Skip to content
ModelListedExternal
Download

Korean Kws3

This model is a fine-tuned version of Kkonjeong/wav2vec2-base-korean on the None dataset. It achieves the following results on the evaluation set: Loss: 0.3971 Accuracy: 1.0

Sourcehuggingface.co/jandyjeong/korean_kws3/tree/2982e93fac807e2f231e2c50c681063ac7ecadd9
KindOtherBase modelKkonjeong/wav2vec2-base-koreanVersionRevision 2982e93Cgrade

Model source

Kind
Other
File size
361 MB
Base model
Kkonjeong/wav2vec2-base-korean
Version
vRevision 2982e93
Updated
Oct 17, 2026
Downloads
4
Source
Hugging Face

Source description

korean_kws3

This model is a fine-tuned version of Kkonjeong/wav2vec2-base-korean on the None dataset. It achieves the following results on the evaluation set:

  • Loss: 0.3971
  • Accuracy: 1.0

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: 0.0001
  • train_batch_size: 8
  • eval_batch_size: 8
  • seed: 42
  • gradient_accumulation_steps: 4
  • total_train_batch_size: 32
  • optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
  • lr_scheduler_type: linear
  • num_epochs: 30
  • mixed_precision_training: Native AMP

Training results

Training LossEpochStepValidation LossAccuracy
No log1.022.09970.0714
No log2.042.08610.2143
No log3.061.91140.2857
No log4.081.87100.2143
No log5.0101.78750.2143
No log6.0121.58510.7143
No log7.0141.45520.9286
No log8.0161.32140.9286
No log9.0181.23840.9286
No log10.0201.19180.9286
No log11.0220.95881.0
No log12.0240.87901.0
No log13.0260.83561.0
No log14.0280.78041.0
No log15.0300.71711.0
No log16.0320.66691.0
No log17.0340.63611.0
No log18.0360.59231.0
No log19.0380.55371.0
No log20.0400.52721.0
No log21.0420.50221.0
No log22.0440.47921.0
No log23.0460.46291.0
No log24.0480.45241.0
No log25.0500.43821.0
No log26.0520.42011.0
No log27.0540.40881.0
No log28.0560.40241.0
No log29.0580.39861.0
No log30.0600.39711.0

Framework versions

  • Transformers 5.0.0
  • Pytorch 2.10.0+cu128
  • Datasets 4.0.0
  • Tokenizers 0.22.2

Sources

1 source
Hugging FacePrimaryVersion 2982e93fac807e2f231e2c50c681063ac7ecadd9
huggingface.co/jandyjeong/korean_kws3/tree/2982e93fac807e2f231e2c50c681063ac7ecadd9

Verified Aug 29

Model artifacts

1 artifact

model.safetensors

safetensors · 361 MB · SHA-256 f2323f5ef6da…6d71 · Hugging Face

Download file

Source excerpts

2 excerpts
ReadmeAug 29huggingface.co/jandyjeong/korean_kws3

--- library_name: transformers base_model: Kkonjeong/wav2vec2-base-korean tags: - generated_from_trainer metrics: - accuracy model-index: - name: korean_kws3 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. --> # korean_kws3 This model is a fine-tuned version of [Kkonjeong/wav2vec2-base-korean](https://huggingface.co/Kkonjeong/wav2vec2-base-korean) on the None dataset. It achieves the following results on the evaluation set: - Loss: 0.3971 - Accuracy: 1.0 ## 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: 0.0001 - train_batch_size: 8 - eval_batch_size: 8 - seed: 42 - gradient_accumulation_steps: 4 - total_train_batch_size: 32 - optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments - lr_scheduler_type: linear - num_epochs: 30 - mixed_precision_training: Native AMP ### Training results | Training Loss | Epoch | Step | Validation Loss | Accuracy | |:-------------:|:-----:|:----:|:---------------:|:--------:| | No log | 1.0 | 2 | 2.0997 | 0.0714 | | No log | 2.0 | 4 | 2.0861 | 0.2143 | | No log | 3.0 | 6 | 1.9114 | 0.2857 | | No log | 4.0 | 8 | 1.8710 | 0.2143 | | No log | 5.0 | 10 | 1.7875 | 0.2143 | | No log | 6.0 | 12 | 1.5851 | 0.7143 | | No log | 7.0 | 14 | 1.4552 | 0.9286 | | No log | 8.0 | 16 | 1.3214 | 0.9286 | | No log | 9.0 | 18 | 1.2384 | 0.9286 | | No log | 10.0 | 20 | 1.1918 | 0.9286 | | No log | 11.0 | 22 | 0.9588 | 1.0 | | No log | 12.0 | 24 | 0.8790 | 1.0 | | No log | 13.0 | 26 | 0.8356 | 1.0 | | No log | 14.0 | 28 | 0.7804 | 1.0 | | No log | 15.0 | 30 | 0.7171 | 1.0 | | No log | 16.0 | 32 | 0.6669 | 1.0 | | No log | 17.0 | 34 | 0.6361 | 1.0 | | No log | 18.0 | 36 | 0.5923 | 1.0 | | No log | 19.0 | 38 | 0.5537 | 1.0 | | No log | 20.0 | 40 | 0.5272 | 1.0 | | No log | 21.0 | 42 | 0.5022 | 1.0 | | No log | 22.0 | 44 | 0.4792 | 1.0 | | No log | 23.0 | 46 | 0.4629 | 1.0 | | No log | 24.0 | 48 | 0.4524 | 1.0 | | No log | 25.0 | 50 | 0.4382 | 1.0 | | No log | 26.0 | 52 | 0.420...

NoteAug 29huggingface.co/jandyjeong/korean_kws3

Source context: 4 downloads · 0 likes · Pipeline audio-classification · Library transformers · Repo jandyjeong/korean_kws3