kid-whisper-tiny-en-myst-ours

Fine-tuned openai/whisper-tiny.en on the MyST children's speech corpus, for research into compression (quantization + pruning) of children's ASR models at small scale.

Model Details

• Base model: openai/whisper-tiny.en (39M parameters)
• Fine-tuning data: MyST corpus, filtered train/dev splits (57,687 train utterances / 9,017 dev utterances)
• Filtering pipeline: custom preprocessing (missing-transcript removal, reference-ASR quality filtering, non-speech-tag handling, short-utterance removal, duration capping for train/dev). See [link to your preprocess_myst.py / thesis Chapter 2 if published] for full filter specification.
• Training framework: HuggingFace Seq2SeqTrainer
• Precision: fp16 mixed-precision training

Training Hyperparameters

Parameter	Value
Learning rate	1e-5
Per-device batch size	16
Gradient accumulation	4 (effective batch 64)
Warmup steps	500
Early stopping	patience=5, metric=WER on dev
Best checkpoint	step 5000

Evaluation

Training-time dev WER (early-stopping metric only): 13.14% Measured on individual (non-concatenated) data/filtered/dev utterances, used solely for checkpoint selection during training. This is not the model's reportable evaluation number.

Reportable baseline WER: 14.53% Measured via HuggingFace ASR pipeline with chunk_length_s=30, beam search (beam=5), batch=4, against the full 3,972-chunk held-out MyST concatenated test set (data/concatenated/test) — the identical protocol used to evaluate every other checkpoint in this research.

Model	Parameters	FP16 Baseline WER
kid-whisper-tiny-en-myst-ours (this model)	39M	14.53%
kid-whisper-small-en-myst (Attia et al.)	244M	9.16%
kid-whisper-medium-en-myst (Attia et al.)	769M	8.94%
kid-whisper-small-myst, multilingual (Attia et al.)	244M	9.91%

All four numbers were measured under the identical evaluation protocol above, making this the first case in this research where the tiny-scale result is directly comparable to the Small/Medium-scale checkpoints rather than relying on separately-reported, differently-measured numbers.

Relationship to Other KID-Whisper Checkpoints

This checkpoint is independently trained and is not derived from or affiliated with:

• aadel4/kid-whisper-small-en-myst / aadel4/kid-whisper-medium-en-myst (Attia et al., fine-tuned Small/Medium Whisper on MyST)
• SatwikDutta/kid-whisper-tiny-en-myst (Dutta et al., fine-tuned tiny.en on a different MyST filtering/split scheme)

This checkpoint was trained specifically to extend a compression study (quantization + pruning) to tiny-scale Whisper using the same data filtering and evaluation pipeline already used for Small/Medium-scale checkpoints in that study, avoiding a training-data confound that would have resulted from mixing in an externally-trained tiny.en checkpoint with a different, undocumented data version.

Intended Use

Research use for studying model compression (post-training quantization, magnitude/importance-based pruning) on children's speech ASR at small model scale. Not evaluated or intended for production deployment.

License

MIT, matching the base openai/whisper-tiny.en license.