language: python

sudo: enabled

python:

    - "3.6"

env:

    global:

     - TRAVIS=true

install:

    - pip install -r requirements.txt

script:
    python preprocess.py --corpus_path corpora/book_review_bert.txt --vocab_path models/google_zh_vocab.txt --dataset_path bert_dataset.pt --processes_num 8 --target bert --seq_length 64

    python pretrain.py --dataset_path bert_dataset.pt --vocab_path models/google_zh_vocab.txt --output_model_path models/bert_model.bin --config_path models/bert/mini_config.json --embedding word_pos_seg --encoder transformer --mask fully_visible --target bert --total_steps 10 --save_checkpoint_steps 10 --report_steps 2 --batch_size 2
    
    mv models/bert_model.bin-10 models/bert_model.bin

    python preprocess.py --corpus_path corpora/book_review.txt --vocab_path models/google_zh_vocab.txt --dataset_path roberta_dataset.pt --processes_num 8 --target mlm --seq_length 64 --dynamic_masking

    python pretrain.py --dataset_path roberta_dataset.pt --vocab_path models/google_zh_vocab.txt --output_model_path models/roberta_model.bin --config_path models/bert/mini_config.json --embedding word_pos_seg --encoder transformer --mask fully_visible --target mlm --total_steps 10 --save_checkpoint_steps 10 --report_steps 2 --batch_size 2
    
    mv models/roberta_model.bin-10 models/roberta_model.bin
    
    python preprocess.py --corpus_path corpora/book_review_bert.txt --vocab_path models/google_zh_vocab.txt --dataset_path albert_dataset.pt --processes_num 8 --target albert --seq_length 64

    python pretrain.py --dataset_path albert_dataset.pt --vocab_path models/google_zh_vocab.txt --output_model_path models/albert_model.bin --config_path models/albert/base_config.json --embedding word_pos_seg --encoder transformer --mask fully_visible --factorized_embedding_parameterization --parameter_sharing --target albert --total_steps 10 --save_checkpoint_steps 10 --report_steps 2 --batch_size 2
    
    mv models/albert_model.bin-10 models/albert_model.bin

    python preprocess.py --corpus_path corpora/book_review.txt --vocab_path models/google_zh_vocab.txt --dataset_path lm_dataset.pt --processes_num 8 --target lm --seq_length 64

    python pretrain.py --dataset_path lm_dataset.pt --vocab_path models/google_zh_vocab.txt --output_model_path models/gpt_model.bin --config_path models/bert/mini_config.json --embedding word_pos --encoder transformer --mask causal --target lm --total_steps 10 --save_checkpoint_steps 10 --report_steps 2 --batch_size 2 
    
    mv models/gpt_model.bin-10 models/gpt_model.bin

    python preprocess.py --corpus_path corpora/book_review_bert.txt --vocab_path models/google_zh_vocab.txt --dataset_path spanbert_dataset.pt --processes_num 8 --target bert --seq_length 64 --span_masking

    python pretrain.py --dataset_path spanbert_dataset.pt --vocab_path models/google_zh_vocab.txt --output_model_path models/spanbert_model.bin --config_path models/bert/mini_config.json --embedding word_pos_seg --encoder transformer --mask fully_visible --target bert --total_steps 10 --save_checkpoint_steps 10 --report_steps 2 --batch_size 2
    
    mv models/spanbert_model.bin-10 models/spanbert_model.bin

    python preprocess.py --corpus_path corpora/book_review_cls.txt --vocab_path models/google_zh_vocab.txt --dataset_path cls_dataset.pt --processes_num 8 --target cls --seq_length 64
    
    python pretrain.py --dataset_path cls_dataset.pt --vocab_path models/google_zh_vocab.txt --output_model_path models/cls_model.bin --config_path models/bert/mini_config.json --embedding word_pos_seg --encoder transformer --mask fully_visible --target cls --total_steps 10 --save_checkpoint_steps 10 --report_steps 2 --batch_size 2 --labels_num 2
    
    mv models/cls_model.bin-10 models/cls_model.bin

    python run_classifier.py --pretrained_model_path models/bert_model.bin --vocab_path models/google_zh_vocab.txt --config_path models/bert/mini_config.json --output_model_path models/classifier_model.bin --train_path datasets/test_data/chnsenticorp_test/train.tsv --dev_path datasets/test_data/chnsenticorp_test/dev.tsv --epochs_num 3 --batch_size 2 --embedding word_pos_seg --encoder transformer --mask fully_visible

    python inference/run_classifier_infer.py --load_model_path models/classifier_model.bin --vocab_path models/google_zh_vocab.txt --config_path models/bert/mini_config.json --test_path datasets/test_data/chnsenticorp_test/test_nolabel.tsv --prediction_path datasets/test_data/chnsenticorp_test/prediction.tsv --embedding word_pos_seg --encoder transformer --mask fully_visible --labels_num 2

    python run_classifier.py --pretrained_model_path models/albert_model.bin --vocab_path models/google_zh_vocab.txt --config_path models/albert/base_config.json --output_model_path models/classifier_model.bin --train_path datasets/test_data/chnsenticorp_test/train.tsv --dev_path datasets/test_data/chnsenticorp_test/dev.tsv --learning_rate 4e-5 --epochs_num 3 --batch_size 2 --embedding word_pos_seg --encoder transformer --mask fully_visible --factorized_embedding_parameterization --parameter_sharing

    python run_mt_classifier.py --pretrained_model_path models/bert_model.bin --vocab_path models/google_zh_vocab.txt --config_path models/bert/mini_config.json --dataset_path_list datasets/test_data/douban_test/ datasets/test_data/chnsenticorp_test/ --epochs_num 1 --batch_size 2 --embedding word_pos_seg --encoder transformer --mask fully_visible
    
    python run_ner.py --pretrained_model_path models/bert_model.bin --vocab_path models/google_zh_vocab.txt --config_path models/bert/mini_config.json --output_model_path models/ner_model.bin --train_path datasets/test_data/msra_ner_test/train.tsv --dev_path datasets/test_data/msra_ner_test/dev.tsv --label2id_path datasets/msra_ner/label2id.json --epochs_num 2 --batch_size 2 --embedding word_pos_seg --encoder transformer --mask fully_visible

    python inference/run_ner_infer.py --load_model_path models/ner_model.bin --vocab_path models/google_zh_vocab.txt --config_path models/bert/mini_config.json --test_path datasets/test_data/msra_ner_test/test_nolabel.tsv --prediction_path datasets/test_data/msra_ner_test/prediction.tsv --label2id_path datasets/msra_ner/label2id.json --embedding word_pos_seg --encoder transformer --mask fully_visible

    python run_cmrc.py --pretrained_model_path models/bert_model.bin --vocab_path models/google_zh_vocab.txt --config_path models/bert/mini_config.json --output_model_path models/cmrc_model.bin --train_path datasets/test_data/cmrc_test/train.json --dev_path datasets/test_data/cmrc_test/dev.json --epochs_num 2 --batch_size 2 --embedding word_pos_seg --encoder transformer --mask fully_visible --seq_length 128
    
    python inference/run_cmrc_infer.py --load_model_path models/cmrc_model.bin --vocab_path models/google_zh_vocab.txt --config_path models/bert/mini_config.json --test_path datasets/test_data/cmrc_test/test.json --prediction_path datasets/test_data/cmrc_test/prediction.json --embedding word_pos_seg --encoder transformer --mask fully_visible --seq_length 128