CoolFace
Modelpublic

marziye-A/whisper-large-v3-full-youtube_80hour_7

sourceHugging Faceupdated 2y agoView on Hugging Face
1likes16downloads
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 large fa - marziye-A

This model is a fine-tuned version of openai/whisper-large on the Common Voice 15.0 dataset. It achieves the following results on the evaluation set:

  • —Loss: 0.1571
  • —Wer: 19.7418

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: 4
  • —evalbatchsize: 4
  • —seed: 42
  • —gradientaccumulationsteps: 4
  • —totaltrainbatch_size: 16
  • —optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • —lrschedulertype: linear
  • —lrschedulerwarmup_steps: 500
  • —num_epochs: 2
  • —mixedprecisiontraining: Native AMP

Training results

Training LossEpochStepValidation LossWer
0.21890.156720000.224829.0575
0.19720.313440000.203525.1376
0.19060.470160000.192325.7159
0.15950.626880000.180622.4166
0.17470.7835100000.175323.0041
0.17440.9402120000.170922.4932
0.13571.0969140000.168720.7782
0.13451.2536160000.164621.3221
0.13621.4103180000.161921.1082
0.1211.5670200000.160120.3781
0.13541.7237220000.158719.8157
0.1221.8804240000.157119.7418

Framework versions

  • —Transformers 4.45.2
  • —Pytorch 2.4.1+cu121
  • —Datasets 3.0.1
  • —Tokenizers 0.20.1