Xlsr 300m Mizonal3 E4 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.
Modellquelle
- Art
- Other
- Dateigröße
- 1,18 GB
- Basismodell
- facebook/wav2vec2-xls-r-300m
- Version
- vRevision 719833d
- Aktualisiert
- 19. Okt. 2026
- Downloads
- 9
- Quelle
- Hugging Face
Quellenbeschreibung
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.
xlsr-300m-mizonal3-E4-lus-v2026.06
This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m 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: 19.7605
- Cer: 3.7878
- Real Time Factor: 0.0024
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-300m-mizonal3-E4-lus-v2026.06")
model = Wav2Vec2ForCTC.from_pretrained("andrewbawitlung/xlsr-300m-mizonal3-E4-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-300m-mizonal3-E1-lus-v2026.06 |
| E2 (Noise) | andrewbawitlung/xlsr-300m-mizonal3-E2-lus-v2026.06 |
| E3 (Speed) | andrewbawitlung/xlsr-300m-mizonal3-E3-lus-v2026.06 |
| E4 (SpecAug) | andrewbawitlung/xlsr-300m-mizonal3-E4-lus-v2026.06 |
| E5 (Combined) | andrewbawitlung/xlsr-300m-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 | 10.9711 | 2.4033 | 99.72 | 91.88 | 1.49e-04 | 2.63 |
| 500 | 1.82 | 3.1841 | 0.4280 | 45.31 | 11.44 | 2.99e-04 | 2.52 |
| 750 | 2.73 | 2.3094 | 0.2775 | 32.74 | 7.43 | 2.56e-04 | 3.22 |
| 1000 | 3.64 | 1.8766 | 0.2251 | 27.46 | 5.96 | 2.12e-04 | 3.20 |
| 1250 | 4.55 | 1.4938 | 0.1916 | 23.92 | 5.01 | 1.68e-04 | 2.80 |
| 1500 | 5.46 | 1.2829 | 0.1783 | 22.01 | 4.54 | 1.24e-04 | 2.71 |
| 1750 | 6.36 | 1.2294 | 0.1607 | 20.72 | 4.14 | 7.96e-05 | 3.18 |
| 2000 | 7.27 | 1.1032 | 0.1517 | 18.84 | 3.80 | 3.55e-05 | 2.87 |
| 2200 | 8.00 | 1.0568 | 0.1485 | 18.31 | 3.66 | 1.76e-07 | 1.83 |
Quellen
1 Quelle
Quellen
1 QuelleVerifiziert 19. Okt.
Modellartefakte
1 Artefakt
Modellartefakte
1 ArtefaktQuellenauszüge
3 Auszüge
Quellenauszüge
3 Auszüge--- language: - lus license: apache-2.0 pipeline_tag: automatic-speech-recognition base_model: facebook/wav2vec2-xls-r-300m tags: - generated_from_trainer datasets: - andrewbawitlung/MiZonal-v3.0 metrics: - wer - cer model-index: - name: xlsr-300m-mizonal3-E4-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: 19.7605 - name: Cer type: cer value: 3.7878 - name: Real Time Factor type: rtf value: 0.0024 ---  > **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. --> # xlsr-300m-mizonal3-E4-lus-v2026.06 This model is a fine-tuned version of [facebook/wav2vec2-xls-r-300m](https://huggingface.co/facebook/wav2vec2-xls-r-300m) 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: 19.7605 - Cer: 3.7878 - Real Time Factor: 0.0024 ## Quick Inference ```python import torch import librosa from transformers import Wav2Vec2Processor, Wav2Vec2ForCTC device = "cuda" if torch.cuda.is_available() else "cpu" processor = Wav2Vec2Processor.from_pretrained("andrewbawitlung/xlsr-300m-mizonal3-E4-lus-v2026.06") model = Wav2Vec2ForCTC.from_pretrained("andrewbawitlung/xlsr-300m-mizonal3-E4-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...
Source context: 9 downloads · 0 likes · Pipeline automatic-speech-recognition · Repo andrewbawitlung/xlsr-300m-mizonal3-E4-lus-v2026.06
Source context: 39 downloads · 0 likes · Pipeline automatic-speech-recognition · Repo andrewbawitlung/xlsr-300m-mizonal3-E4-lus-v2026.06
