xlsr-1b-mizonal3-E2-lus-v2026.06

This model is a fine-tuned version of facebook/wav2vec2-xls-r-1b 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: 24.8748
• Cer: 5.1604
• Real Time Factor: 0.0043

Quick Inference

import torch
import librosa
from transformers import Wav2Vec2Processor, Wav2Vec2ForCTC

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

processor = Wav2Vec2Processor.from_pretrained("andrewbawitlung/xlsr-1b-mizonal3-E2-lus-v2026.06")
model = Wav2Vec2ForCTC.from_pretrained("andrewbawitlung/xlsr-1b-mizonal3-E2-lus-v2026.06").to(device)

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

with torch.no_grad():
    logits = model(input_values).logits

predicted_ids = torch.argmax(logits, dim=-1)
transcription = processor.batch_decode(predicted_ids)[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/xlsr-1b-mizonal3-E1-lus-v2026.06
E2 (Noise)	andrewbawitlung/xlsr-1b-mizonal3-E2-lus-v2026.06
E3 (Speed)	andrewbawitlung/xlsr-1b-mizonal3-E3-lus-v2026.06
E4 (SpecAug)	andrewbawitlung/xlsr-1b-mizonal3-E4-lus-v2026.06
E5 (Combined)	andrewbawitlung/xlsr-1b-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	0.8755	0.4581	48.95	12.17	1.49e-04	1.40
500	0.91	0.5465	0.3804	40.83	10.36	2.99e-04	1.24
750	1.37	0.4093	0.2644	30.86	7.10	2.81e-04	1.03
1000	1.82	0.3209	0.2372	26.79	6.12	2.62e-04	0.81
1250	2.28	0.2007	0.2154	24.54	5.36	2.42e-04	0.46
1500	2.73	0.1880	0.2011	23.05	5.06	2.23e-04	0.47
1750	3.19	0.1394	0.1852	23.79	4.93	2.04e-04	0.49
2000	3.64	0.1953	0.1968	22.90	4.93	1.84e-04	0.92
2250	4.10	0.2069	0.2769	24.44	5.02	1.65e-04	0.58
2500	4.55	0.3208	0.3434	24.33	5.06	1.46e-04	0.76
2750	5.01	0.3164	0.3326	23.62	5.10	1.27e-04	0.26
3000	5.46	0.2487	0.2662	22.49	4.84	1.07e-04	0.46
3250	5.92	0.2050	0.2446	22.45	4.77	8.81e-05	0.41
3500	6.38	0.2008	0.2391	22.36	4.75	6.88e-05	0.29
3750	6.83	0.1897	0.2370	23.05	4.89	4.96e-05	0.19
4000	7.29	0.1851	0.2370	22.47	4.78	3.03e-05	0.12
4250	7.74	0.1884	0.2363	22.61	4.81	1.10e-05	0.00