<!-- 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. -->
wav2vec2-xls-r-300m-arabic-suit
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: 0.2986
- Wer: 22.4877
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: 5e-05
- train_batch_size: 8
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 4
- total_train_batch_size: 32
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 500
- training_steps: 12000
- mixed_precision_training: Native AMP
Training results
Training Loss | Epoch | Step | Validation Loss | Wer |
---|---|---|---|---|
0.6952 | 0.83 | 1000 | 0.5802 | 56.5975 |
0.4528 | 1.66 | 2000 | 0.4097 | 39.5698 |
0.3064 | 2.5 | 3000 | 0.3433 | 32.3567 |
0.232 | 3.33 | 4000 | 0.3192 | 28.1373 |
0.1677 | 4.16 | 5000 | 0.2956 | 25.8399 |
0.1474 | 4.99 | 6000 | 0.2748 | 24.2858 |
0.2104 | 5.82 | 7000 | 0.3265 | 27.7863 |
0.1689 | 6.66 | 8000 | 0.3081 | 26.2716 |
0.1312 | 7.49 | 9000 | 0.3112 | 25.0516 |
0.1041 | 8.32 | 10000 | 0.3071 | 23.7715 |
0.0913 | 9.15 | 11000 | 0.3044 | 22.8781 |
0.0963 | 9.98 | 12000 | 0.2986 | 22.4877 |
Framework versions
- Transformers 4.20.1
- Pytorch 1.10.0+cu113
- Datasets 2.1.0
- Tokenizers 0.12.1