Whisper Small Mizonal3 E2 Lus v2026.06
> Disclaimer / Notice: Details for these are in Peer Review and publications of the paper will be made available soon for more details.
Model source
- Kind
- Text model
- File size
- 922 MB
- Base model
- openai/whisper-small
- Version
- vmodel
- Updated
- Oct 19, 2026
- Downloads
- 9
- Source
- Hugging Face
Source description
Mizo Automatic Speech Recognition (ASR) Models v3.0
Disclaimer / Notice: Details for these are in Peer Review and publications of the paper will be made available soon for more details.
whisper-small-mizonal3-E2-lus-v2026.06
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: 22.3422
- Cer: 8.0155
- Real Time Factor: 0.0255
Quick Inference
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-E2-lus-v2026.06")
model = WhisperForConditionalGeneration.from_pretrained("andrewbawitlung/whisper-small-mizonal3-E2-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 | 0.46 | 2.4902 | 0.5650 | 37.47 | 12.96 | 1.49e-04 | 29.28 |
| 500 | 0.91 | 2.6669 | 0.7054 | 39.37 | 16.05 | 2.99e-04 | 27.87 |
| 750 | 1.37 | 2.0718 | 0.6896 | 38.85 | 25.33 | 2.81e-04 | 24.14 |
| 1000 | 1.82 | 1.5676 | 0.6065 | 31.95 | 11.21 | 2.62e-04 | 15.83 |
| 1250 | 2.28 | 1.0195 | 0.6181 | 30.51 | 12.53 | 2.42e-04 | 20.85 |
| 1500 | 2.73 | 0.8523 | 0.5772 | 30.07 | 12.88 | 2.23e-04 | 11.63 |
| 1750 | 3.19 | 0.5502 | 0.5918 | 28.73 | 11.14 | 2.04e-04 | 7.99 |
| 2000 | 3.64 | 0.5025 | 0.5752 | 30.36 | 11.42 | 1.84e-04 | 12.55 |
| 2250 | 4.10 | 0.3019 | 0.5884 | 33.95 | 15.83 | 1.65e-04 | 5.13 |
| 2500 | 4.55 | 0.2702 | 0.5884 | 40.83 | 23.26 | 1.46e-04 | 4.48 |
| 2750 | 5.01 | 0.2087 | 0.5710 | 23.39 | 8.77 | 1.27e-04 | 5.55 |
| 3000 | 5.46 | 0.1572 | 0.5716 | 25.11 | 10.18 | 1.07e-04 | 6.26 |
| 3250 | 5.92 | 0.1184 | 0.5987 | 32.94 | 15.67 | 8.81e-05 | 2.89 |
| 3500 | 6.38 | 0.0607 | 0.5804 | 21.96 | 7.76 | 6.88e-05 | 0.80 |
| 3750 | 6.83 | 0.0562 | 0.5838 | 21.34 | 7.26 | 4.96e-05 | 9.40 |
| 4000 | 7.29 | 0.0172 | 0.5829 | 21.33 | 7.35 | 3.03e-05 | 0.17 |
| 4250 | 7.74 | 0.0087 | 0.5896 | 22.58 | 8.39 | 1.10e-05 | 0.07 |
| 4392 | 8.00 | 0.5884 | 21.92 | 8.03 |
Sources
1 source
Sources
1 sourceVerified Oct 19
Model artifacts
1 artifact
Model artifacts
1 artifactSource excerpts
3 excerpts
Source excerpts
3 excerpts--- language: - lus license: apache-2.0 pipeline_tag: automatic-speech-recognition base_model: openai/whisper-small tags: - generated_from_trainer datasets: - andrewbawitlung/MiZonal-v3.0 metrics: - wer - cer model-index: - name: whisper-small-mizonal3-E2-lus-v2026.06 results: - task: name: Automatic Speech Recognition type: automatic-speech-recognition dataset: name: MiZonal v3.0 type: andrewbawitlung/MiZonal-v3.0 config: default split: test metrics: - name: Wer type: wer value: 22.3422 - name: Cer type: cer value: 8.0155 - name: Real Time Factor type: rtf value: 0.0255 ---  > **Disclaimer / Notice:** Details for these are in Peer Review and publications of the paper will be made available soon for more details. <!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. --> # whisper-small-mizonal3-E2-lus-v2026.06 This model is a fine-tuned version of [openai/whisper-small](https://huggingface.co/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: 22.3422 - Cer: 8.0155 - Real Time Factor: 0.0255 ## 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-E2-lus-v2026.06") model = WhisperForConditionalGeneration.from_pretrained("andrewbawitlung/whisper-small-mizonal3-E2-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...
Source context: 39 downloads · 0 likes · Pipeline automatic-speech-recognition · Repo andrewbawitlung/whisper-small-mizonal3-E2-lus-v2026.06
Source context: 9 downloads · 0 likes · Pipeline automatic-speech-recognition · Repo andrewbawitlung/whisper-small-mizonal3-E2-lus-v2026.06
