EasyTPP Documentation

EasyTPP is an easy-to-use development and application toolkit for Neural Temporal Point Process (Neural TPP), with key features in configurability, compatibility and reproducibility. We hope this project could benefit both researchers and practitioners with the goal of easily customized development and open benchmarking.

GETTING STARTED

• Introduction
  • Framework
  • Install
  • Prepare Data
  • Model Training and Prediction
• Installation
  • Requirements
  • Install
• Quick Start
  • Download Dataset
  • Setup the configuration file
  • Train the Model

USER GUIDE

• Dataset
  • Required format
  • Data processing
  • Preprocessed Datasets
• Model Training
  • Step 1: Setup the config file containing data and model configs
  • Step 2: Run the training script
  • Checkout the output
• Model Prediction
  • Step 1: Setup the config file
  • Step 2: Run the evaluation script
  • Checkout the output

DEVELOPER GUIDE

• Model Customization
  • Create a new TPP Model Class
  • Rewrite Relevant Methods

ADVANCED TOPICS

• Thinning Algorithm
  • The implementation of the algorithm
  • One-step prediction
  • Multi-step prediction
• Tensorboard
  • Step 1: Activate the usage of tensorboard in Config file
  • Step 2: Launching the tensorboard
• Performance Benchmarks
  • Simulated dataset
• Implementation Details
  • Basic structure
  • Neural Hawkes Process (NHP)
  • Attentive Neural Hawkes Process (AttNHP)
  • Transformer Hawkes Process (THP)
  • Self-Attentive Hawkes Process (SAHP)
  • Recurrent Marked Temporal Point Processes (RMTPP)
  • Intensity Free Learning of Temporal Point Process (IntensityFree)
  • Fully Neural Network based Model for General Temporal Point Processes (FullyNN)
  • ODE-based Temporal Point Process (ODETPP)
  • Attentive Neural Hawkes Network (ANHN)

API REFERENCE

• Config
  • Config
  • DataConfig
  • DataSpecConfig
  • ModelConfig
  • BaseConfig
  • RunnerConfig
  • HPOConfig
  • HPORunnerConfig
• Preprocess
  • TPPDataLoader
  • EventTokenizer
  • TPPDataset
  • get_data_loader()
• Model
  • model.tf_model module
  • model.torch_model module
• Runner
  • Runner
  • TPPRunner
• Hyper-parameter Optimization
• Tf and Torch Wrapper
• Utilities
  • py_assert()
  • make_config_string()
  • create_folder()
  • save_yaml_config()
  • load_yaml_config()
  • RunnerPhase
  • LogConst
  • load_pickle()
  • has_key()
  • array_pad_cols()
  • MetricsHelper
  • MetricsTracker
  • set_device()
  • set_optimizer()
  • set_seed()
  • save_pickle()
  • count_model_params()
  • Registrable
  • Timer
  • concat_element()
  • get_stage()
  • to_dict()
  • parse_uri_to_protocol_and_path()
  • is_master_process()
  • is_local_master_process()
  • dict_deep_update()
  • DefaultRunnerConfig
  • rk4_step_method()
  • is_tf_available()
  • is_tensorflow_probability_available()
  • is_torchvision_available()
  • is_torch_cuda_available()
  • is_tf_gpu_available()
  • is_torch_gpu_available()
  • is_torch_available()
  • requires_backends()
  • PaddingStrategy
  • ExplicitEnum
  • TruncationStrategy
  • is_torch_device()
  • is_numpy_array()