damages of any character arising as a

      result of this License or out of the use or inability to use the

      Work (including but not limited to damages for loss of goodwill,

      work stoppage, computer failure or malfunction, or any and all

      other commercial damages or losses), even if such Contributor

      has been advised of the possibility of such damages.

   9. Accepting Warranty or Additional Liability. While redistributing

      the Work or Derivative Works thereof, You may choose to offer,

      and charge a fee for, acceptance of support, warranty, indemnity,

      or other liability obligations and/or rights consistent with this

      License. However, in accepting such obligations, You may act only

      on Your own behalf and on Your sole responsibility, not on behalf

      of any other Contributor, and only if You agree to indemnify,

      defend, and hold each Contributor harmless for any liability

      incurred by, or claims asserted against, such Contributor by reason

      of your accepting any such warranty or additional liability.

   END OF TERMS AND CONDITIONS

   APPENDIX: How to apply the Apache License to your work.

      To apply the Apache License to your work, attach the following

      boilerplate notice, with the fields enclosed by brackets "[]"

      replaced with your own identifying information. (Don't include

      the brackets!)  The text should be enclosed in the appropriate

      comment syntax for the file format. We also recommend that a

      file or class name and description of purpose be included on the

      same "printed page" as the copyright notice for easier

      identification within third-party archives.

   Copyright [yyyy] [name of copyright owner]

   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.

Megatron is a large, powerful transformer. This repo is for ongoing research on training large, powerful transformer language models at scale. Currently, we support multinode training of [BERT](https://arxiv.org/pdf/1810.04805.pdf) in mixed precision. Our codebase is capable of training BERT Large on 64 V100 GPUs in 3 days. We achieved a final language modeling perplexity of 3.15 and SQuAD F1-score of 90.7.

# Setup

We officially support only python3.6.

To use this repo please install the latest supported versions of PyTorch with GPU support. 

Additionally, part of this codebase leverages tensorflow-cpu to perform dataloading of TFRecords. We recommend creating a virtual environment (to avoid breaking existing tf installations) and install our `reuirements.txt`.

```

python -m pip install virtualenv

virtualenv bert_env

source bert_env/bin/activate

pip install -r requirements.txt

```

# Usage

We've provided 4 scripts that pretrain BERT. All saved checkpoints can be used for finetuning according to [existing implementations](https://github.com/huggingface). Save model checkpoints with `--save`.

## BERT Pretraining

`bash scripts/pretrain_bert.sh`

This script runs single gpu BERT pretraining and is mainly for debugging purposes.

To use this script place your `--train-data` in loose json format with one json per line. The text field of your json dictionaries should correspond to `--text-key`. 

```

python pretrain_bert.py \

    --batch-size 4 \

    --tokenizer-type BertWordPieceTokenizer \

    --cache-dir temp_cache_dir \

    --tokenizer-model-type bert-large-uncased \

    --vocab-size 30522 \

    --train-data wikipedia \

    --loose-json \

    --text-key text \

    --split 1000,1,1 \

    --lazy-loader \

    --max-preds-per-seq 80 \

    --seq-length 512 \

    --max-position-embeddings 512 \

    --num-layers 24 \

    --hidden-size 1024 \

    --intermediate-size 4096 \

    --num-attention-heads 16 \

    --hidden-dropout 0.1 \

    --attention-dropout 0.1 \

    --train-iters 1000000 \

    --lr 0.0001 \

    --lr-decay-style linear \

    --lr-decay-iters 990000 \

    --warmup .01 \

    --weight-decay 1e-2 \

    --clip-grad 1.0 \

    --fp16 \

    --fp32-layernorm \

    --fp32-embedding \

    --hysteresis 2 \

    --num-workers 2 

```

## Distributed BERT Pretraining

`bash scripts/pretrain_bert_distributed.sh`

To use this script, follow the same data preparation procedure as in [earlier sections](#bert-pretraining). This script uses the pytorch distributed launcher to launch distributed training. As such, multinode training can be achieved by properly setting environment variables for the `env://` init method. See the official pytorch [documentation](https://pytorch.org/docs/stable/distributed.html#launch-utility) for further description of these [environment variables](https://pytorch.org/docs/stable/distributed.html#environment-variable-initialization). By default multinode training uses the nccl distributed backend.

## Distributed BERT Pretraining with TFRecords

`bash scripts/pretrain_bert_tfrecords_distributed.sh`

This script takes advantage of TensorFlow BERT's [`create_pretraining.py`](https://github.com/NVIDIA/DeepLearningExamples/blob/master/TensorFlow/LanguageModeling/BERT/create_pretraining_data.py) script to pre-cache the dataset in the TFRecord format. To convert the data to pytorch tensors we use a `TFRecordDataset` and tensorflow eager mode to turn the TFRecords into numpy matrices before loading them into pytorch gpu tensors. This greatly reduces the overhead of dataprocessing and speeds up training. Pass a whitespace-separated list of TFRecord paths to `--train-data` and enable the `--use-tfrecords` flag. Multinode training can be achieved as described in the [previous section](#distributed-bert-pretraining).

## Train Custom Sentence Piece Tokenizer and Pretrain BERT

`bash scripts/pretrain_bert_sentencepiece.sh`

This script runs BERT pretraining with a `sentencepiece` tokenizer. If no sentencepiece tokenizer exists at `--tokenizer-path` one will be trained automatically. The sentencepiece tokenizer can be used with the previous scripts (NOTE: sentencepiece training can only happen during single gpu pretraining). `<--tokenizer-path>.vocab` can be used with [`create_pretraining_data.py`](https://github.com/NVIDIA/DeepLearningExamples/blob/master/TensorFlow/LanguageModeling/BERT/create_pretraining_data.py) to make a TFRecord dataset with the given tokenization.

# Collecting Wikipedia Training Data

We recommend following the wikipedia data extraction process specified by google research: "the recommended pre-processing is to download [the latest dump](https://dumps.wikimedia.org/enwiki/latest/enwiki-latest-pages-articles.xml.bz2), extract the text with [WikiExtractor.py](https://github.com/attardi/wikiextractor), and then apply any necessary cleanup to convert it into plain text." 

We recommend using the `--json` argument when using WikiExtractor, which will dump the wikipedia data into loose json format (one json per line), making it more manageable and readily consumable by our codebase.

Once the json dataset is ready make sure to set the path in line 27 of `data_utils/corpora.py`.

If your system is memory limited we also recommend running pretraining with the `--lazy-loader` argument as we've done. After preprocessing the dataset once, this will allow the dataset to be lazily loaded from disk, as opposed to storing it in memory.

# coding=utf-8

# Copyright (c) 2019, NVIDIA CORPORATION.  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.

"""utils for creating datasets"""

import os

import math

from .samplers import DistributedBatchSampler

from .datasets import json_dataset, csv_dataset, split_ds, ConcatDataset, SplitDataset, bert_sentencepair_dataset

from .lazy_loader import exists_lazy, make_lazy, lazy_array_loader

from .tokenization import Tokenization, CommandToken, Tokenizer, CharacterLevelTokenizer, BertWordPieceTokenizer, make_tokenizer

from . import corpora

TRAIN_DATA = 0

VAL_DATA = 1

TEST_DATA = 2

def should_split(split):

    """

    given split proportions checks if should split

    Examples:

    >>> should_split([10,0,0]) 

    False

    >>> should_split([1,.1,.2])

    True

    """

    return max(split)/sum(split) != 1.

def get_ext(path):

    """gets path extension"""

    return os.path.splitext(path)[1]

def get_dataset(path, **kwargs):

    """gets dataset object based on keyword args and file at `path`"""

    if supported_corpus(path):

        return corpora.NAMED_CORPORA[path](**kwargs)

    ext = get_ext(path)

    if ext =='.json':

        text = json_dataset(path, **kwargs)

    elif ext in ['.csv', '.tsv']:

        text = csv_dataset(path, **kwargs)

    else:

        raise NotImplementedError('data file type %s is not supported'%(ext))

    return text

def supported_corpus(corpus_name):

    """checks if corpus name is defined in `corpora.py`"""

    return corpus_name in corpora.NAMED_CORPORA

def make_dataset(path, seq_length, text_key, label_key, lazy=False, process_fn=None, split=[1.],

                delim=',', loose=False, binarize_sent=False, drop_unlabeled=False, tokenizer=None,

                tokenizer_type='CharacterLevelTokenizer', tokenizer_model_path=None, vocab_size=None,

                model_type='bpe', pad_token=0, character_converage=1.0, non_binary_cols=None, **kwargs):

    """function to create datasets+tokenizers for common options"""

    if isinstance(process_fn, str):

        process_fn = eval(process_fn)

    if non_binary_cols is not None:

        # multilabel dataset support (only for csvs)

        label_key = non_binary_cols

    def get_dataset_from_path(path_):

        if lazy:

            # get lazily loaded dataset

            named_corpora = False

            if supported_corpus(path_):

                named_corpora = True

                name = path_

                path_ = corpora.NAMED_CORPORA[path_].PATH

            if not exists_lazy(path_, data_type='data'):

                # create cached version of dataset for lazy loading if it doesn't exist

                text = get_dataset(name if named_corpora else path_, text_key=text_key, label_key=label_key, binarize_sent=binarize_sent,

                    delim=delim, drop_unlabeled=drop_unlabeled, loose_json=loose)

                make_lazy(path_, text.X, data_type='data')

            text = lazy_array_loader(path_, data_type='data', map_fn=process_fn)

        else:

            # get dataset

            text = get_dataset(path_, text_key=text_key, label_key=label_key, binarize_sent=binarize_sent,

                    delim=delim, drop_unlabeled=drop_unlabeled, loose_json=loose, preprocess_fn=process_fn)

        return text

    # get one or multiple datasets and concatenate

    if isinstance(path, str):

        path = [path]

    datasets = [get_dataset_from_path(p) for p in path]

    if len(datasets) == 1:

        ds = datasets[0]

    else:

        ds = ConcatDataset(datasets)

    # make tokenizer for dataset

    if tokenizer is None:

        tokenizer = make_tokenizer(tokenizer_type, ds, tokenizer_model_path, vocab_size, model_type, 

                                    pad_token, character_converage, **kwargs)

    ds_type = ''

    if 'ds_type' in kwargs:

        ds_type = kwargs['ds_type']

    ds.SetTokenizer(tokenizer)

    # Split dataset into train/val/test (and wrap bert dataset)

    if should_split(split):

        ds = split_ds(ds, split)

        if ds_type.lower() == 'bert':

            ds = [bert_sentencepair_dataset(d, max_seq_len=seq_length) for d in ds]

    else:

        if ds_type.lower() == 'bert':

            ds = bert_sentencepair_dataset(ds, max_seq_len=seq_length)

    return ds, tokenizer