Edit model card

whisper-small-id

This model is a fine-tuned version of openai/whisper-small on the mozilla-foundation/common_voice_17_0 id dataset. It achieves the following results on the evaluation set:

  • Loss: 0.0878
  • Wer: 0.0590 (5.9%)

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: 16
  • seed: 42
  • gradient_accumulation_steps: 2
  • total_train_batch_size: 32
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • lr_scheduler_warmup_steps: 500
  • training_steps: 20000
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss Wer
0.1875 0.8457 1000 0.1400 0.1099
0.0852 1.6913 2000 0.1043 0.0857
0.0387 2.5370 3000 0.0914 0.0757
0.0153 3.3827 4000 0.0860 0.0818
0.008 4.2283 5000 0.0878 0.0698
0.005 5.0740 6000 0.0878 0.0745
0.0033 5.9197 7000 0.0834 0.0651
0.0029 6.7653 8000 0.0815 0.0627
0.0014 7.6110 9000 0.0853 0.0627
0.0013 8.4567 10000 0.0861 0.0641
0.0005 9.3023 11000 0.0857 0.0633
0.0005 10.1480 12000 0.0856 0.0620
0.0007 10.9937 13000 0.0866 0.0605
0.0005 11.8393 14000 0.0871 0.0614
0.0002 12.6850 15000 0.0850 0.0596
0.0004 13.5307 16000 0.0849 0.0600
0.0001 14.3763 17000 0.0868 0.0592
0.0002 15.2220 18000 0.0873 0.0593
0.0001 16.0677 19000 0.0875 0.0585
0.0001 16.9133 20000 0.0878 0.0590

Framework versions

  • Transformers 4.42.0.dev0
  • Pytorch 2.3.0+cu121
  • Datasets 2.19.1
  • Tokenizers 0.19.1
Downloads last month
3
Safetensors
Model size
242M params
Tensor type
F32
·

Finetuned from

Dataset used to train Bagus/whisper-small-id-cv17

Evaluation results