metadata
language:
- es
license: apache-2.0
base_model: openai/whisper-small
tags:
- generated_from_trainer
datasets:
- facebook/voxpopuli
metrics:
- wer
model-index:
- name: Whisper small es - m2
results:
- task:
name: Automatic Speech Recognition
type: automatic-speech-recognition
dataset:
name: voxpopuli
type: facebook/voxpopuli
config: es
split: None
args: 'config: es, split: test, train'
metrics:
- name: Wer
type: wer
value: 10.901096153044639
Whisper small es - m2
This model is a fine-tuned version of openai/whisper-small on the voxpopuli dataset. It achieves the following results on the evaluation set:
- Loss: 0.2611
- Wer: 10.9011
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: 1e-05
- train_batch_size: 16
- 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: 2500
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 0.2532 | 0.1571 | 500 | 0.3079 | 11.9764 |
| 0.2254 | 0.3142 | 1000 | 0.2858 | 10.9469 |
| 0.2303 | 0.4713 | 1500 | 0.2729 | 11.0053 |
| 0.2213 | 0.6283 | 2000 | 0.2657 | 10.8511 |
| 0.2375 | 0.7854 | 2500 | 0.2611 | 10.9011 |
Framework versions
- Transformers 4.42.4
- Pytorch 2.3.1+cu121
- Datasets 2.19.1
- Tokenizers 0.19.1