CoolFace
Modelpublic

aghannam/whisper-medium-ar-original

sourceHugging Faceapache-2.0updated 4y agoView on Hugging Face
0likes16downloads
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-ar-original

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

  • —Loss: 0.1852
  • —Wer: 14.1086

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: 24
  • —evalbatchsize: 24
  • —seed: 42
  • —optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • —lrschedulertype: linear
  • —lrschedulerwarmup_steps: 500
  • —training_steps: 8000
  • —mixedprecisiontraining: Native AMP

Training results

Training LossEpochStepValidation LossWer
0.1151.014000.120418.6541
0.07742.028000.107415.5844
0.04383.0312000.116016.4699
0.02334.0416000.127915.1122
0.01315.0520000.135015.5254
0.00516.0624000.145514.9941
0.00357.0728000.146414.1677
0.00328.0832000.154514.8170
0.00139.0936000.162313.8725
0.001310.140000.154313.4002
0.000611.1144000.165314.1677
0.000612.1248000.169913.7544
0.000313.1352000.170513.4593
0.000114.1456000.173313.6954
0.000215.1560000.176813.8725
0.000116.1664000.178613.7544
0.017.1768000.182613.9906
0.018.1872000.183914.0496
0.019.1976000.184814.0496
0.020.280000.185214.1086

Framework versions

  • —Transformers 4.25.1
  • —Pytorch 1.12.1
  • —Datasets 2.8.0
  • —Tokenizers 0.13.2