DeepClassifier

DeepClassifier is a python package based on pytorch, which is easy-use and general for text classification task. You can install DeepClassifier by pip install -U deepclassifier。 If you want to know more information about DeepClassifier, please see the documentation. So let's start!🤩

If you think DeepClassifier is good, please star and fork it to give me motivation to continue maintenance！🤩 And it's my pleasure that if Deepclassifier is helpful to you!🥰

Installation

Just like other Python packages, DeepClassifier also can be installed through pip.The command of installation is pip install -U deepclassifier.

Models

Here is a list of models that have been integrated into DeepClassifier. In the future, we will integrate more models into DeepClassifier. Welcome to join us!🤩

1. TextCNN: Convolutional Neural Networks for Sentence Classification ,2014 EMNLP
2. RCNN: Recurrent Convolutional Neural Networks for Text Classification,2015,IJCAI
3. DPCNN: Deep Pyramid Convolutional Neural Networks for Text Categorization ,2017,ACL
4. HAN: Hierarchical Attention Networks for Document Classification, 2016,ACL
5. BERT: BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding,2018, ACL
6. BertTextCNN: BERT+TextCNN
7. BertRCNN: BERT+RCNN
8. BertDPCNN: BERT+DPCNN
9. BertHAN: BERT+HAN ...

Quick start

I wiil show you that how to use DeepClassifier below.🥰 Click [here] to display the complete code.

you can define model like that(take BertTextCNN model as example):👇

from deepclassifier.models import BertTextCNN

# parameters of model
embedding_dim = 768  # if you use bert, the default is 768.
dropout_rate = 0.2
num_class = 2
bert_path = "/Users/codewithzichao/Desktop/bert-base-uncased/"

my_model = BertTextCNN(embedding_dim=embedding_dim,
                       dropout_rate=dropout_rate,
                       num_class=num_class,
                       bert_path=bert_path)

optimizer = optim.Adam(my_model.parameters())
loss_fn = nn.CrossEntropyLoss()

After defining model, you can train/test/predict model like that:👇

from deepclassifier.trainers import Trainer

model_name = "berttextcnn"
save_path = "best.ckpt"
writer = SummaryWriter("logfie/1")
max_norm = 0.25
eval_step_interval = 20

my_trainer =Trainer(model_name=model_name,model=my_model,
                    train_loader=train_loader,dev_loader=dev_loader,
                    test_loader=test_loader, optimizer=optimizer, 
                    loss_fn=loss_fn,save_path=save_path, epochs=1, 
                    writer=writer, max_norm=max_norm,
                    eval_step_interval=eval_step_interval)

# training
my_trainer.train()
# print the best F1 value on dev set
print(my_trainer.best_f1)

# testing
p, r, f1 = my_trainer.test()
print(p, r, f1)

# predict
pred_data = DataLoader(pred_data, batch_size=1)
pred_label = my_trainer.predict(pred_data)
print(pred_label)

Contact me

If you want any questions about DeepClassifier, welcome to submit issue or pull requests! And welcome to communicate with me through [email protected].🥳

Citation

@misc{zichao2020deepclassifier,
author = {Zichao Li},
title = {DeepClassifier: use-friendly and flexiable package of NLP based text classification models},
year = {2020},
publisher = {GitHub},
journal = {GitHub Repository},
howpublished = {\url{https://github.com/codewithzichao/DeepClassifier}},
}