An Apache 2.0 NLP research library, built on PyTorch, for developing state-of-the-art deep learning models on a wide variety of linguistic tasks.

PLEASE NOTE that AllenNLP is currently pre-release! There are many rough edges. We will have our first release in early September.

The fastest way to get an environment to run AllenNLP is with Docker. Once you have installed Docker just run docker run -it --rm allennlp/allennlp-cpu to get an environment that will run on the cpu (use allennlp-gpu if you have a CUDA-supported GPU). Now you can do any of the following:

• Run a model on example sentences with allennlp/run bulk.
• Start a web service to host our models with allennlp/run serve.
• Interactively code against AllenNLP from the Python interpreter with python.

Built on PyTorch, AllenNLP makes it easy to design and evaluate new deep learning models for nearly any NLP problem, along with the infrastructure to easily run them in the cloud or on your laptop. AllenNLP was designed with the following principles:

• Hyper-modular and lightweight. Use the parts which you like seamlessly with PyTorch.
• Extensively tested and easy to extend. Test coverage is above 90% and the example models provide a template for contributions.
• Take padding and masking seriously, making it easy to implement correct models without the pain.
• Experiment friendly. Run reproducible experiments from a json specification with comprehensive logging.

AllenNLP includes reference implementations of high quality models for Semantic Role Labelling, Question and Answering (BiDAF), Entailment (decomposable attention), and more.

AllenNLP is built and maintained by the Allen Institute for Artificial Intelligence, in close collaboration with researchers at the University of Washington and elsewhere. With a dedicated team of best-in-field researchers and software engineers, the AllenNLP project is uniquely positioned to provide state of the art models with high quality engineering.

Conda will set up a virtual environment with the exact version of Python used for development along with all the dependencies needed to run AllenNLP.

1. Download and install Conda.

2. Create a Conda environment with Python 3.

   conda create -n allennlp python=3.5
   

3. Now activate the Conda environment.

   source activate allennlp
   

4. Install the required dependencies.

   INSTALL_TEST_REQUIREMENTS="true" ./scripts/install_requirements.sh
   

5. Visit http://pytorch.org/ and install the relevant pytorch package.

6. Set the PYTHONHASHSEED for repeatable experiments.

   export PYTHONHASHSEED=2157
   

You should now be able to test your installation with pytest -v. Congratulations!

Docker provides a virtual machine with everything set up to run AllenNLP--whether you will leverage a GPU or just run on a CPU. Docker provides more isolation and consistency, and also makes it easy to distribute your environment to a compute cluster.

It is easy to run a pre-built Docker development environment. AllenNLP is configured with Docker Cloud to build a new image on every update to the master branch. To download an image from Docker Hub:

docker pull allennlp/allennlp-cpu:latest

You can alternatively download an environment set up to use a GPU.

docker pull allennlp/allennlp-gpu:latest

Following are instructions on creating a Docker environment that use the CPU. To use the GPU, use the same instructions but substitute gpu for cpu. The following command will take some time, as it completely builds the environment needed to run AllenNLP.

docker build --file Dockerfile.cpu --tag allennlp/allennlp-cpu .

You should now be able to see this image listed by running docker images allennlp-cpu.

REPOSITORY          TAG                 IMAGE ID            CREATED             SIZE
allennlp/allennlp-cpu        latest              b66aee6cb593        5 minutes ago       2.38GB

You can run the image with docker run --rm -it allennlp/allennlp-cpu. The --rm flag cleans up the image on exit and the -it flags make the session interactive so you can use the bash shell the Docker image starts.

The Docker environment uses Conda to install Python and automatically enters the Conda environment "allennlp".

You can test your installation by running pytest -v.

Kubernetes will deploy your Docker images into the cloud, so you can have a reproducible development environment on AWS.

1. Set up kubectl to connect to your Kubernetes cluster.

2. Run kubectl create -f /path/to/kubernetes-dev-environment.yaml. This will create a "job" on the cluster which you can later connect to using bash. Note that you will be using the last Dockerfile that would pushed, and so the source code may not match what you have locally.

3. Retrieve the name of the pod created with kubectl describe job <JOBNAME> --namespace=allennlp. The pod name will be your job name followed by some additional characters.

4. Get a shell inside the container using kubectl exec -it <PODNAME> bash

5. When you are done, don't forget to kill your job using kubectl delete -f /path/to/kubernetes-dev-environment.yaml

AllenNLP is an open-source project backed by the Allen Institute for Artificial Intelligence (AI2). AI2 is a non-profit institute with the mission to contribute to humanity through high-impact AI research and engineering. To learn more about who specifically contributed to this codebase, see our contributors page.