Skip to content

Latest commit

 

History

History
23 lines (12 loc) · 1.34 KB

docker.md

File metadata and controls

23 lines (12 loc) · 1.34 KB
Raw

To run the experiments inside a docker container, the following commands should be run in the docker directory.

First, install docker CE (engine) and nvidia-docker. If you want to run without sudo, [add yourself to the docker group](sudo usermod -a -G docker <user> && reboot). Otherwise run sudo before each command below:

docker build --network=host -t baseline -f Dockerfile ../
nvidia-docker run --net=host -v /data:/data:ro -it baseline bash

We assume all necessary files (datasests, embeddings etc.) are stored in /data/ in your machine.

For convenience, we also provide a build script, and a run script which can be run as ./run_docker.sh -g 0 -n <container_name>, eg. ./run_docker.sh -g 0 -n test. The script assumes that following directories exist in your machine:

  • /data/embeddings: location for embedding files (W2v, Glove), see embeddings.json in mead.

  • /data/datasets: dataset locations (PTB, CoNLL), see datasets.json in mead.

  • /data/model-checkpoints: directory to store saved models (model graph + data)

  • /data/model-store: directory to store exported models.