How to use custom word embeddings?

[havardox]

Is it possible to use custom pre-trained word embeddings? The current ones all seem to be in English and I want to load embeddings in other languages.

Relevant section of docs: https://entity-embed.readthedocs.io/en/latest/guide/field_types.html#semantic-fields

[fjsj]

Yes it is if you get ones compatible with torchtext Vocab. The version of torchtext that the project uses is not the latest one, that's the issue I think.

Also note the semantic embedding part of the model is mostly straightforward:

entity-embed/entity_embed/models.py

Lines 55 to 66 in 1bd9223

	class SemanticEmbedNet(nn.Module):
	def __init__(self, field_config, embedding_size):
	super().__init__()
	self.embedding_size = embedding_size
	self.dense_net = nn.Sequential(
	nn.Embedding.from_pretrained(field_config.vocab.vectors),
	nn.Dropout(p=field_config.embed_dropout_p),
	)
	def forward(self, x, **kwargs):
	return self.dense_net(x)

You may switch that for your own semantic embedding and train as you wish.