CoolFace
Modelpublic

AigizK/bashkir-whisper-small

sourceHugging Faceapache-2.0updated 4y agoView on Hugging Face
0likes11downloads
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 Bashkir

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

  • —Loss: 0.2589
  • —Wer: 15.0723

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: 8
  • —seed: 42
  • —optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • —lrschedulertype: linear
  • —lrschedulerwarmup_steps: 500
  • —training_steps: 30000
  • —mixedprecisiontraining: Native AMP

Training results

Training LossEpochStepValidation LossWer
0.16371.0120000.255526.4682
0.13752.0140000.222321.5394
0.08513.0260000.208619.6725
0.05734.0280000.217818.4280
0.0365.03100000.231217.8248
0.02386.04120000.262117.4096
0.07337.04140000.212016.5656
0.01118.05160000.268216.2291
0.01559.05180000.267715.9242
0.004110.06200000.317815.9534
0.002312.01220000.321816.0536
0.062113.01240000.231315.6169
0.002214.02260000.288715.1083
0.019915.02280000.255315.1848
0.008316.03300000.258915.0723

Framework versions

  • —Transformers 4.26.0.dev0
  • —Pytorch 1.13.0+cu117
  • —Datasets 2.7.1.dev0
  • —Tokenizers 0.13.2