metadata
license: apache-2.0
tags:
- generated_from_trainer
datasets:
- common_voice
model-index:
- name: wav2vec2-large-xlsr-300-arabic
results: []
wav2vec2-large-xlsr-300-arabic
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: 0.3384
- Wer: 0.3105
- Cer: 0.0879
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: 64
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 128
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 1000
- num_epochs: 10
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer | Cer |
|---|---|---|---|---|---|
| 0.7383 | 1.8 | 500 | 0.4292 | 0.4065 | 0.1189 |
| 0.664 | 3.6 | 1000 | 0.4245 | 0.3978 | 0.1175 |
| 0.6064 | 5.4 | 1500 | 0.3854 | 0.3625 | 0.1048 |
| 0.5221 | 7.19 | 2000 | 0.3819 | 0.3400 | 0.0976 |
| 0.4591 | 8.99 | 2500 | 0.3384 | 0.3105 | 0.0879 |
Framework versions
- Transformers 4.16.0.dev0
- Pytorch 1.10.1+cu102
- Datasets 1.17.1.dev0
- Tokenizers 0.11.0