CoolFace
Modelpublic

asusevski/whisper-small-sr

sourceHugging Faceapache-2.0updated 4y agoView on Hugging Face
2likes13downloads
Model Card

<!-- 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 Small Serbian

This model is a fine-tuned version of openai/whisper-small on the mozilla-foundation/commonvoice11_0 sr dataset. It achieves the following results on the evaluation set:

  • —Loss: 0.5670
  • —Wer: 22.9057

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

The model was trained on the training + validation splits of the Serbian split of the Common Voice dataset and evaluated on the test split of the Serbian split from the Common Voice dataset.

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

  • —learning_rate: 1e-05
  • —trainbatchsize: 64
  • —evalbatchsize: 32
  • —seed: 42
  • —optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • —lrschedulertype: linear
  • —lrschedulerwarmup_steps: 500
  • —training_steps: 800
  • —mixedprecisiontraining: Native AMP

Training results

Training LossEpochStepValidation LossWer
0.42575.01000.437732.4160
0.077910.02000.392823.7556
0.010815.03000.485623.4318
0.010420.04000.563725.4958
0.006925.05000.528923.1485
0.002230.06000.567022.9057
0.001235.07000.574623.0271
0.000640.08000.581023.1890

Framework versions

  • —Transformers 4.26.0.dev0
  • —Pytorch 1.13.1+cu117
  • —Datasets 2.8.1.dev0
  • —Tokenizers 0.13.2