Cape Cray Model

The Cape open source project has partnered with Cray Supercomputers to train even more accurate models.
This model is (as of time of release) the best in the world at full document machine reading for wikipedia trivia questions! https://competitions.codalab.org/competitions/17208#results.

Standard production model with top 100k glove embeddings

This archive contains the latest standard model with top 100k glove embeddings.
It is automatically downloaded and extracted when launching predictions with the default settings.
Including all embeddings might improve performance if your data has many uncommon tokens.
To download all 2.2M embeddings set env variable CAPE_DOWNLOAD_ALL_GLOVE_EMBEDDINGS=true.

Standard production model

This archive contains the latest standard model.
It is automatically downloaded and extracted when launching predictions with the default settings.