metadata
license: apache-2.0
tags:
- generated_from_trainer
datasets:
- common_voice
metrics:
- wer
model-index:
- name: ArabicSpeechToTextModel
results:
- task:
name: Automatic Speech Recognition
type: automatic-speech-recognition
dataset:
name: common_voice
type: common_voice
config: ar
split: test
args: ar
metrics:
- name: Wer
type: wer
value: 0.252583913221449
ArabicSpeechToTextModel
This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the common_voice dataset. It achieves the following results on the evaluation set:
- Loss: 0.2868
- Wer: 0.2526
Model description
More information needed
Intended uses & limitations
More information needed
Training and evaluation data
More information needed
Training procedure
Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 0.0003
- train_batch_size: 2
- eval_batch_size: 8
- seed: 8
- gradient_accumulation_steps: 8
- total_train_batch_size: 16
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 500
- num_epochs: 4
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 5.6373 | 0.52 | 700 | 0.7918 | 0.7011 |
| 0.71 | 1.03 | 1400 | 0.5274 | 0.4574 |
| 0.4809 | 1.55 | 2100 | 0.4143 | 0.3755 |
| 0.387 | 2.06 | 2800 | 0.3700 | 0.3345 |
| 0.3086 | 2.58 | 3500 | 0.3362 | 0.3018 |
| 0.2716 | 3.09 | 4200 | 0.3081 | 0.2717 |
| 0.2073 | 3.61 | 4900 | 0.2868 | 0.2526 |
Framework versions
- Transformers 4.30.2
- Pytorch 2.0.1+cu118
- Datasets 2.13.1
- Tokenizers 0.13.3