cstr/punctuate-all-GGUF

Punctuate-All (GGUF)

GGUF conversion of kredor/punctuate-all for use with CrispASR.

Adds punctuation to unpunctuated ASR output. 12 languages with ASCII punctuation output. Smaller and faster alternative to fullstop-punc-multilang (base vs large).

Model Details

• Architecture: XLM-RoBERTa-base — 12L, d=768, 12 heads, d_ffn=3072, GELU
• Parameters: ~278M
• Classifier: Linear(768, 6) — 6 punctuation classes
• Labels: none, . (period), , (comma), ? (question), - (dash), : (colon)
• Vocabulary: SentencePiece (250,002 tokens)
• Max sequence: 512 tokens (auto-chunked)
• Languages: en, de, fr, es, bg, it, pl, nl, cs, pt, sk, sl
• License: MIT

Usage with CrispASR

crispasr --backend wav2vec2 -m wav2vec2.gguf --punc-model punctuate-all-q4_k.gguf -f audio.wav

Available Files

File	Quant	Size	Description
punctuate-all-f16.gguf	F16	901 MB	Half precision
punctuate-all-q4_k.gguf	Q4_K	154 MB	Recommended

Original Model

• Source: kredor/punctuate-all
• Training: WMT/Europarl dataset

Provenance and EU AI Act Art. 53 note

• Upstream model: kredor/punctuate-all — published by kredor.
• Upstream licence: mit. This repository redistributes under the same terms; it grants no rights the upstream licence does not.
• What was done here: format conversion and/or quantisation only (GGUF/GGML). No training, no fine-tuning, no merging, no distillation, no change to architecture, vocabulary or capability. Only the numeric representation of the upstream weights differs.
• Training data: documented — where it is documented at all — by the upstream provider; see the upstream model card. No training data was used, added or selected by this repository.
• Provider status: under Regulation (EU) 2024/1689 the upstream authors remain the provider of this model. Converting the serialisation format does not make this repository the provider of a new general-purpose AI model, and no such claim is made. Questions about training content, copyright policy or model capability belong upstream.