This model is a fine-tuned version of openai/whisper-small on the MiZonal v3.0 dataset. Note: ~1 hour of conversational speech was added to this dataset version.

It achieves the following results on the evaluation set:

Wer: 20.4869
Cer: 6.0524
Real Time Factor: 0.0238

Quick Inference

python

import torch
import librosa
from transformers import WhisperProcessor, WhisperForConditionalGeneration

device = "cuda" if torch.cuda.is_available() else "cpu"

processor = WhisperProcessor.from_pretrained("andrewbawitlung/whisper-small-mizonal3-E4-lus-v2026.06")
model = WhisperForConditionalGeneration.from_pretrained("andrewbawitlung/whisper-small-mizonal3-E4-lus-v2026.06").to(device)

audio, sr = librosa.load("your_audio.wav", sr=16000)
input_features = processor(audio, sampling_rate=16000, return_tensors="pt").input_features.to(device)

with torch.no_grad():
    predicted_ids = model.generate(input_features, max_new_tokens=256)

transcription = processor.batch_decode(predicted_ids, skip_special_tokens=True)[0]
print(transcription)

Model description

Experiment Configurations

This repository is part of a series of experiments. The different configurations are:

E1 (Baseline): Standard training configuration.
E2 (Noise): Training with background noise augmentation.
E3 (Speed): Training with speed perturbation augmentation.
E4 (SpecAug): Training with SpecAugment (time and frequency masking).
E5 (Combined): Training with a combination of all augmentations.

All Models in this Family

Experiment	Hugging Face Repository
E1 (Baseline)	andrewbawitlung/whisper-small-mizonal3-E1-lus-v2026.06
E2 (Noise)	andrewbawitlung/whisper-small-mizonal3-E2-lus-v2026.06
E3 (Speed)	andrewbawitlung/whisper-small-mizonal3-E3-lus-v2026.06
E4 (SpecAug)	andrewbawitlung/whisper-small-mizonal3-E4-lus-v2026.06
E5 (Combined)	andrewbawitlung/whisper-small-mizonal3-E5-lus-v2026.06

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 0.0003
train_batch_size: 8
eval_batch_size: 8
seed: 42
optimizer: OptimizerNames.ADAMW_TORCH_FUSED
lr_scheduler_type: SchedulerType.LINEAR
num_epochs: 8

Training results

step	epoch	train_loss	eval_loss	eval_wer	eval_cer	learning_rate	grad_norm
250	0.91	0.6510	0.5886	38.47	15.37	1.49e-04	8.01
500	1.82	0.7070	0.6960	41.22	16.54	2.99e-04	7.35
750	2.73	0.5222	0.5778	47.78	26.32	2.56e-04	4.32
1000	3.64	0.3354	0.5246	29.65	12.14	2.12e-04	3.21
1250	4.55	0.2034	0.5009	28.23	11.60	1.68e-04	3.17
1500	5.46	0.1231	0.4744	23.27	7.47	1.24e-04	2.51
1750	6.36	0.0741	0.4535	20.68	6.55	7.96e-05	2.13
2000	7.27	0.0303	0.4445	18.82	5.57	3.55e-05	0.87