aimetrics / jarvis

Jarvis is a toolbox built on top of TensorFlow2.0 that allows developers and researchers to easily build neural networks in TensorFlow, particularly CTR models for large-scale advertising and recommendation scenarios.

Jarvis is a toolbox built on top of TensorFlow2.0 that allows developers and researchers to easily build neural networks in TensorFlow, particularly CTR models for large-scale advertising and recommendation scenarios. It provides the implementation of Meitu's FLEN model.

Note that Jarvis is still actively under development, so feedback and contributions are welcome. Feel free to submit your contributions as a pull request.

Jarvis features:

• Scalability: fast training on large-scale networks with tens of millions of sparse features
• Extensible: easily register new models and criteria.
• Supported tasks:
  • CTR prediction
  • Multi-task learning (coming)
  • online learning (todo)

Getting Started

Requirements and Installation

Please see environment.yml for more details

Usage

You can use python scripts/flen.py to run FLEN model on Avazu dataset.

Expected output:

Variant	AUC	Logloss
FLEN	0.7519	0.3944
FLEND	0.7528	0.3944

Avazu dataset

Download the tfrecord format dataset from here.
Alternatively, You can use python tools/dataset/avazu.py to prepare Avazu dataset yourself.

Customization

Implement Your Own Model

If you have a well-perform algorithm and are willing to implement it in our toolkit to help more people, you can create a pull request, detailed information can be found here.