sarpba's picture
Update README.md
a65312a verified
metadata
library_name: transformers
language:
  - hu
license: apache-2.0
base_model: openai/whisper-small
tags:
  - generated_from_trainer
metrics:
  - wer
model-index:
  - name: whisper-small-finetuned-hu
    results: []

whisper-small-finetuned-hu

This model is a fine-tuned version of openai/whisper-small on the custom dataset. It achieves the following results on the evaluation set:

  • Loss: 0.02658
  • Wer: 0.08494

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: 2.5e-05
  • train_batch_size: 32
  • eval_batch_size: 32
  • seed: 42
  • distributed_type: multi-GPU
  • num_devices: 2
  • total_train_batch_size: 64
  • total_eval_batch_size: 64
  • optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
  • lr_scheduler_type: linear
  • lr_scheduler_warmup_steps: 500
  • num_epochs: 2
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss Wer
0.0559 0.0902 2000 0.0575 0.2634
0.0481 0.1804 4000 0.0488 0.1917
0.0415 0.2707 6000 0.0438 0.1329
0.0408 0.3609 8000 0.0408 0.1234
0.0393 0.4511 10000 0.0388 0.1173
0.0375 0.5413 12000 0.0372 0.1119
0.0342 0.6316 14000 0.0357 0.1101
0.0335 0.7218 16000 0.0349 0.1071
0.0323 0.8120 18000 0.0331 0.1037
0.0325 0.9022 20000 0.0326 0.1035
0.0305 0.9925 22000 0.0315 0.0974
0.02 1.0827 24000 0.0312 0.0992
0.0207 1.1729 26000 0.0310 0.0937
0.0203 1.2631 28000 0.0301 0.0941
0.0215 1.3534 30000 0.0296 0.0913
0.0199 1.4436 32000 0.0289 0.0911
0.0197 1.5338 34000 0.0285 0.0890
0.0187 1.6240 36000 0.0279 0.0887
0.0188 1.7143 38000 0.0276 0.0882
0.0186 1.8045 40000 0.0271 0.0856
0.0181 1.8947 42000 0.0266 0.0849
0.0176 1.9849 44000 0.0264 0.0863

Framework versions

  • Transformers 4.47.0
  • Pytorch 2.5.1+cu118
  • Datasets 3.1.0
  • Tokenizers 0.21.0