CoolFace
Modelpublic

wangpuupup/whisper-small-nl-dy-augmentation

sourceHugging Faceapache-2.0updated 4y agoView on Hugging Face
0likes8downloads
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 dysarthric Dutch

This model is a fine-tuned version of qmeeus/whisper-small-nl on the data/copas copas-full dataset. It achieves the following results on the evaluation set:

  • —Loss: 0.4702
  • —Wer: 22.1638

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: 0.0001
  • —trainbatchsize: 32
  • —evalbatchsize: 32
  • —seed: 42
  • —gradientaccumulationsteps: 2
  • —totaltrainbatch_size: 64
  • —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
0.16180.055000.378728.9235
0.05831.0510000.373225.7702
0.03822.0515000.400125.4621
0.03163.0520000.408124.7010
0.01694.0525000.432524.1935
0.01535.0530000.432533.4179
0.00746.0535000.436723.9398
0.00967.0540000.439023.3055
0.00548.0545000.444123.7042
0.00329.0450000.449323.2693
0.00410.0455000.452423.3418
0.004811.0460000.449823.7224
0.00112.0465000.457722.8887
0.000213.0470000.457722.0913
0.000114.0475000.461622.1276
0.000115.0480000.463922.2726
0.000116.0485000.466222.1095
0.000117.0490000.468422.1457
0.000118.0495000.469722.1457
0.000119.04100000.470222.1638

Framework versions

  • —Transformers 4.26.0.dev0
  • —Pytorch 1.12.1+cu116
  • —Datasets 2.4.0
  • —Tokenizers 0.12.1