|
--- |
|
language: |
|
- ko |
|
license: apache-2.0 |
|
tags: |
|
- generated_from_trainer |
|
datasets: |
|
- Bingsu/zeroth-korean |
|
metrics: |
|
- wer |
|
pipeline_tag: automatic-speech-recognition |
|
base_model: openai/whisper-large-v2 |
|
model-index: |
|
- name: whisper-large-v2-Ko |
|
results: |
|
- task: |
|
type: automatic-speech-recognition |
|
name: Automatic Speech Recognition |
|
dataset: |
|
name: Bingsu/zeroth-korean |
|
type: Bingsu/zeroth-korean |
|
metrics: |
|
- type: wer |
|
value: 2.9 |
|
name: Wer |
|
- task: |
|
type: automatic-speech-recognition |
|
name: Automatic Speech Recognition |
|
dataset: |
|
name: google/fleurs |
|
type: google/fleurs |
|
config: ko_kr |
|
split: test |
|
metrics: |
|
- type: wer |
|
value: 20.66 |
|
name: WER |
|
--- |
|
|
|
<!-- 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-large-v2-Ko |
|
|
|
This model is a fine-tuned version of [openai/whisper-large-v2](https://huggingface.co/openai/whisper-large-v2) on the None dataset. |
|
It achieves the following results on the evaluation set: |
|
- Loss: 0.0617 |
|
- Wer: **2.9** |
|
|
|
## Model description |
|
|
|
More information needed |
|
|
|
## Intended uses & limitations |
|
|
|
More information needed |
|
|
|
## Training and evaluation data |
|
|
|
More information needed |
|
|
|
## Training procedure |
|
|
|
***** train metrics ***** |
|
epoch = 50.0 |
|
train_loss = 0.0234 |
|
train_runtime = 16:20:18.00 |
|
train_samples = 22262 |
|
train_samples_per_second = 19.042 |
|
train_steps_per_second = 0.085 |
|
|
|
### Training hyperparameters |
|
|
|
The following hyperparameters were used during training: |
|
- learning_rate: 1e-05 |
|
- train_batch_size: 32 |
|
- eval_batch_size: 16 |
|
- seed: 42 |
|
- distributed_type: multi-GPU |
|
- num_devices: 7 |
|
- total_train_batch_size: 224 |
|
- total_eval_batch_size: 112 |
|
- 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 | |
|
|:-------------:|:-----:|:----:|:---------------:|:------:| |
|
| 0.0299 | 10.0 | 1000 | 0.0745 | 0.0447 | |
|
| 0.0085 | 20.0 | 2000 | 0.0608 | 0.0353 | |
|
| 0.0036 | 30.0 | 3000 | 0.0593 | 0.0302 | |
|
| 0.0013 | 40.0 | 4000 | 0.0609 | 0.0282 | |
|
| 0.0008 | 50.0 | 5000 | 0.0617 | 0.0290 | |
|
|
|
|
|
### Framework versions |
|
|
|
- Transformers 4.27.0.dev0 |
|
- Pytorch 1.12.1+cu113 |
|
- Datasets 2.10.1 |
|
- Tokenizers 0.13.2 |
