metadata
base_model: openai/whisper-base
datasets:
- Jpep26/AfterProcessing
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: AfterProcessing
type: Jpep26/AfterProcessing
args: 'config: ko, split: valid'
metrics:
- type: wer
value: 0.7054380664652568
name: Wer
Test
This model is a fine-tuned version of openai/whisper-base on the AfterProcessing dataset. It achieves the following results on the evaluation set:
- Loss: 0.9167
- Cer: 0.8486
- Wer: 0.7054
- Mean: 0.7770
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: 200
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Cer | Wer | Mean |
|---|---|---|---|---|---|---|
| 2.9457 | 0.6410 | 50 | 2.6739 | 0.5791 | 0.8056 | 0.6923 |
| 1.7821 | 1.2821 | 100 | 1.6827 | 0.4622 | 0.6561 | 0.5592 |
| 1.2153 | 1.9231 | 150 | 1.1411 | 0.4216 | 0.6022 | 0.5119 |
| 0.8636 | 2.5641 | 200 | 0.9167 | 0.8486 | 0.7054 | 0.7770 |
Framework versions
- Transformers 4.45.0.dev0
- Pytorch 2.1.2+cu121
- Datasets 2.20.0
- Tokenizers 0.19.1