CoolFace
Modelpublic

shivam/xls-r-300m-hindi

sourceHugging Faceapache-2.0updated 5y agoView on Hugging Face
0likes59downloads
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 - HI dataset. It achieves the following results on the evaluation set:

  • —Loss: 0.8111
  • —Wer: 0.5177

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: 7.5e-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: 2000
  • —num_epochs: 50.0
  • —mixedprecisiontraining: Native AMP

Training results

Training LossEpochStepValidation LossWer
4.97332.595005.06971.0
3.38395.1810003.35181.0
2.05967.7715001.39920.7869
1.610210.3620001.07120.6754
1.458712.9525000.92800.6361
1.366715.5430000.92810.6155
1.304218.1335000.90370.5921
1.254420.7340000.89960.5824
1.227423.3245000.89340.5797
1.176325.9150000.86430.5760
1.14928.555000.82510.5544
1.120731.0960000.85060.5527
1.09133.6865000.83700.5366
1.061336.2770000.83450.5352
1.049538.8675000.83800.5321
1.034541.4580000.82850.5269
1.029744.0485000.78360.5141
1.02746.6390000.81200.5180
0.987649.2295000.81090.5188

Framework versions

  • —Transformers 4.16.0.dev0
  • —Pytorch 1.10.1+cu113
  • —Datasets 1.18.1.dev0
  • —Tokenizers 0.11.0