CoolFace
Modelpublic

yesj1234/xlsr_cycle1_ko

sourceHugging Faceapache-2.0updated 3y agoView on Hugging Face
0likes7downloads
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. -->

ko-xlsr

This model is a fine-tuned version of facebook/wav2vec2-large-xlsr-53 on the ./SAMPLE_SPEECH.PY - NA dataset. It achieves the following results on the evaluation set:

  • —Loss: 0.5156
  • —Cer: 0.1228

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: 4
  • —evalbatchsize: 4
  • —seed: 42
  • —distributed_type: multi-GPU
  • —num_devices: 4
  • —gradientaccumulationsteps: 8
  • —totaltrainbatch_size: 128
  • —totalevalbatch_size: 16
  • —optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • —lrschedulertype: linear
  • —lrschedulerwarmup_steps: 500
  • —num_epochs: 40

Training results

Training LossEpochStepValidation LossCer
1.7783.3110001.27730.3050
1.10376.6320000.77160.1888
0.95299.9430000.67260.1659
0.842413.2640000.61380.1512
0.76716.5750000.58850.1433
0.720119.8860000.56820.1378
0.66423.270000.55830.1333
0.629626.5180000.54160.1298
0.602129.8390000.53770.1272
0.56833.14100000.52410.1246
0.551936.45110000.51840.1228
0.539539.77120000.51560.1227

Framework versions

  • —Transformers 4.33.0
  • —Pytorch 2.0.1+cu117
  • —Datasets 2.14.4
  • —Tokenizers 0.13.3