CoolFace
Modelpublic

DrishtiSharma/wav2vec2-xls-r-pa-IN-a1

sourceHugging Faceapache-2.0updated 5y agoView on Hugging Face
0likes102downloads
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 is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the MOZILLA-FOUNDATION/COMMONVOICE8_0 - PA-IN dataset. It achieves the following results on the evaluation set:

  • —Loss: 1.1508
  • —Wer: 0.4908

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: 16
  • —evalbatchsize: 8
  • —seed: 42
  • —optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • —lrschedulertype: linear
  • —lrschedulerwarmup_steps: 1500
  • —num_epochs: 100.0
  • —mixedprecisiontraining: Native AMP

Training results

Training LossEpochStepValidation LossWer
3.58419.265003.25140.9941
0.399218.5210000.87900.6107
0.240927.7815001.00120.6366
0.144737.0420001.01670.6276
0.110946.325001.06380.5653
0.079755.5630001.14470.5715
0.063664.8135001.15030.5316
0.046674.0740001.22270.5386
0.037283.3345001.12140.5225
0.023992.5950001.13750.4998

Framework versions

  • —Transformers 4.17.0.dev0
  • —Pytorch 1.10.2+cu102
  • —Datasets 1.18.2.dev0
  • —Tokenizers 0.11.0