metadata
library_name: transformers
language:
- en
license: apache-2.0
base_model: openai/whisper-large-v3
tags:
- generated_from_trainer
datasets:
- stillerman/libristutter-4.7k
metrics:
- wer
model-index:
- name: Whisper Large V3 Stutter - Ariel Cerda
results:
- task:
name: Automatic Speech Recognition
type: automatic-speech-recognition
dataset:
name: Libristutter 4.7k
type: stillerman/libristutter-4.7k
args: 'config: en, split: test'
metrics:
- name: Wer
type: wer
value: 18.279313632030505
Whisper Large V3 Stutter - Ariel Cerda
This model is a fine-tuned version of openai/whisper-large-v3 on the Libristutter 4.7k dataset. It achieves the following results on the evaluation set:
- Loss: 0.4938
- Wer: 18.2793
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: 5000
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 0.0354 | 3.7453 | 1000 | 0.3009 | 18.2972 |
| 0.0028 | 7.4906 | 2000 | 0.4106 | 16.5157 |
| 0.0004 | 11.2360 | 3000 | 0.4474 | 20.5076 |
| 0.0002 | 14.9813 | 4000 | 0.4774 | 17.5941 |
| 0.0001 | 18.7266 | 5000 | 0.4938 | 18.2793 |
Framework versions
- Transformers 4.44.2
- Pytorch 2.4.1+cu121
- Datasets 3.0.1
- Tokenizers 0.19.1