Modelpublic

willwade/byt5-p2g-multilingual-tiny

sourceHugging Facecc-by-sa-4.0updated 1mo agoView on Hugging Face

0likes36downloads

Model Card

ByT5 multilingual P2G (tiny) — 17M params

Phoneme-to-grapheme inversion of the harmonized retrain. Input: <lang>: phoneme tokens. Output: word spelling. Trained on a 4.12M-pair harmonized corpus.

Results (4k stratified test sample)

	score
micro exact	0.397
macro exact	0.557

Files

—HF-format weights at root (~70 MB)
—onnx/ — validated encoder+decoder pair

Licence

CC BY-SA 4.0