Thank you for sharing pre-trained embeddings for Wikidata5m at https://graphvite.io/docs/latest/pretrained_model.html. The current format for these pretrained weights is a set of pickle files, while I would expect that a flat vocabulary file (or json) together with a serialized numpy array would also allow the same functionality with more convenience and less safety risks.
Do you have, or could you make available such a copy of these embeddings?
Thank you!
Hello,
Thank you for sharing pre-trained embeddings for Wikidata5m at https://graphvite.io/docs/latest/pretrained_model.html. The current format for these pretrained weights is a set of pickle files, while I would expect that a flat vocabulary file (or json) together with a serialized numpy array would also allow the same functionality with more convenience and less safety risks.
Do you have, or could you make available such a copy of these embeddings? Thank you!