metadata
license: apache-2.0
base_model: facebook/wav2vec2-base
tags:
- generated_from_trainer
metrics:
- wer
model-index:
- name: w2v2-base-pretrained_lr1e-4_at0.8_da0.3
results: []
w2v2-base-pretrained_lr1e-4_at0.8_da0.3
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: 2.2956
- Wer: 0.1811
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
- training_steps: 3500
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 21.7228 | 17.86 | 250 | 3.7139 | 1.0 |
| 3.1757 | 35.71 | 500 | 3.1149 | 1.0 |
| 2.8079 | 53.57 | 750 | 1.7093 | 1.0038 |
| 0.301 | 71.43 | 1000 | 1.4764 | 0.2443 |
| 0.0637 | 89.29 | 1250 | 1.7091 | 0.2200 |
| 0.0311 | 107.14 | 1500 | 1.7022 | 0.2055 |
| 0.021 | 125.0 | 1750 | 2.0876 | 0.2016 |
| 0.0154 | 142.86 | 2000 | 2.2718 | 0.1974 |
| 0.0114 | 160.71 | 2250 | 2.1541 | 0.1845 |
| 0.0089 | 178.57 | 2500 | 2.2868 | 0.1854 |
| 0.0074 | 196.43 | 2750 | 2.3831 | 0.1914 |
| 0.0062 | 214.29 | 3000 | 2.2381 | 0.1841 |
| 0.0054 | 232.14 | 3250 | 2.4147 | 0.1824 |
| 0.0056 | 250.0 | 3500 | 2.2956 | 0.1811 |
Framework versions
- Transformers 4.35.0
- Pytorch 2.0.0
- Datasets 2.14.6
- Tokenizers 0.14.1