metadata
base_model: openai/whisper-base
datasets:
- Jpep26/NoErrorDataset
language:
- ko
library_name: transformers
license: apache-2.0
metrics:
- wer
tags:
- hf-asr-leaderboard
- generated_from_trainer
model-index:
- name: Test
results:
- task:
type: automatic-speech-recognition
name: Automatic Speech Recognition
dataset:
name: NoErrorDataset
type: Jpep26/NoErrorDataset
args: 'config: ko, split: valid'
metrics:
- type: wer
value: 0.48766120044578887
name: Wer
Test
This model is a fine-tuned version of openai/whisper-base on the NoErrorDataset dataset. It achieves the following results on the evaluation set:
- Loss: 0.6020
- Cer: 0.4962
- Wer: 0.4877
- Mean: 0.4919
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: 32
- eval_batch_size: 16
- 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: 200
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Cer | Wer | Mean |
|---|---|---|---|---|---|---|
| 2.3893 | 0.3817 | 50 | 2.0607 | 0.4421 | 0.7020 | 0.5720 |
| 1.2402 | 0.7634 | 100 | 1.0999 | 0.3408 | 0.5773 | 0.4591 |
| 0.7512 | 1.1450 | 150 | 0.7303 | 0.7268 | 0.5418 | 0.6343 |
| 0.593 | 1.5267 | 200 | 0.6020 | 0.4962 | 0.4877 | 0.4919 |
Framework versions
- Transformers 4.44.2
- Pytorch 2.0.1+cu117
- Datasets 2.18.0
- Tokenizers 0.19.1