metadata
license: apache-2.0
base_model: facebook/wav2vec2-base
tags:
- generated_from_trainer
datasets:
- vivos
metrics:
- wer
model-index:
- name: wav2vec2-vivos-asr
results:
- task:
name: Automatic Speech Recognition
type: automatic-speech-recognition
dataset:
name: vivos
type: vivos
config: default
split: None
args: default
metrics:
- name: Wer
type: wer
value: 0.39171506989212995
wav2vec2-vivos-asr
This model is a fine-tuned version of facebook/wav2vec2-base on the vivos dataset. It achieves the following results on the evaluation set:
- Loss: 0.7105
- Wer: 0.3917
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
- gradient_accumulation_steps: 2
- total_train_batch_size: 64
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_steps: 400
- num_epochs: 20
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 5.7574 | 2.0 | 292 | 3.6165 | 1.0 |
| 3.4366 | 4.0 | 584 | 3.5022 | 1.0 |
| 2.7297 | 6.0 | 876 | 1.4138 | 0.7707 |
| 1.0147 | 8.0 | 1168 | 0.8998 | 0.5443 |
| 0.6909 | 10.0 | 1460 | 0.7924 | 0.4759 |
| 0.5341 | 12.0 | 1752 | 0.7368 | 0.4337 |
| 0.4472 | 14.0 | 2044 | 0.7149 | 0.4063 |
| 0.4269 | 16.0 | 2336 | 0.7197 | 0.4002 |
| 0.3627 | 18.0 | 2628 | 0.7151 | 0.3961 |
| 0.3487 | 20.0 | 2920 | 0.7105 | 0.3917 |
Framework versions
- Transformers 4.42.3
- Pytorch 2.1.2
- Datasets 2.20.0
- Tokenizers 0.19.1