metadata
license: apache-2.0
tags:
- whisper-event
- generated_from_trainer
datasets:
- facebook/voxpopuli
metrics:
- wer
base_model: openai/whisper-small
model-index:
- name: Whisper Small Croatian
results:
- task:
type: automatic-speech-recognition
name: Automatic Speech Recognition
dataset:
name: facebook/voxpopuli
type: facebook/voxpopuli
config: hr
split: test
args: hr
metrics:
- type: wer
value: 25.37305060024683
name: Wer
Whisper Small Croatian
This model is a fine-tuned version of openai/whisper-small on the facebook/voxpopuli hr dataset. It achieves the following results on the evaluation set:
- Loss: 0.6325
- Wer: 25.3731
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: 64
- eval_batch_size: 32
- 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: 5000
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 0.0028 | 24.01 | 1000 | 0.5303 | 26.1752 |
| 0.0006 | 49.01 | 2000 | 0.5849 | 25.4123 |
| 0.0003 | 74.01 | 3000 | 0.6141 | 25.6311 |
| 0.0002 | 99.01 | 4000 | 0.6325 | 25.3731 |
| 0.0002 | 124.01 | 5000 | 0.6405 | 25.4348 |
Framework versions
- Transformers 4.26.0.dev0
- Pytorch 1.13.0+cu117
- Datasets 2.7.1.dev0
- Tokenizers 0.13.2