metadata
license: apache-2.0
base_model: facebook/wav2vec2-base
tags:
- generated_from_trainer
metrics:
- wer
model-index:
- name: w2v2-base-pretrained_lr1e-4_at1_da1
results: []
w2v2-base-pretrained_lr1e-4_at1_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: 6.1479
- Wer: 1.0644
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: 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: 1000
- num_epochs: 60
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 18.0956 | 6.1 | 250 | 4.6463 | 1.0 |
| 3.3157 | 12.2 | 500 | 3.7016 | 1.0 |
| 3.0923 | 18.29 | 750 | 3.9545 | 1.0 |
| 1.5621 | 24.39 | 1000 | 4.1576 | 1.0610 |
| 0.145 | 30.49 | 1250 | 4.6917 | 1.0503 |
| 0.0861 | 36.59 | 1500 | 5.4079 | 1.0635 |
| 0.0553 | 42.68 | 1750 | 5.8676 | 1.0601 |
| 0.0393 | 48.78 | 2000 | 6.3465 | 1.0618 |
| 0.0308 | 54.88 | 2250 | 6.1479 | 1.0644 |
Framework versions
- Transformers 4.35.0
- Pytorch 2.0.0
- Datasets 2.14.6
- Tokenizers 0.14.1