|
---
|
|
language:
|
|
- ko
|
|
license: apache-2.0
|
|
tags:
|
|
- generated_from_trainer
|
|
base_model: openai/whisper-base
|
|
datasets:
|
|
- AIHub/noise
|
|
model-index:
|
|
- name: Whisper Base Noise Ko - Dearlie
|
|
results: []
|
|
---
|
|
|
|
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
|
should probably proofread and complete it, then remove this comment. -->
|
|
|
|
# Whisper Base Noise Ko - Dearlie
|
|
|
|
This model is a fine-tuned version of [openai/whisper-base](https://huggingface.co/openai/whisper-base) on the Noise Data dataset.
|
|
It achieves the following results on the evaluation set:
|
|
- Loss: 1.0514
|
|
- Cer: 36.7799
|
|
|
|
## 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: 0.0003
|
|
- 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: 6000
|
|
- mixed_precision_training: Native AMP
|
|
|
|
### Training results
|
|
|
|
| Training Loss | Epoch | Step | Validation Loss | Cer |
|
|
|:-------------:|:------:|:----:|:---------------:|:-------:|
|
|
| 1.6152 | 0.8780 | 1000 | 1.6157 | 68.7925 |
|
|
| 1.1303 | 1.7559 | 2000 | 1.2280 | 52.3127 |
|
|
| 0.7779 | 2.6339 | 3000 | 1.0609 | 43.5822 |
|
|
| 0.4133 | 3.5119 | 4000 | 1.0210 | 41.4074 |
|
|
| 0.175 | 4.3898 | 5000 | 1.0462 | 38.0307 |
|
|
| 0.0468 | 5.2678 | 6000 | 1.0514 | 36.7799 |
|
|
|
|
|
|
### Framework versions
|
|
|
|
- Transformers 4.41.0.dev0
|
|
- Pytorch 2.3.0+cu121
|
|
- Datasets 2.19.0
|
|
- Tokenizers 0.19.1
|
|
|
