---
language:
- lus
license: apache-2.0
pipeline_tag: automatic-speech-recognition
base_model: Qwen/Qwen3-ASR-1.7B
tags:
- generated_from_trainer
datasets:
- andrewbawitlung/MiZonal-v3.0
metrics:
- wer
- cer
model-index:
- name: qwen3-asr-1.7b-mizonal3-E3-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: 21.3802
    - name: Cer
      type: cer
      value: 4.4457
    - name: Real Time Factor
      type: rtf
      value: 0.0589
---
![Mizo Automatic Speech Recognition (ASR) Models v3.0](banner.jpg)

# qwen3-asr-1.7b-mizonal3-E3-lus-v2026.06

This model is a fine-tuned version of [Qwen/Qwen3-ASR-1.7B](https://huggingface.co/Qwen/Qwen3-ASR-1.7B) on the **MiZonal v3.0** dataset.

It achieves the following results on the evaluation set:
- Wer: 21.3802
- Cer: 4.4457
- Real Time Factor: 0.0589

## Quick Inference


```python
import torch
from qwen_asr import Qwen3ASRModel

# Load the model
model = Qwen3ASRModel.from_pretrained(
    "andrewbawitlung/qwen3-asr-1.7b-mizonal3-E3-lus-v2026.06", 
    dtype=torch.bfloat16, 
    device_map="cuda:0"  # Adjust device as needed
)

# Transcribe audio
results = model.transcribe("your_audio.wav")
print(results)
```


## 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/qwen3-asr-1.7b-mizonal3-E1-lus-v2026.06](https://huggingface.co/andrewbawitlung/qwen3-asr-1.7b-mizonal3-E1-lus-v2026.06) |
| **E2 (Noise)** | [andrewbawitlung/qwen3-asr-1.7b-mizonal3-E2-lus-v2026.06](https://huggingface.co/andrewbawitlung/qwen3-asr-1.7b-mizonal3-E2-lus-v2026.06) |
| **E3 (Speed)** | [andrewbawitlung/qwen3-asr-1.7b-mizonal3-E3-lus-v2026.06](https://huggingface.co/andrewbawitlung/qwen3-asr-1.7b-mizonal3-E3-lus-v2026.06) |
| **E4 (SpecAug)** | [andrewbawitlung/qwen3-asr-1.7b-mizonal3-E4-lus-v2026.06](https://huggingface.co/andrewbawitlung/qwen3-asr-1.7b-mizonal3-E4-lus-v2026.06) |
| **E5 (Combined)** | [andrewbawitlung/qwen3-asr-1.7b-mizonal3-E5-lus-v2026.06](https://huggingface.co/andrewbawitlung/qwen3-asr-1.7b-mizonal3-E5-lus-v2026.06) |

### Training hyperparameters

The following hyperparameters were used during training:
- learning_rate: 2e-05
- train_batch_size: 32
- 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 |
| --- | --- | --- | --- | --- | --- | --- | --- |
| 200 | 0.9709 | 0.4075 | 0.2658 | 22.4385 | 4.9848 | 1.79e-05 | 6.1250 |
| 400 | 1.9417 | 0.1853 | 0.2452 | 19.5096 | 4.4547 | 1.55e-05 | 5.7188 |
| 600 | 2.9126 | 0.0737 | 0.2774 | 19.1982 | 4.4847 | 1.30e-05 | 3.8750 |
| 800 | 3.8835 | 0.0359 | 0.3090 | 19.7626 | 4.5784 | 1.05e-05 | 3.2031 |
| 1000 | 4.8544 | 0.0132 | 0.3367 | 19.1398 | 4.4158 | 8.04e-06 | 2.7344 |
| 1200 | 5.8252 | 0.0077 | 0.3561 | 19.3344 | 4.5642 | 5.56e-06 | 0.9766 |
| 1400 | 6.7961 | 0.0053 | 0.3702 | 19.2469 | 4.5271 | 3.08e-06 | 0.6406 |
| 1600 | 7.7670 | 0.0052 | 0.3726 | 19.2955 | 4.5713 | 6.07e-07 | 0.4590 |