<!-- 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-large-xls-r-300m-ar-2
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: 0.4764
- Wer: 0.3073
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: 0.001
- train_batch_size: 32
- eval_batch_size: 16
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 64
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 500
- num_epochs: 30
- mixed_precision_training: Native AMP
Training results
Training Loss | Epoch | Step | Validation Loss | Wer |
---|---|---|---|---|
1.0851 | 1.18 | 400 | 0.5614 | 0.4888 |
0.691 | 2.35 | 800 | 0.6557 | 0.5558 |
0.6128 | 3.53 | 1200 | 0.5852 | 0.5070 |
0.543 | 4.71 | 1600 | 0.5591 | 0.4838 |
0.5185 | 5.88 | 2000 | 0.6649 | 0.5514 |
0.4816 | 7.06 | 2400 | 0.5598 | 0.4689 |
0.4336 | 8.24 | 2800 | 0.5384 | 0.4515 |
0.405 | 9.41 | 3200 | 0.4987 | 0.4138 |
0.3811 | 10.59 | 3600 | 0.5427 | 0.4644 |
0.3539 | 11.76 | 4000 | 0.4881 | 0.4159 |
0.3299 | 12.94 | 4400 | 0.5160 | 0.4198 |
0.3096 | 14.12 | 4800 | 0.5019 | 0.4077 |
0.2881 | 15.29 | 5200 | 0.5146 | 0.4140 |
0.2894 | 16.47 | 5600 | 0.4861 | 0.4026 |
0.2461 | 17.65 | 6000 | 0.4765 | 0.3742 |
0.2371 | 18.82 | 6400 | 0.4679 | 0.3672 |
0.2182 | 20.0 | 6800 | 0.4699 | 0.3603 |
0.1942 | 21.18 | 7200 | 0.4769 | 0.3519 |
0.1823 | 22.35 | 7600 | 0.4719 | 0.3497 |
0.1682 | 23.53 | 8000 | 0.4876 | 0.3456 |
0.1526 | 24.71 | 8400 | 0.4591 | 0.3300 |
0.137 | 25.88 | 8800 | 0.4819 | 0.3314 |
0.1283 | 27.06 | 9200 | 0.4823 | 0.3213 |
0.1174 | 28.24 | 9600 | 0.4879 | 0.3174 |
0.1104 | 29.41 | 10000 | 0.4764 | 0.3073 |
Framework versions
- Transformers 4.17.0
- Pytorch 1.11.0
- Datasets 1.18.4
- Tokenizers 0.11.6