CoolFace
Modelpublic

mageec/wave2vec2_capstone

sourceHugging Faceapache-2.0updated 2y agoView on Hugging Face
0likes8downloads
Model Card

wave2vec2_capstone

This model is a fine-tuned version of facebook/wav2vec2-base on the commonvoice16_1 dataset. It achieves the following results on the evaluation set:

  • —Loss: 0.2796
  • —Accuracy: 0.9400
  • —F1 score: 0.9399

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: 0.0003
  • —trainbatchsize: 9
  • —evalbatchsize: 9
  • —seed: 42
  • —gradientaccumulationsteps: 12
  • —totaltrainbatch_size: 108
  • —optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • —lrschedulertype: linear
  • —lrschedulerwarmup_ratio: 0.1
  • —num_epochs: 8
  • —mixedprecisiontraining: Native AMP

Training results

Training LossEpochStepValidation LossAccuracyF1 score
0.89511.07761.16170.66510.6607
0.66082.015520.63450.81880.8188
0.44263.023280.47920.86720.8677
0.35764.031050.38260.89170.8929
0.1945.038810.32550.91250.9130
0.16356.046570.29030.92080.9206
0.09037.054330.29900.93000.9299
0.04058.062080.27960.94000.9399

Framework versions

  • —Transformers 4.36.2
  • —Pytorch 2.1.2+cu121
  • —Datasets 2.16.0
  • —Tokenizers 0.15.0