CoolFace
Modelpublic

sanchit-gandhi/wav2vec2-2-rnd-no-adapter-regularisation

sourceHugging Faceupdated 5y agoView on Hugging Face
0likes74downloads
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. -->

This model was trained from scratch on the librispeech_asr dataset. It achieves the following results on the evaluation set:

  • —Loss: 0.7177
  • —Wer: 0.1283

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: 3e-05
  • —trainbatchsize: 8
  • —evalbatchsize: 8
  • —seed: 42
  • —gradientaccumulationsteps: 4
  • —totaltrainbatch_size: 32
  • —optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • —lrschedulertype: linear
  • —lrschedulerwarmup_steps: 1000
  • —num_epochs: 25.0
  • —mixedprecisiontraining: Native AMP

Training results

Training LossEpochStepValidation LossWer
6.12281.6815006.04901.1433
5.41733.3630005.34531.4878
4.16355.0445004.41850.9644
2.12466.7360003.20890.5026
1.888.4175001.98860.3438
1.260610.0990001.44720.2487
0.749211.77105001.17160.1949
0.886813.45120001.01460.1702
0.507815.13135000.88210.1548
0.451516.82150000.81810.1417
0.390218.5165000.77650.1364
0.357520.18180000.73670.1333
0.290321.86195000.72110.1301
0.269823.54210000.71770.1283

Framework versions

  • —Transformers 4.17.0.dev0
  • —Pytorch 1.10.2+cu113
  • —Datasets 1.18.3
  • —Tokenizers 0.11.0