Downloads · 30 days
1
9% of all-time downloads
k3rb3l/wav2vec2-xls-r-300m-en-atc-uwb-atcc-coreml
wav2vec2-xls-r-300m-en-atc-uwb-atcc-coreml is a automatic speech recognition model from k3rb3l. Use it when you need speech turned into text. The card lists the license as apache-2.0.
Jzuluaga/wav2vec2-xls-r-300m-en-atc-uwb-atcc converted to Core ML for on-device inference on iPhone/iPad (iOS 16+, FP16 mlprogram).
Downloads · 30 days
1
9% of all-time downloads
All-time downloads
11
Public
Repo size
631 MB
Likes
0
Public
Click a slice to open those files.
.bin631 MB · 100%
From the Hugging Face model README
Jzuluaga/wav2vec2-xls-r-300m-en-atc-uwb-atcc converted to Core ML for
on-device inference on iPhone/iPad (iOS 16+, FP16 mlprogram).
audio: raw mono 16 kHz Float32, shape (1, samples), 1–60 s
(flexible). Feature-extractor normalization is baked into the graph — feed
raw samples.logits: CTC logits (1, frames, vocab). Greedy-decode: argmax
over vocab, collapse repeats, drop the [PAD] blank (id 28), map ids via
vocab.json, | (id 0) → space.Chunk audio longer than 60 s (e.g. 10 s windows with 2 s stride).
Converted with coremltools from the original PyTorch weights (traced, FP16).
Greedy CTC transcripts of the validation clip match the PyTorch baseline:
match: True, max |logit Δ| = 0.1379.
Note: the original repo decodes with a KenLM language model (pyctcdecode). That LM is not part of this package — on-device decoding is greedy CTC, which scores slightly lower than the LM-rescored path.
Conversion script: gmet/tools/convert_wav2vec2_coreml.py