/DS222-Poincare-Embedding

Distributed experiments on Poincare Embeddings using TensorFlow, PyTorch and HOROVOD.

Primary LanguagePython

Poincaré Embeddings for Learning Hierarchical Representations using Distributed frameworks

Implementation of Poincaré Embeddings for Learning Hierarchical Representations using distributed frameworks such as TensorFlow, PyTorch and HOROVOD.

pytorch/plots/mammal_closure.tsv_poincare_dim2_e1000.png

Installation

Simply clone this repository via

git clone https://github.com/sobalgi/DS222-Poincare-Embedding.git
cd DS222-Poincare-Embedding

Dependencies

  • Python 3 with NumPy
  • PyTorch
  • TensorFlow
  • Horovod
  • Scikit-Learn
  • NLTK (to generate the WordNet data)

Folders

For TensorFLow implementations

cd tensorflow

For PyTorch implementations

cd pytorch

For Numpy implementation

cd numpy

The steps to reproduce the results of the experiments are given in the respective readme files of the folders.

References

Original implementaion of Poincare Embeddings from Facebook AI Research. (with unresolved issues!!!)

FAIR code

Instructions to setup and use Horovod with PyTorch and TensorFlow.

HOROVOD