Code and data to support the paper "PAQ 65 Million Probably-Asked Questions andWhat You Can Do With Them"
Repo reproducing experimental results in "Addressing the Topological Defects of Disentanglement"
code for "AttentiveNAS Improving Neural Architecture Search via Attentive Sampling"
Codebase for paper "N-Bref A High-fidelity Decompiler Exploiting Programming Structures"
Replication code for "Causal Network Motifs Identifying Heterogeneous Spillover Effects in A/B Tests", TheWebConf2021.
A large-scale multilingual speech corpus for representation learning, semi-supervised learning and interpretation
Research and experimental code related to Opacus, an open-source library for training PyTorch models with Differential Privacy
RLStructures is a library to facilitate the implementation of new reinforcement learning algorithms. It includes a library, a tutorial, and different RL algorithms provided as examples.
Code accompanying paper, Forward Prediction for Physical Reasoning
NLG Best Practices for Data-Efficient Modeling How to Train Production-Ready Models with Little Data
The lightweight PyBullet wrapper for robotics researchers. Scale your research with less boilerplate.
Private computation framework library allows developers to perform randomized controlled trials, without leaking information about who participated or what action an individual took. It uses secure multiparty computation to guarantee this privacy. It is suitable for conducting A/B testing, or measuring advertising lift and learning the aggregate statistics without sharing information on the individual level.
COVID deterioration prediction based on chest X-ray radiographs via MoCo-trained image representations
This is a Tensor Train based compression library to compress sparse embedding tables used in large-scale machine learning models such as recommendation and natural language processing. We showed this library can reduce the total model size by up to 100x in Facebook’s open sourced DLRM model while achieving same model quality. Our implementation is faster than the state-of-the-art implementations. Existing the state-of-the-art library also decompresses the whole embedding tables on the fly therefore they do not provide memory reduction during runtime of the training. Our library decompresses only the requested rows therefore can provide 10,000 times memory footprint reduction per embedding table. The library also includes a software cache to store a portion of the entries in the table in decompressed format for faster lookup and process.
Applications using the GTN library and code to reproduce experiments in "Differentiable Weighted Finite-State Transducers"
