CoolFace
Modelpublic

r-sharma-coder/hubert-large-ll60k-librispeech-multi-gpu

sourceHugging Faceapache-2.0updated 3y agoView on Hugging Face
0likes9downloads
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. -->

hubert-large-ll60k-librispeech-multi-gpu

This model is a fine-tuned version of facebook/hubert-large-ll60k on the MACABDUL9/LIBRISPEECH-HUBERT-DISCRETE-TOKENS - NA dataset. It achieves the following results on the evaluation set:

  • Loss: 0.0616
  • Wer: 0.0460

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: 2
  • evalbatchsize: 8
  • seed: 42
  • distributed_type: multi-GPU
  • num_devices: 4
  • totaltrainbatch_size: 8
  • totalevalbatch_size: 32
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lrschedulertype: linear
  • lrschedulerwarmup_steps: 500
  • num_epochs: 4.0

Training results

Training LossEpochStepValidation LossWer
2.96920.031002.93361.0
2.79360.062002.85141.0
0.88920.083000.71730.5438
0.47250.114000.42930.3818
0.53470.145000.34740.3284
0.31430.176000.29660.2852
0.41440.27000.24980.2383
0.5160.228000.21560.2174
0.25370.259000.18100.1877
0.30080.2810000.16990.1737
0.30780.3111000.14850.1600
0.29220.3412000.14440.1448
0.23030.3613000.13190.1334
0.18580.3914000.12860.1353
0.08370.4215000.11620.1262
0.1770.4516000.10970.1100
0.34760.4817000.11400.1114
0.25380.518000.11180.1086
0.12960.5319000.10520.1016
0.39050.5620000.10340.0950
0.20850.5921000.10010.1003
0.35630.6222000.09740.0958
0.12580.6423000.10670.1018
0.09720.6724000.09380.0930
0.25880.725000.09130.0861
0.20660.7326000.09460.0902
0.17060.7627000.09440.0956
0.11290.7828000.09180.0904
0.18050.8129000.09130.0890
0.23280.8430000.09070.0893
0.26990.8731000.09180.0849
0.16530.932000.09020.0827
0.26370.9233000.08730.0836
0.25110.9534000.09020.0851
0.20530.9835000.09070.0842
0.07191.0136000.08080.0752
0.06331.0437000.08740.0775
0.0631.0738000.08550.0757
0.06531.0939000.09160.0766
0.03811.1240000.08640.0756
0.04571.1541000.08500.0761
0.03991.1842000.08420.0842
0.04031.2143000.08340.0754
0.04621.2344000.08330.0753
0.03121.2645000.08150.0731
0.04321.2946000.08120.0748
0.0321.3247000.07830.0683
0.05021.3548000.07860.0699
0.07431.3749000.08390.0678
0.02291.450000.07710.0664
0.06721.4351000.07820.0662
0.07581.4652000.08080.0696
0.03091.4953000.07830.0710
0.0431.5154000.07480.0682
0.0761.5455000.07780.0685
0.0411.5756000.07920.0726
0.07841.657000.07630.0639
0.0971.6358000.07510.0668
0.09151.6559000.07480.0661
0.05111.6860000.07670.0641
0.02541.7161000.07540.0656
0.0241.7462000.07320.0657
0.0411.7763000.07130.0673
0.0341.7964000.06860.0647
0.06041.8265000.07330.0658
0.04561.8566000.07430.0658
0.04311.8867000.07600.0637
0.11491.9168000.07570.0620
0.04311.9369000.07270.0618
0.04251.9670000.07170.0619
0.04521.9971000.07080.0637
0.01722.0272000.07400.0620
0.04132.0573000.06950.0628
0.0182.0774000.07490.0626
0.03032.175000.07230.0625
0.01962.1376000.07410.0629
0.0152.1677000.06990.0583
0.01452.1978000.07350.0602
0.02872.2179000.07260.0619
0.07352.2480000.07050.0576
0.01612.2781000.07150.0563
0.05492.382000.06920.0563
0.02272.3383000.07150.0545
0.01562.3584000.07020.0536
0.0452.3885000.06720.0533
0.02862.4186000.06700.0537
0.01762.4487000.06930.0554
0.01282.4788000.06960.0549
0.02792.4989000.06850.0541
0.03282.5290000.06740.0574
0.04162.5591000.06740.0581
0.03642.5892000.06420.0540
0.02232.6193000.06860.0544
0.01622.6394000.06800.0521
0.03922.6695000.06460.0512
0.0272.6996000.06630.0544
0.01922.7297000.06540.0529
0.03392.7598000.06540.0526
0.02692.7799000.06510.0545
0.02632.8100000.06360.0521
0.04442.83101000.06500.0520
0.03222.86102000.06220.0513
0.01712.89103000.06410.0516
0.01212.91104000.06330.0515
0.03472.94105000.06100.0518
0.04322.97106000.06050.0512
0.01873.0107000.06230.0501
0.01473.03108000.06520.0486
0.01123.05109000.06480.05
0.01463.08110000.06380.0514
0.00643.11111000.06650.0486
0.02233.14112000.06540.0487
0.02613.17113000.06550.0488
0.01443.2114000.06380.0488
0.01953.22115000.06550.0491
0.00913.25116000.06480.0492
0.01713.28117000.06330.0483
0.05383.31118000.06540.0488
0.00473.34119000.06250.0490
0.0153.36120000.06340.0476
0.03233.39121000.06550.0494
0.01733.42122000.06370.0495
0.01883.45123000.06380.0479
0.0043.48124000.06370.0459
0.0223.5125000.06370.0467
0.01393.53126000.06380.0466
0.0243.56127000.06330.0467
0.00473.59128000.06500.0468
0.00993.62129000.06360.0476
0.04183.64130000.06310.0473
0.02983.67131000.06200.0476
0.03823.7132000.06170.0468
0.02813.73133000.06190.0464
0.02563.76134000.06210.0463
0.00243.78135000.06120.0463
0.01073.81136000.06110.0456
0.00843.84137000.06130.0456
0.00393.87138000.06140.0463
0.00323.9139000.06110.0461
0.00433.92140000.06140.0458
0.00733.95141000.06150.0460
0.01453.98142000.06160.0460

Framework versions

  • Transformers 4.39.0.dev0
  • Pytorch 2.2.1
  • Datasets 2.18.0
  • Tokenizers 0.15.1