# Embeddings¶

## Pretrained¶

Pretrained embeddings can be configured in the data section of the YAML configuration, e.g.:

data:
source_embedding:
path: data/glove/glove-100000.txt

SequenceToSequence models take a share_embeddings argument in the constructor that accepts a EmbeddingsSharingLevel value to enable different level of embeddings sharing.
See for example the custom model definition transformer_shared_embedding.py that shares all embeddings, including the output softmax weights, for a Transformer model.