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: 20.9090
Cer: 6.6145
Real Time Factor: 0.0256

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-E1-lus-v2026.06")
model = WhisperForConditionalGeneration.from_pretrained("andrewbawitlung/whisper-small-mizonal3-E1-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	0.5595	0.5548	61.74	28.44	1.49e-04	5.02
500	1.82	0.6523	0.7330	39.35	15.07	2.99e-04	8.09
750	2.73	0.5007	0.6280	37.52	17.94	2.56e-04	6.30
1000	3.64	0.3050	0.5721	29.66	11.26	2.12e-04	3.05
1250	4.55	0.1731	0.5281	27.26	10.38	1.68e-04	3.45
1500	5.46	0.0961	0.4887	23.05	7.51	1.24e-04	2.42
1750	6.36	0.0464	0.4911	22.28	7.40	7.96e-05	2.54
2000	7.27	0.0174	0.4611	19.28	5.81	3.55e-05	0.62