simplifies code by using high-level TF interfaces: tf.layers, tf.data, tf.estimator;

produces lots of clean Notebooks to make life easy;

can be run on CPU in reasonable time;

Python 3 and Jupyter Notebook are required

(CPU User) $ pip3 install tensorflow sklearn scipy bunch tqdm wget
(GPU User) $ pip3 install tensorflow-gpu sklearn scipy bunch tqdm wget

Penn Treebank(PTB 语料库)

• Skip-Gram Notebook

• CBOW Notebook

IMDB Movie Reviews Sentiment(电影评论情感分类)

• TF-IDF + LR Notebook

• Text-CNN Notebook

• Gated CNN Notebook

• Bi-RNN Notebook

• ConvLSTM Notebook

• Attention-Pooling Notebook

Language Model(语言模型)

• RNN LM Notebook

• Character-Aware LM Notebook

• Dilated-CNN LM Notebook

• Self-Attention LM Notebook

Speech Recognition(语音识别)

• CTC Notebook

User-Item Matching Folder

Question Matching Folder

POS Tagging(词性识别)

• Bi-RNN + CRF Notebook

• CNN + CRF Notebook

Chinese Segmentation(中文分词)

• Bi-RNN + CRF Notebook

• CNN + CRF Notebook

Learning to Sort(机器排序)

• Seq2Seq + Attention + Beam-Search Notebook

• Pointer Network Notebook

• Transformer Notebook

Learning to Dialog(机器对话)

• Transformer Notebook

Learning to Reconstruct(重建句子)

• VAE (RNN) Notebook

  • TF Distributions: Notebook

• VAE (Dilated-CNN) Notebook

  • TF Distributions: Notebook

• VAE (Self-Attention) Notebook

Learning to Control(定向重建句子)

• VAE + Discriminator Folder

Memory Network(记忆网络)

• End-to-End Memory Network Notebook

• Dynamic Memory Network Notebook

Knowledge Representation(知识表示)

• RDF + SPARQL Notebook (WN18)

Link Prediction (链路预测)

• DistMult (1-1 Scoring) Notebook (WN18)

• DistMult (1-N Scoring) Notebook (WN18)

• ConvE (1-N Scoring) Notebook (WN18)

Loading Data (导入数据)

• TFRecord: Fixed Length

• TFRecord: Padded Batch

Project Template(项目模版)

• TF Estimator Template

• TF Dataset Template