metadata
language:
- ko
license: apache-2.0
tags:
- hf-asr-leaderboard
- generated_from_trainer
datasets:
- hyojin99/EBRC
base_model: openai/whisper-base
model-index:
- name: ft_model
results: []
ft_model
This model is a fine-tuned version of openai/whisper-base on the EBRC dataset. It achieves the following results on the evaluation set:
- Loss: 0.4127
- Cer: 18.2381
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: 50
- training_steps: 12000
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Cer |
|---|---|---|---|---|
| 0.5845 | 0.4 | 1000 | 0.5512 | 34.8150 |
| 0.4598 | 0.8 | 2000 | 0.4840 | 22.6069 |
| 0.3077 | 1.2 | 3000 | 0.4570 | 20.7128 |
| 0.3212 | 1.6 | 4000 | 0.4381 | 21.4198 |
| 0.3027 | 2.0 | 5000 | 0.4181 | 20.1164 |
| 0.219 | 2.4 | 6000 | 0.4180 | 19.6479 |
| 0.2373 | 2.8 | 7000 | 0.4089 | 18.6477 |
| 0.1342 | 3.2 | 8000 | 0.4127 | 18.3603 |
| 0.1601 | 3.6 | 9000 | 0.4104 | 18.4824 |
| 0.1489 | 4.0 | 10000 | 0.4084 | 18.0628 |
| 0.1308 | 4.4 | 11000 | 0.4134 | 18.2856 |
| 0.114 | 4.8 | 12000 | 0.4127 | 18.2381 |
Framework versions
- Transformers 4.40.0.dev0
- Pytorch 2.2.1+cu121
- Datasets 2.18.0
- Tokenizers 0.15.2