README.md

metadata

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-E5-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: 19.466
          - name: Cer
            type: cer
            value: 4.1407
          - name: Real Time Factor
            type: rtf
            value: 0.0564

qwen3-asr-1.7b-mizonal3-E5-lus-v2026.06

This model is a fine-tuned version of Qwen/Qwen3-ASR-1.7B on the MiZonal v3.0 dataset.

It achieves the following results on the evaluation set:

• Wer: 19.4660
• Cer: 4.1407
• Real Time Factor: 0.0564

Quick Inference

import torch
from qwen_asr import Qwen3ASRModel

# Load the model
model = Qwen3ASRModel.from_pretrained(
    "andrewbawitlung/qwen3-asr-1.7b-mizonal3-E5-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
E2 (Noise)	andrewbawitlung/qwen3-asr-1.7b-mizonal3-E2-lus-v2026.06
E3 (Speed)	andrewbawitlung/qwen3-asr-1.7b-mizonal3-E3-lus-v2026.06
E4 (SpecAug)	andrewbawitlung/qwen3-asr-1.7b-mizonal3-E4-lus-v2026.06
E5 (Combined)	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	0.7286	0.4891	0.2692	23.1099	5.0590	1.86e-05	7.5938
400	1.4554	0.1987	0.2398	18.8285	4.2161	1.67e-05	5.7188
600	2.1821	0.0896	0.2608	18.6338	4.3610	1.49e-05	4.9375
800	2.9107	0.0570	0.2731	19.0620	4.2974	1.30e-05	3.2656
1000	3.6375	0.0286	0.3041	18.9452	4.2373	1.11e-05	3.0781
1200	4.3643	0.0138	0.3248	18.4490	4.1402	9.29e-06	5.3750
1400	5.0911	0.0067	0.3406	18.7701	4.2603	7.43e-06	0.6484
1600	5.8197	0.0069	0.3434	18.5268	4.2020	5.58e-06	1.6250
1800	6.5464	0.0058	0.3537	18.5268	4.2020	3.72e-06	0.6523
2000	7.2732	0.0048	0.3551	18.5755	4.2462	1.86e-06	0.4805
2200	8.0000	0.0048	0.3554	18.6241	4.1896	9.28e-09	1.2188