Commit History
-
remove TDT decoder conversion script (not CTC)
46daad3 -
add config.json, remove non-CTC scripts
- Restore config.json (model metadata: vocab_size, token IDs) - Remove convert/parakeet-tdt-v2-0.6b/ (TDT 0.6b, not CTC) - Remove scripts/run_benchmarks.py (general FluidAudio, not CTC) - Keep convert/parakeet-tdt-ctc-110m/ (CTC conversion scripts)20c78b6 -
restore convert/ and scripts/ directories (scripts only)
Keep conversion and benchmark Python scripts. Omit build artifacts: compiled models, audio files, plots, PDFs, .DS_Store, pyproject.toml, uv.lock, and intermediate outputs.a7fc9b6 -
cleanup: flatten repo to CTC-only models at root
Move AudioEncoder.mlmodelc, MelSpectrogram.mlmodelc, and vocab.json from parakeet-ctc-110m-coreml/ subdirectory to repo root. Remove everything not needed by the Swift FluidAudio library: - TDT model copies (parakeet-tdt-0.6b-v2-coreml/, parakeet-tdt-v2-0.6b/) - Conversion scripts (convert/, parakeet-tdt-ctc-110m/, scripts/) - CLI benchmarks (cli/) — already in FluidAudio repo - Duplicate model copies (models/) - Python artifacts (pyproject.toml, uv.lock) - HuggingFace tokenizer files (tokenizer.json, tokenizer.model, etc.) - .DS_Store files Matches the clean flat structure of parakeet-ctc-0.6b-coreml.8036e60
-
Upload 401 files
Verified signature26cfae7
