Whisper Tiny PT with Common Voice 11

This model is a fine-tuned version of openai/whisper-tiny on the Common Voice 11.0 dataset. It achieves the following results on the evaluation set:

  • Loss: 0.5205
  • Wer: 33.2447

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: 8
  • eval_batch_size: 1
  • seed: 42
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • training_steps: 16000

Training results

Training Loss Epoch Step Validation Loss Wer
0.3154 0.44 1000 0.4987 36.2196
0.3252 0.88 2000 0.4586 33.6213
0.1989 1.32 3000 0.4457 32.7455
0.3112 1.76 4000 0.4356 31.4097
0.1329 2.2 5000 0.4348 31.1559
0.1193 2.64 6000 0.4343 31.4046
0.0723 3.07 7000 0.4424 31.5869
0.0698 3.51 8000 0.4497 32.0827
0.0865 3.95 9000 0.4497 31.0945
0.0522 4.39 10000 0.4716 32.2190
0.0542 4.83 11000 0.4761 32.6944
0.061 5.27 12000 0.4983 32.0691
0.0459 5.71 13000 0.4985 32.4968
0.0338 6.15 14000 0.5123 33.3129
0.0492 6.59 15000 0.5217 33.2686
0.0194 7.03 16000 0.5205 33.2447

Framework versions

  • Transformers 4.25.0.dev0
  • Pytorch 1.13.0+cu117
  • Datasets 2.6.1
  • Tokenizers 0.13.1
Downloads last month
15
Inference Examples
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social visibility and check back later, or deploy to Inference Endpoints (dedicated) instead.

Dataset used to train lgris/whisper-tiny-cv11-pt

Evaluation results