Views
No views yet
Changes made: this is a format conversion only (NVIDIA NeMo.nemo→ MLXsafetensors, bf16). No retraining, fine-tuning or quantization — the weights are derived deterministically from the public NVIDIA checkpoint. Unofficial community conversion — not affiliated with or endorsed by NVIDIA.
/ AED model (text)
model.safetensors bf16, ~1.96 GB
config.json mlx-audio canary schema
tokenizer.model SentencePiece (16384, 25 langs)
ctc/ native timestamps (word/sentence alignment)
model.safetensors CTC-model FastConformer encoder (own fine-tuned weights), bf16
config.json, tokenizer.model
ctc_head.npz CTC head (W 16385×1024, b; blank id 16384)dw_striding ×8 subsampling, rel_pos attention. See the
original model card for training data, benchmarks
and full details.1from mlx_audio.stt.utils import load
2model = load("CogniSoftOrg/canary-1b-v2-mlx-bf16")
3res = model.generate("audio_16k.wav", source_lang="fr", target_lang="fr", use_pnc=True)
4print(res.text)ctc/ encoder + ctc_head.npz give native ~80 ms timestamps: run the AED for the text and
CTC-forced-align the AED tokens onto the CTC frames (WhisperX-style, all-Canary).scaled_dot_product_attention raises a dtype error.
Run the decoder in float32 (cast weights at load) or patch the two SDPA call sites to cast
the mask to q.dtype.language= kwarg is silently filtered by generate_transcription → pass source_lang /
target_lang directly to model.generate.no_repeat_ngram_size=3.ctc/ model recognizes fast but garbles far-field text (CTC limitation) — use
it for timing only, keep the AED for text.GOVERNING TERMS: Use of this model is governed by the CC-BY-4.0 license). Attribution
to NVIDIA is required; this conversion is provided as-is, without warranty.1@article{canary-1b-v2,
2 title={Canary-1B-v2 & Parakeet-TDT-0.6B-v3: Efficient and High-Performance Models for Multilingual ASR and AST},
3 journal={arXiv:2509.14128}, year={2025}
4}
5@article{granary,
6 title={Granary: Speech Recognition and Translation Dataset in 25 European Languages},
7 journal={arXiv:2505.13404}, year={2025}
8}