An implementation of Meta's wav2vec 2.0 speech transcription using the Burn ML Framework.

This crate is a work in progress, and does not have a stable API yet.

Some tests in this repo require the test data in [test-data/], which is retrievable with git-annex. First, install git-annex, and then run:

git config annex.private true
git annex get test-data

Some files under test-data/ are only available via a git-annex "special remote". For example, to download from GitHub LFS:

git annex enableremote github
git annex get test-data

All tests should now complete when run. Consider running tests in --release mode, as burn runs very slowly with optimizations disabled.

cargo test --release --workspace

GitHub is the project's bug tracker. Please search for similar existing issues before submitting a new one.

Licensed under MIT.