CoolFace
Modelpublic

Vkt/model-facebookptbrlarge

sourceHugging Faceapache-2.0updated 4y agoView on Hugging Face
0likes331downloads
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. -->

model-facebookptbrlarge

This model is a fine-tuned version of facebook/wav2vec2-large-xlsr-53-portuguese on the common_voice dataset. It achieves the following results on the evaluation set:

  • Loss: 0.2206
  • Wer: 0.1322

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: 8
  • evalbatchsize: 8
  • seed: 42
  • gradientaccumulationsteps: 2
  • totaltrainbatch_size: 16
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lrschedulertype: linear
  • lrschedulerwarmup_steps: 500
  • num_epochs: 30
  • mixedprecisiontraining: Native AMP

Training results

Training LossEpochStepValidation LossWer
5.89750.294000.41310.3336
0.51310.578000.41030.3293
0.48460.8612000.34930.3028
0.41741.1416000.30550.2730
0.41051.4320000.32830.3041
0.40281.7224000.35390.3210
0.3862.028000.29250.2690
0.32242.2932000.28420.2665
0.31222.5736000.27810.2472
0.30872.8640000.27940.2692
0.28783.1544000.27950.2537
0.29153.4348000.27640.2478
0.28163.7252000.27610.2366
0.2834.056000.26410.2587
0.24484.2960000.24890.2417
0.2474.5764000.25380.2422
0.254.8668000.26600.2306
0.22565.1572000.24770.2267
0.22255.4376000.23640.2195
0.22175.7280000.23190.2139
0.22726.084000.24890.2427
0.20166.2988000.24040.2181
0.19736.5892000.25320.2273
0.21016.8696000.25900.2100
0.19467.15100000.24140.2108
0.18457.43104000.24850.2124
0.18617.72108000.24050.2124
0.18518.01112000.24490.2062
0.15878.29116000.25100.2048
0.16948.58120000.22900.2059
0.16378.86124000.23760.2063
0.15949.15128000.23070.1967
0.15379.44132000.22740.2017
0.14989.72136000.23220.2025
0.151610.01140000.23230.1971
0.133610.29144000.22490.1920
0.13410.58148000.22580.2055
0.13810.86152000.22500.1906
0.1311.15156000.24230.1920
0.130211.44160000.22940.1849
0.125311.72164000.21930.1889
0.121912.01168000.23500.1869
0.114912.29172000.23500.1903
0.116112.58176000.22770.1899
0.112912.87180000.24160.1855
0.109113.15184000.22890.1815
0.107313.44188000.23830.1799
0.113513.72192000.23060.1819
0.107514.01196000.22830.1742
0.097114.3200000.22710.1851
0.096714.58204000.23950.1809
0.103914.87208000.22860.1808
0.098415.15212000.23030.1821
0.092215.44216000.22540.1745
0.088215.73220000.22800.1836
0.085916.01224000.23550.1779
0.083216.3228000.23470.1740
0.085416.58232000.23420.1739
0.087416.87236000.23160.1719
0.080817.16240000.22910.1730
0.074117.44244000.23080.1674
0.081517.73248000.23290.1655
0.076418.01252000.25140.1711
0.071918.3256000.22750.1578
0.066518.58260000.23670.1614
0.069318.87264000.21850.1593
0.066219.16268000.22660.1678
0.061219.44272000.23320.1602
0.062319.73276000.22830.1670
0.065920.01280000.21420.1626
0.058120.3284000.21980.1646
0.06320.59288000.22510.1588
0.061820.87292000.21860.1554
0.054921.16296000.22510.1490
0.05821.44300000.23660.1559
0.054321.73304000.22620.1535
0.052922.02308000.23580.1519
0.05322.3312000.21980.1513
0.055222.59316000.22340.1503
0.049222.87320000.21910.1516
0.048823.16324000.23210.1500
0.047923.45328000.21520.1420
0.045323.73332000.22020.1453
0.048524.02336000.22350.1468
0.045124.3340000.21920.1455
0.04124.59344000.21380.1438
0.043524.87348000.23350.1423
0.040425.16352000.22200.1409
0.037425.45356000.23660.1437
0.040525.73360000.22330.1428
0.038526.02364000.22080.1414
0.037326.3368000.22650.1420
0.036526.59372000.21740.1402
0.03726.88376000.22490.1397
0.037927.16380000.21730.1374
0.035427.45384000.22120.1381
0.03427.73388000.23130.1364
0.034728.02392000.22300.1356
0.031828.31396000.22310.1357
0.030528.59400000.22810.1366
0.030728.88404000.22590.1342
0.031529.16408000.22520.1332
0.031429.45412000.22180.1328
0.030729.74416000.22060.1322

Framework versions

  • Transformers 4.17.0
  • Pytorch 1.8.1+cu111
  • Datasets 2.2.1
  • Tokenizers 0.12.1