<!-- 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-teacher-en-asr-timit
This model is a fine-tuned version of facebook/wav2vec2-large on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.4160
- Wer: 0.2984
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: 64
- eval_batch_size: 8
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 1000
- num_epochs: 30
- mixed_precision_training: Native AMP
Training results
Training Loss | Epoch | Step | Validation Loss | Wer |
---|---|---|---|---|
4.4952 | 3.17 | 200 | 3.0555 | 1.0 |
2.5341 | 6.35 | 400 | 0.8144 | 0.7441 |
0.694 | 9.52 | 600 | 0.4154 | 0.4572 |
0.3593 | 12.7 | 800 | 0.4260 | 0.3890 |
0.2567 | 15.87 | 1000 | 0.4166 | 0.3614 |
0.1988 | 19.05 | 1200 | 0.3912 | 0.3346 |
0.1338 | 22.22 | 1400 | 0.4000 | 0.3178 |
0.1044 | 25.4 | 1600 | 0.4425 | 0.3071 |
0.0786 | 28.57 | 1800 | 0.4160 | 0.2984 |
Framework versions
- Transformers 4.25.1
- Pytorch 1.12.1+cu113
- Datasets 1.18.3
- Tokenizers 0.13.2