CoolFace
Modelpublic

lorenzoncina/whisper-medium-ar

sourceHugging Faceapache-2.0updated 3y agoView on Hugging Face
0likes6downloads
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 Medium Arabic

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

  • —Loss: 0.4218
  • —Wer: 47.5307

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
  • —trainbatchsize: 16
  • —evalbatchsize: 4
  • —seed: 42
  • —optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • —lrschedulertype: linear
  • —lrschedulerwarmup_steps: 500
  • —training_steps: 10000
  • —mixedprecisiontraining: Native AMP

Training results

Training LossEpochStepValidation LossWer
0.22150.110000.336149.9307
0.11341.0720000.329056.76
0.07652.0430000.340054.3947
0.04173.0140000.359952.5320
0.03643.1150000.374055.5653
0.00944.0860000.415256.4307
0.00775.0570000.421847.5307
0.00186.0280000.455650.0493
0.00126.1290000.476054.8147
0.00097.09100000.471148.7533

Framework versions

  • —Transformers 4.28.0.dev0
  • —Pytorch 2.0.0+cu117
  • —Datasets 2.11.1.dev0
  • —Tokenizers 0.13.2