Bert for Multi-task Learning

What is it

This a project that uses BERT to do multi-task learning with multiple GPU support.

Why do I need this

In the original BERT code, neither multi-task learning or multiple GPU training is possible. Plus, the original purpose of this project is NER which dose not have a working script in the original BERT code.

To sum up, you can use this project if you:

Need multi-task learning
Need multiple GPU training
Need Sequence labeling

How to run pre-defined problems

How to use pretrained model

Pretrained models for CWS and NER are released. You can use simplified api defined in src/estimator_wrapper.py as follow.

    params = Params()
    m = ChineseNER(params, gpu=1)
    print(m.ner(['''正如老K教练所说，勒布朗詹姆斯的领袖气质开始凸显。面对外界的质疑，勒布朗表
示，“梦十队”一定会在伦敦奥运会上成功卫冕，“我们的球队不只想变得更强，而
且想拿到金牌，”詹姆斯说，“很多人都认为表示没法拿到金牌。他们认为我们在
身高上存在缺陷，说我们没有全身心地投入到国家队当中。但是我们会全身心地投
入，去迎接挑战。我现在已经迫不及待要出场比赛了。”''']))

How to train

This project is still in very early stage, and the available problems are quite limited. You can find available problems and baseline here

Avalable trained checkpoint

CWS (download trained checkpoint here)
NER (download trained checkpoint here)
CTBPOS (download trained checkpoint here)

Multitask Training

There are two types of chaining operations can be used to chain problems.

&. If two problems have the same inputs, they can be chained using &. Problems chained by & will be trained at the same time.
|. If two problems don't have the same inputs, they need to be chained using |. Problems chained by | will be sampled to train at every instance.

For example, CWS|NER|WeiboNER&WeiboSegment, one problem will be sampled at each turn, say WeiboNER&WeiboSegment, then WeiboNER and WeiboSegment will trained for this turn together.

You can train using the following command.

python main.py --problem "CWS|NER|WeiboNER&WeiboSegment" --schedule train --model_dir "tmp/multitask"

For evaluation, you need to separate the problems.

python main.py --problem "CWS" --schedule eval --model_dir "tmp/multitask"
python main.py --problem "NER" --schedule eval --model_dir "tmp/multitask"
python main.py --problem "WeiboNER&WeiboSegment" --schedule eval --model_dir "tmp/multitask"

How to add problems

Implement data preprocessing function and import it into src/data_preprocessing/__init__.py. One example can be found below.
Add problem config to self.problem_type and self.num_classes in src/params.py

def WeiboFakeCLS(params, mode):
    """Just a test problem to test multiproblem support

    Arguments:
        params {Params} -- params
        mode {mode} -- mode
    """
    tokenizer = FullTokenizer(vocab_file=params.vocab_file)

    inputs_list = [['科','技','全','方','位','资','讯','智','能','，','快','捷','的','汽','车','生','活','需','要','有','三','屏','一','云','爱','你'],
 ['对', '，', '输', '给', '一', '个', '女', '人', '，', '的', '成', '绩', '。', '失', '望']]
    target_list = [0, 1]

    label_encoder = get_or_make_label_encoder(
        'WeiboFakeCLS', mode, target_list, 0)

    return create_single_problem_generator('WeiboFakeCLS',
                                           inputs_list,
                                           new_target_list,
                                           label_encoder,
                                           params,
                                           tokenizer)

TODO

~~Add multiple GPU support AdamWeightDecayOptimizer~~
~~Add Pretraining~~
Add better evaluation
~~Add more ner problem~~

ohwe/bert-multitask-learning