cstr/wav2vec2-large-xlsr-53-german-GGUF

Wav2Vec2 Large XLSR-53 German -- GGUF

GGUF conversions and quantisations of jonatasgrosman/wav2vec2-large-xlsr-53-german for use with CrispStrobe/CrispASR.

Available variants

File	Quant	Size	Notes
wav2vec2-large-xlsr-53-german-q4_k.gguf	Q4_K	~222 MB	Best size/quality tradeoff

Model details

• Architecture: Wav2Vec2ForCTC — CNN feature extractor + 24L transformer encoder (1024d, 16 heads, stable (pre-norm)) + CTC head
• Parameters: 315M
• Language: German
• Vocab: 38 characters (CTC greedy decode)
• License: APACHE-2.0
• Source: jonatasgrosman/wav2vec2-large-xlsr-53-german
• Fine-tuned on German CommonVoice. Most popular German wav2vec2 model (12.9K downloads).

Usage with CrispASR

./build/bin/crispasr --backend wav2vec2 -m wav2vec2-large-xlsr-53-german-q4_k.gguf -f german_audio.wav -l de

# Auto-download (default German model):
./build/bin/crispasr --backend wav2vec2 -m auto --auto-download -l de -f audio.wav

Provenance and EU AI Act Art. 53 note

• Upstream model: jonatasgrosman/wav2vec2-large-xlsr-53-german — published by jonatasgrosman.
• Upstream licence: apache-2.0. 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.