Emmental Tutorials

Introduction Tutorials

We have several introductory tutorials to help get you started with using Emmental.

Chexnet Tutorial

In this tutorial, we build an Emmental application to predicted 14 common diagnoses using convolutional neural networks in over 100,000 NIH chest x-rays proposed from the CheXNet paper.

GLUE Tutorial

In this tutorial, we build an Emmental application to tackle the General Language Understanding Evaluation (GLUE) benchmark which is a collection of resources for training, evaluating, and analyzing natural language understanding systems.

SuperGLUE Tutorial

In this tutorial, we build an Emmental application to tackle SuperGLUE, a new benchmark in the same style as GLUE with a set of more difficult language understanding tasks. Our submission achieved a new start-of-the-art score on June 15, 2019, under the name of Stanford Hazy Research. This code has also been refactored and used in the Snorkel project in the snorkel-superglue repository.

Data augmentation Tutorial

In this tutorial, we build an Emmental application to tackle image classification benchmarks (i.e. MNIST, CIFAR-10, and CIFAR-100) to study how to make data augmentation more efficient. More specifically, we propose an uncertainty-based random sampling scheme which, among the transformed data points, picks those with the highest losses, i.e. those "providing the most information".

Text classification Tutorial

In this tutorial, we build an Emmental application to tackle the text classification. More specifically, we analyze how information transfer among different tasks in multi-task learning settings.

Multilabel classification Tutorial

In this tutorial, we build an Emmental application to tackle the multilabel classification problem. More specifically, we identify and classify toxic online comments into six categories: toxic, severe_toxic, obscene, threat, insult, and identity_hate.

Installation

For the Python dependencies, we recommend using a virtualenv. Once you have cloned the repository, change directories to the root of the repository and run

virtualenv -p python3 .venv

Once the virtual environment is created, activate it by running:

source .venv/bin/activate

Any Python libraries installed will now be contained within this virtual environment. To deactivate the environment, simply run:

deactivate

Then, install Emmental and any other python dependencies by running:

pip install -r requirements.txt