anton-l's picture
anton-l HF staff
Update README.md
8574177
metadata
language:
  - all
license: apache-2.0
tags:
  - fleurs-lang_id
  - google/xtreme_s
  - generated_from_trainer
datasets:
  - google/xtreme_s
metrics:
  - accuracy
model-index:
  - name: xtreme_s_xlsr_300m_fleurs_langid
    results: []

xtreme_s_xlsr_300m_fleurs_langid

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the GOOGLE/XTREME_S - FLEURS.ALL dataset. It achieves the following results on the evaluation set:

  • Accuracy: 0.7271
  • Accuracy Af Za: 0.3865
  • Accuracy Am Et: 0.8818
  • Accuracy Ar Eg: 0.9977
  • Accuracy As In: 0.9858
  • Accuracy Ast Es: 0.8362
  • Accuracy Az Az: 0.8386
  • Accuracy Be By: 0.4085
  • Accuracy Bn In: 0.9989
  • Accuracy Bs Ba: 0.2508
  • Accuracy Ca Es: 0.6947
  • Accuracy Ceb Ph: 0.9852
  • Accuracy Cmn Hans Cn: 0.9799
  • Accuracy Cs Cz: 0.5353
  • Accuracy Cy Gb: 0.9716
  • Accuracy Da Dk: 0.6688
  • Accuracy De De: 0.7807
  • Accuracy El Gr: 0.7692
  • Accuracy En Us: 0.9815
  • Accuracy Es 419: 0.9846
  • Accuracy Et Ee: 0.5230
  • Accuracy Fa Ir: 0.8462
  • Accuracy Ff Sn: 0.2348
  • Accuracy Fi Fi: 0.9978
  • Accuracy Fil Ph: 0.9564
  • Accuracy Fr Fr: 0.9852
  • Accuracy Ga Ie: 0.8468
  • Accuracy Gl Es: 0.5016
  • Accuracy Gu In: 0.973
  • Accuracy Ha Ng: 0.9163
  • Accuracy He Il: 0.8043
  • Accuracy Hi In: 0.9354
  • Accuracy Hr Hr: 0.3654
  • Accuracy Hu Hu: 0.8044
  • Accuracy Hy Am: 0.9914
  • Accuracy Id Id: 0.9869
  • Accuracy Ig Ng: 0.9360
  • Accuracy Is Is: 0.0217
  • Accuracy It It: 0.8
  • Accuracy Ja Jp: 0.7385
  • Accuracy Jv Id: 0.5824
  • Accuracy Ka Ge: 0.8611
  • Accuracy Kam Ke: 0.4184
  • Accuracy Kea Cv: 0.8692
  • Accuracy Kk Kz: 0.8727
  • Accuracy Km Kh: 0.7030
  • Accuracy Kn In: 0.9630
  • Accuracy Ko Kr: 0.9843
  • Accuracy Ku Arab Iq: 0.9577
  • Accuracy Ky Kg: 0.8936
  • Accuracy Lb Lu: 0.8897
  • Accuracy Lg Ug: 0.9253
  • Accuracy Ln Cd: 0.9644
  • Accuracy Lo La: 0.1580
  • Accuracy Lt Lt: 0.4686
  • Accuracy Luo Ke: 0.9922
  • Accuracy Lv Lv: 0.6498
  • Accuracy Mi Nz: 0.9613
  • Accuracy Mk Mk: 0.7636
  • Accuracy Ml In: 0.6962
  • Accuracy Mn Mn: 0.8462
  • Accuracy Mr In: 0.3911
  • Accuracy Ms My: 0.3632
  • Accuracy Mt Mt: 0.6188
  • Accuracy My Mm: 0.9705
  • Accuracy Nb No: 0.6891
  • Accuracy Ne Np: 0.8994
  • Accuracy Nl Nl: 0.9093
  • Accuracy Nso Za: 0.8873
  • Accuracy Ny Mw: 0.4691
  • Accuracy Oci Fr: 0.1533
  • Accuracy Om Et: 0.9512
  • Accuracy Or In: 0.5447
  • Accuracy Pa In: 0.8153
  • Accuracy Pl Pl: 0.7757
  • Accuracy Ps Af: 0.8105
  • Accuracy Pt Br: 0.7715
  • Accuracy Ro Ro: 0.4122
  • Accuracy Ru Ru: 0.9794
  • Accuracy Rup Bg: 0.9468
  • Accuracy Sd Arab In: 0.5245
  • Accuracy Sk Sk: 0.8624
  • Accuracy Sl Si: 0.0300
  • Accuracy Sn Zw: 0.8843
  • Accuracy So So: 0.8803
  • Accuracy Sr Rs: 0.0257
  • Accuracy Sv Se: 0.0145
  • Accuracy Sw Ke: 0.9199
  • Accuracy Ta In: 0.9526
  • Accuracy Te In: 0.9788
  • Accuracy Tg Tj: 0.9883
  • Accuracy Th Th: 0.9912
  • Accuracy Tr Tr: 0.7887
  • Accuracy Uk Ua: 0.0627
  • Accuracy Umb Ao: 0.7863
  • Accuracy Ur Pk: 0.0134
  • Accuracy Uz Uz: 0.4014
  • Accuracy Vi Vn: 0.7246
  • Accuracy Wo Sn: 0.4555
  • Accuracy Xh Za: 1.0
  • Accuracy Yo Ng: 0.7353
  • Accuracy Yue Hant Hk: 0.7985
  • Accuracy Zu Za: 0.4696
  • Loss: 1.3789
  • Loss Af Za: 2.6778
  • Loss Am Et: 0.4615
  • Loss Ar Eg: 0.0149
  • Loss As In: 0.0764
  • Loss Ast Es: 0.4560
  • Loss Az Az: 0.5677
  • Loss Be By: 1.9231
  • Loss Bn In: 0.0024
  • Loss Bs Ba: 2.4954
  • Loss Ca Es: 1.2632
  • Loss Ceb Ph: 0.0426
  • Loss Cmn Hans Cn: 0.0650
  • Loss Cs Cz: 1.9334
  • Loss Cy Gb: 0.1274
  • Loss Da Dk: 1.4990
  • Loss De De: 0.8820
  • Loss El Gr: 0.9839
  • Loss En Us: 0.0827
  • Loss Es 419: 0.0516
  • Loss Et Ee: 1.9264
  • Loss Fa Ir: 0.6520
  • Loss Ff Sn: 5.4283
  • Loss Fi Fi: 0.0109
  • Loss Fil Ph: 0.1706
  • Loss Fr Fr: 0.0591
  • Loss Ga Ie: 0.5174
  • Loss Gl Es: 1.2657
  • Loss Gu In: 0.0850
  • Loss Ha Ng: 0.3234
  • Loss He Il: 0.8299
  • Loss Hi In: 0.4190
  • Loss Hr Hr: 2.9754
  • Loss Hu Hu: 0.8345
  • Loss Hy Am: 0.0329
  • Loss Id Id: 0.0529
  • Loss Ig Ng: 0.2523
  • Loss Is Is: 6.5153
  • Loss It It: 0.8113
  • Loss Ja Jp: 1.3968
  • Loss Jv Id: 2.0009
  • Loss Ka Ge: 0.6162
  • Loss Kam Ke: 2.2192
  • Loss Kea Cv: 0.5567
  • Loss Kk Kz: 0.5592
  • Loss Km Kh: 1.7358
  • Loss Kn In: 0.1063
  • Loss Ko Kr: 0.1519
  • Loss Ku Arab Iq: 0.2075
  • Loss Ky Kg: 0.4639
  • Loss Lb Lu: 0.4454
  • Loss Lg Ug: 0.3764
  • Loss Ln Cd: 0.1844
  • Loss Lo La: 3.8051
  • Loss Lt Lt: 2.5054
  • Loss Luo Ke: 0.0479
  • Loss Lv Lv: 1.3713
  • Loss Mi Nz: 0.1390
  • Loss Mk Mk: 0.7952
  • Loss Ml In: 1.2999
  • Loss Mn Mn: 0.7621
  • Loss Mr In: 3.7056
  • Loss Ms My: 3.0192
  • Loss Mt Mt: 1.5520
  • Loss My Mm: 0.1514
  • Loss Nb No: 1.1194
  • Loss Ne Np: 0.4231
  • Loss Nl Nl: 0.3291
  • Loss Nso Za: 0.5106
  • Loss Ny Mw: 2.7346
  • Loss Oci Fr: 5.0983
  • Loss Om Et: 0.2297
  • Loss Or In: 2.5432
  • Loss Pa In: 0.7753
  • Loss Pl Pl: 0.7309
  • Loss Ps Af: 1.0454
  • Loss Pt Br: 0.9782
  • Loss Ro Ro: 3.5829
  • Loss Ru Ru: 0.0598
  • Loss Rup Bg: 0.1695
  • Loss Sd Arab In: 2.6198
  • Loss Sk Sk: 0.5583
  • Loss Sl Si: 6.0923
  • Loss Sn Zw: 0.4465
  • Loss So So: 0.4492
  • Loss Sr Rs: 4.7575
  • Loss Sv Se: 6.5858
  • Loss Sw Ke: 0.4235
  • Loss Ta In: 0.1818
  • Loss Te In: 0.0808
  • Loss Tg Tj: 0.0912
  • Loss Th Th: 0.0462
  • Loss Tr Tr: 0.7340
  • Loss Uk Ua: 4.6777
  • Loss Umb Ao: 1.4021
  • Loss Ur Pk: 8.4067
  • Loss Uz Uz: 4.3297
  • Loss Vi Vn: 1.1304
  • Loss Wo Sn: 2.2281
  • Loss Xh Za: 0.0009
  • Loss Yo Ng: 1.3345
  • Loss Yue Hant Hk: 1.0728
  • Loss Zu Za: 3.7279
  • Predict Samples: 77960

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
  • train_batch_size: 8
  • eval_batch_size: 1
  • seed: 42
  • distributed_type: multi-GPU
  • num_devices: 8
  • total_train_batch_size: 64
  • total_eval_batch_size: 8
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • lr_scheduler_warmup_steps: 2000
  • num_epochs: 5.0
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Accuracy Validation Loss
0.5296 0.26 1000 0.4016 2.6633
0.4252 0.52 2000 0.5751 1.8582
0.2989 0.78 3000 0.6332 1.6780
0.3563 1.04 4000 0.6799 1.4479
0.1617 1.3 5000 0.6679 1.5066
0.1409 1.56 6000 0.6992 1.4082
0.01 1.82 7000 0.7071 1.2448
0.0018 2.08 8000 0.7148 1.1996
0.0014 2.34 9000 0.6410 1.6505
0.0188 2.6 10000 0.6840 1.4050
0.0007 2.86 11000 0.6621 1.5831
0.1038 3.12 12000 0.6829 1.5441
0.0003 3.38 13000 0.6900 1.3483
0.0004 3.64 14000 0.6414 1.7070
0.0003 3.9 15000 0.7075 1.3198
0.0002 4.16 16000 0.7105 1.3118
0.0001 4.42 17000 0.7029 1.4099
0.0 4.68 18000 0.7180 1.3658
0.0001 4.93 19000 0.7236 1.3514

Framework versions

  • Transformers 4.18.0.dev0
  • Pytorch 1.10.1+cu111
  • Datasets 1.18.4.dev0
  • Tokenizers 0.11.6