Modelpublic

pgilles/zipa-large-crctc-luxembourgish-ipapack20

sourceHugging Facemitupdated 2mo agoView on Hugging Face

0likes8downloads

Model Card

ZIPA Large CR-CTC — Luxembourgish + IPAPack++ ~20% (finetune)

Fine-tuned from `anyspeech/zipa-large-crctc-ns-800k` (ZIPA Zipformer CR-CTC large with diacritics).

Luxembourgish MFA-phone IPA ASR model (University of Luxembourg ASR group), trained on Luxembourgish phone data mixed with ~20% IPAPack++ multilingual phone replay.

Item	Value
Architecture	Zipformer large CR-CTC
Base model	`anyspeech/zipa-large-crctc-ns-800k` (with diacritics)
Output	Space-separated MFA phone IPA tokens
Tokenizer	`lb_phones.model`
IPAPack++ replay ratio	~0.20
License	MIT

Files

File	Description
`best-valid-loss.pt`	Lean PyTorch checkpoint (`model` / `model_avg`; no optimizer)
`config.json`	Hyperparameters and finetune metadata
`lb_phones.model` / `lb_phones.vocab`	SentencePiece phone tokenizer
`phones.json` / `phones.vocab`	Phone inventory helpers
`tokens.txt`	Token id table

License

MIT, matching upstream ZIPA.

Attribution

—ZIPA: https://github.com/lingjzhu/zipa
—Base: https://huggingface.co/anyspeech/zipa-large-crctc-ns-800k
—Fine-tuning: unilux