Repository navigation
Added
- Vocabularies are now kept at high level in
fasttextlt.fasttext.FastTextVocab, which allows you
to save and load them independently if you're keeping the weights somewhere else and are just
interested in getting the (sub)word ids. (fix: #2)
Changed
- We now assume standard sizes and little-endianness when loading a model and enforce it on save.
Full Changelog: v0.1.0...v0.2.0