metadata
license: apache-2.0
tags:
- generated_from_trainer
metrics:
- wer
model-index:
- name: wav2vec2-large-asr-th-2
results: []
wav2vec2-large-asr-th-2
This model is a fine-tuned version of facebook/wav2vec2-large-xlsr-53 on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.5485
- Wer: 0.6163
- Cer: 0.1880
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: 0.0001
- train_batch_size: 20
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 40
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 500
- training_steps: 3000
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer | Cer |
|---|---|---|---|---|---|
| 3.7562 | 0.39 | 500 | 3.6183 | 1.0 | 0.9999 |
| 2.3585 | 0.79 | 1000 | 1.6482 | 0.9893 | 0.4815 |
| 1.1946 | 1.18 | 1500 | 0.7852 | 0.7434 | 0.2404 |
| 1.0082 | 1.58 | 2000 | 0.6294 | 0.6566 | 0.2069 |
| 0.9679 | 1.97 | 2500 | 0.5688 | 0.6272 | 0.1922 |
| 0.9845 | 2.37 | 3000 | 0.5485 | 0.6163 | 0.1880 |
Framework versions
- Transformers 4.27.3
- Pytorch 1.13.1+cu116
- Datasets 2.10.1
- Tokenizers 0.13.2