CoolFace
Modelpublic

oliverguhr/wav2vec2-large-xlsr-53-german-cv9

sourceHugging Faceapache-2.0updated 3y agoView on Hugging Face
1likes181downloads
Model Card

wav2vec2-large-xlsr-53-german-cv9

This model is a fine-tuned version of ./facebook/wav2vec2-large-xlsr-53 on the MOZILLA-FOUNDATION/COMMONVOICE9_0 - DE dataset.

It achieves the following results on the test set:

  • —CER: 2.273015898213336
  • —Wer: 9.480663281840769

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: 0.0001
  • —trainbatchsize: 16
  • —evalbatchsize: 32
  • —seed: 42
  • —gradientaccumulationsteps: 8
  • —totaltrainbatch_size: 128
  • —optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • —lrschedulertype: linear
  • —lrschedulerwarmup_ratio: 0.1
  • —num_epochs: 50.0
  • —mixedprecisiontraining: Native AMP

Training results

Training LossEpochStepValidation LossEval Wer
0.41291.035570.30150.2499
0.21212.071140.15960.1567
0.14553.0106710.13770.1354
0.14364.0142280.13010.1282
0.11445.0177850.12250.1245
0.12196.0213420.12540.1208
0.1047.0248990.11980.1232
0.10168.0284560.11490.1174
0.10939.0320130.11860.1186
0.085810.0355700.11820.1164
0.10211.0391270.11910.1186
0.083412.0426840.11610.1096
0.091613.0462410.11470.1107
0.081114.0497980.11740.1136
0.081415.0533550.11320.1114
0.086516.0569120.11340.1097
0.070117.0604690.10960.1054
0.089118.0640260.11100.1076
0.07119.0675830.11410.1074
0.072620.0711400.10940.1093
0.064721.0746970.10880.1095
0.064322.0782540.11050.1044
0.076423.0818110.10720.1042
0.060524.0853680.10950.1026
0.072225.0889250.11440.1066
0.059726.0924820.10870.1022
0.06227.0960390.10730.1027
0.053628.0995960.10680.1027
0.061629.01031530.10970.1037
0.064230.01067100.11170.1020
0.055531.01102670.11090.0990
0.063232.01138240.11040.0977
0.048233.01173810.11080.0958
0.060134.01209380.10950.0957
0.050835.01244950.10790.0973
0.052636.01280520.10680.0967
0.048737.01316090.10810.0966
0.049538.01351660.10990.0956
0.052839.01387230.10910.0923
0.043940.01422800.11110.0928
0.046741.01458370.11310.0943
0.040742.01493940.11150.0944
0.04643.01529510.11060.0935
0.044744.01565080.10830.0919
0.043445.01600650.10930.0909
0.047246.01636220.10920.0921
0.041447.01671790.11060.0922
0.050148.01707360.10940.0918
0.038849.01742930.10990.0918
0.042850.01778500.11030.0915

Framework versions

  • —Transformers 4.19.0.dev0
  • —Pytorch 1.11.0+cu113
  • —Datasets 2.0.0
  • —Tokenizers 0.11.6