allennlp.tools.create_elmo_embeddings_from_vocab#

main#

main(vocab_path:str, elmo_config_path:str, elmo_weights_path:str, output_dir:str, batch_size:int, device:int, use_custom_oov_token:bool=False)

Creates ELMo word representations from a vocabulary file. These word representations are independent - they are the result of running the CNN and Highway layers of the ELMo model, but not the Bidirectional LSTM. ELMo requires 2 additional tokens: and . The first token in this file is assumed to be an unknown token.

This script produces two artifacts: A new vocabulary file with the and tokens inserted and a glove formatted embedding file containing word : vector pairs, one per line, with all values separated by a space.