metadata
license: apache-2.0
base_model: facebook/wav2vec2-base
tags:
- generated_from_trainer
metrics:
- wer
model-index:
- name: w2v2-base-pretrained_lr5e-5_at0.2_da1
results: []
w2v2-base-pretrained_lr5e-5_at0.2_da1
This model is a fine-tuned version of facebook/wav2vec2-base on the None dataset. It achieves the following results on the evaluation set:
- Loss: 1.0942
- Wer: 0.1674
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: 32
- 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: 500
- training_steps: 4000
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 17.4656 | 3.91 | 250 | 3.8210 | 1.0 |
| 3.2203 | 7.81 | 500 | 3.1655 | 1.0 |
| 2.5403 | 11.72 | 750 | 1.2547 | 0.9979 |
| 0.5746 | 15.62 | 1000 | 0.5996 | 0.5088 |
| 0.2573 | 19.53 | 1250 | 0.7483 | 0.2046 |
| 0.152 | 23.44 | 1500 | 0.9229 | 0.1862 |
| 0.1082 | 27.34 | 1750 | 0.9192 | 0.1833 |
| 0.0748 | 31.25 | 2000 | 1.0565 | 0.1747 |
| 0.0603 | 35.16 | 2250 | 0.9710 | 0.1815 |
| 0.0485 | 39.06 | 2500 | 1.0599 | 0.1704 |
| 0.0399 | 42.97 | 2750 | 1.0942 | 0.1730 |
| 0.034 | 46.88 | 3000 | 1.0842 | 0.1670 |
| 0.0309 | 50.78 | 3250 | 1.0670 | 0.1632 |
| 0.0269 | 54.69 | 3500 | 1.1369 | 0.1649 |
| 0.0244 | 58.59 | 3750 | 1.0229 | 0.1666 |
| 0.0228 | 62.5 | 4000 | 1.0942 | 0.1674 |
Framework versions
- Transformers 4.35.0
- Pytorch 2.0.0
- Datasets 2.14.6
- Tokenizers 0.14.1