CoolFace
Modelpublic

infinitejoy/wav2vec2-large-xls-r-300m-abkhaz-cv8

sourceHugging Faceapache-2.0updated 5y agoView on Hugging Face
2likes64downloads
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. -->

wav2vec2-large-xls-r-300m-abkhaz-cv8

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the MOZILLA-FOUNDATION/COMMONVOICE8_0 - AB dataset. It achieves the following results on the evaluation set:

  • —Loss: 0.1614
  • —Wer: 0.2907

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: 7e-05
  • —trainbatchsize: 32
  • —evalbatchsize: 16
  • —seed: 42
  • —optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • —lrschedulertype: linear
  • —lrschedulerwarmup_steps: 4000
  • —num_epochs: 50.0
  • —mixedprecisiontraining: Native AMP

Training results

Training LossEpochStepValidation LossWer
1.28814.2640000.37640.6461
1.07678.5380000.26570.5164
0.984112.79120000.23300.4445
0.927417.06160000.21340.3929
0.878121.32200000.19450.3886
0.838125.59240000.18400.3737
0.805429.85280000.17560.3523
0.776334.12320000.17450.3299
0.747438.38360000.16770.3074
0.729842.64400000.16490.2963
0.712546.91440000.16170.2931

Framework versions

  • —Transformers 4.16.0.dev0
  • —Pytorch 1.10.1+cu102
  • —Datasets 1.18.3
  • —Tokenizers 0.11.0