metadata
library_name: peft
language:
- ro
license: apache-2.0
base_model: openai/whisper-small
tags:
- base_model:adapter:openai/whisper-small
- lora
- transformers
datasets:
- VladS159/romanian_speech_dataset_with_5_percent_synthetic_data
metrics:
- wer
model-index:
- name: Whisper Small Ro - PEFT
results:
- task:
type: automatic-speech-recognition
name: Automatic Speech Recognition
dataset:
name: Romanian Speech Dataset + 5% Synthetic
type: VladS159/romanian_speech_dataset_with_5_percent_synthetic_data
metrics:
- type: wer
value: 106.59810174871058
name: Wer
Whisper Small Ro - PEFT
This model is a fine-tuned version of openai/whisper-small on the Romanian Speech Dataset + 5% Synthetic dataset. It achieves the following results on the evaluation set:
- Loss: 0.4248
- Wer: 106.5981
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: 0.001
- train_batch_size: 8
- eval_batch_size: 8
- seed: 42
- optimizer: Use adamw_torch_fused with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 100
- training_steps: 100
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 2.0921 | 0.0113 | 50 | 1.0811 | 95.9014 |
| 0.6328 | 0.0227 | 100 | 0.4248 | 106.5981 |
Framework versions
- PEFT 0.18.1.dev0
- Transformers 4.57.1
- Pytorch 2.9.1+rocm6.4
- Datasets 3.6.0
- Tokenizers 0.22.1