license: cc-by-4.0
language:
- en
- he
base_model: Hebrish/Sabich
base_model_relation: quantized
pipeline_tag: automatic-speech-recognition
library_name: mlx
tags:
- mlx
- parakeet
- code-switching
- hebrew
- english
- automatic-speech-recognition
extra_gated_prompt: >-
Access to this model is by request and is governed by the Hebrish terms of use
(https://hebrish.app/terms). Non-commercial use is free; commercial use needs
a commercial licence from access-requests@hebrish.app. Please also complete
the attestation form: https://hebrish.app/legal/hebrish-terms-of-use.pdf
extra_gated_fields:
Name: text
Organisation: text
Intended use: text
I agree to use this model under the Hebrish terms of use (non-commercial unless I hold a commercial licence): checkbox
Sabich v1.0: MLX
MLX weights for Apple silicon of Sabich v1.0, the kitchen-sink Parakeet TDT 0.6B v2 fine-tune: English-Hebrew code-switching plus tech vocabulary, brand names, numbers in digits and Irish-English phrasing, in one
speaker's voice. Hebrew words come out in Latin romanization. Results, training data and limitations
are on the main card. This build is Sabich v1.0 (run r7-all); it is replaced when the line has a new version.
Website: hebrish.app. Use of this model and every converted or quantised build of it is governed by the Hebrish terms of use. Access is by request: use the button on this page, or write to access-requests@hebrish.app. Non-commercial use (personal, research, education, open-source projects) is free; please complete the attestation form. Commercial use needs a commercial licence. Versions released under CC-BY-4.0 stay under that licence; the terms do not restrict rights it grants.
Usage
pip install parakeet-mlx
parakeet-mlx audio.wav --model Hebrish/Sabich-MLX --fp32
import mlx.core as mx
from parakeet_mlx import from_pretrained
model = from_pretrained("Hebrish/Sabich-MLX", dtype=mx.float32)
print(model.transcribe("audio.wav").text)
python -m mlx_audio.stt.generate --model Hebrish/Sabich-MLX --audio audio.wav --output-path out --format txt
This is not an iPhone build: the Hebrish apps, and an iOS app, use the sherpa-onnx build (Hebrish/Sabich-sherpa-onnx-int8).
Conversion and check
Converted from training run r7-all with the remap behind the stock mlx-community Parakeet
repos (senstella's script),
fp32 weights (parakeet-mlx casts to bf16 at load unless asked for fp32; it cannot load quantized
weights, so there is no 8-bit build). Checked with parakeet-mlx 0.5.3 on MLX's Linux CPU backend, against the NeMo
PyTorch output on held-out code-switching test clips:
| Identical to NeMo | WER vs NeMo | WER vs reference | |
|---|---|---|---|
| NeMo (PyTorch) | 5.4% | ||
| MLX bf16 (default) | 9/10 | 0.5% | 5.9% |
| MLX fp32 | 9/10 | 0.5% | 5.9% |
WER here is raw (cased and punctuated) on a small sample; the main card has the full evaluation. The conversion is a direct key remap (no retraining or quantization).