Modelpublic

nsa01n/cohere-cs-decoder-middle

sourceHugging Faceapache-2.0updated 12d agoView on Hugging Face

0likes18downloads

Model Card

nsa01n/cohere-cs-decoder-middle

Decoder-only fine-tune of `CohereLabs/cohere-transcribe-arabic-07-2026` for Arabic/English code-switching.

Arabic podcast speech routinely borrows English words, and a plain Arabic ASR model transcribes them phonetically in Arabic script ("الويكند"). This model is trained to emit them in Latin script instead ("ال weekend"), which is what downstream consumers of the transcript actually want.

What was trained

Variant `middle` — the middle third of the decoder layers were unfrozen. The Conformer audio encoder is frozen in every variant of this series, including full; "full" means the full decoder, not the full model.


Decoder stack	`model.decoder` (8 layers)
Layers trained	[3, 4]
Encoder trainable params	0
Total params	2066M
Trainable params	70M (3.366%)

Training data

`Ahmed1/cohere-asr-cs` — VAD-segmented Arabic podcast clips (16 kHz mono), transcribed with the base model and then rewritten so English loanwords appear in Latin script. Splits are grouped by source episode, so no episode appears in more than one split.

Trained on 1,900 clips, validated on 454.

Hyperparameters

parameter	value
gradientaccumulationsteps	2
learning_rate	8e-05
lrschedulertype	cosine
maxgradnorm	1.0
numtrainepochs	6
optim	adamwbnb8bit
perdevicetrainbatchsize	16
savetotallimit	1
warmup_ratio	0.03
weight_decay	0.01
effective batch size	32
precision	bf16
seed	42

Final metrics: {"train_runtime": 408.5423, "train_samples_per_second": 27.904, "train_steps_per_second": 0.881, "total_flos": 5.372639220298678e+19, "train_loss": 0.14237111575073666, "epoch": 6.0}

Evaluation

Code-switching (the task)

metric	this model	base
clips it switched script on	43/464 (9.3%)	0/464 (0.0%)
clip-level recall	0.304	0.000
Latin word F1	0.335	0.000
Latin word precision	0.723	0.000
Latin word recall	0.218	0.000
hallucinated Latin tokens	18	0

Measured on 464 held-out clips, 125 of which contain English loanwords (216 Latin word tokens).

Transcription (context, not the target)

metric	value
WER	0.0536
CER	0.0293

WER rises relative to the base model, and that is expected. The references are the base model's own transcriptions with loanwords rewritten into Latin script, so the base reproduces them almost exactly (~1% WER) while never code-switching. Every script change this model makes correctly still counts as edits against a base-shaped reference.

Usage

python

from transformers import AutoModelForSpeechSeq2Seq, AutoProcessor

processor = AutoProcessor.from_pretrained("nsa01n/cohere-cs-decoder-middle")
model = AutoModelForSpeechSeq2Seq.from_pretrained("nsa01n/cohere-cs-decoder-middle")

Feed it 16 kHz mono audio, ideally VAD-segmented to under 30 s per clip — that is how it was trained and the base model caps at 35 s.

Limitations

—Two podcast shows only (sawalef-business, soqrat); other domains and dialects are out of distribution.
—Training targets were machine-generated (base-model transcription + an LLM rewrite pass), not human-verified, so its ceiling is the base model's accuracy on this audio.
—The encoder was never trained, so acoustic robustness is unchanged from the base model.