metadata

library_name: transformers
license: apache-2.0
base_model: facebook/wav2vec2-xls-r-300m
tags:
  - generated_from_trainer
datasets:
  - fleurs
metrics:
  - wer
model-index:
  - name: wav2vec2-xls-r-300m-lg-CV-Fleurs_filtered-100hrs-v12
    results:
      - task:
          name: Automatic Speech Recognition
          type: automatic-speech-recognition
        dataset:
          name: fleurs
          type: fleurs
          config: lg_ug
          split: None
          args: lg_ug
        metrics:
          - name: Wer
            type: wer
            value: 0.45568513119533527

wav2vec2-xls-r-300m-lg-CV-Fleurs_filtered-100hrs-v12

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

Loss: 0.5693
Wer: 0.4557
Cer: 0.0926

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: 4
eval_batch_size: 2
seed: 42
gradient_accumulation_steps: 2
total_train_batch_size: 8
optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
lr_scheduler_type: linear
num_epochs: 70
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
0.8586	1.0	7125	0.4810	0.5615	0.1256
0.423	2.0	14250	0.4661	0.5633	0.1450
0.3588	3.0	21375	0.4151	0.5	0.1075
0.32	4.0	28500	0.3937	0.4999	0.1084
0.291	5.0	35625	0.3880	0.4866	0.1026
0.2685	6.0	42750	0.3779	0.4860	0.1035
0.2498	7.0	49875	0.3598	0.4660	0.0973
0.2347	8.0	57000	0.3553	0.4573	0.0936
0.2198	9.0	64125	0.3584	0.4630	0.0950
0.2072	10.0	71250	0.3571	0.4742	0.0983
0.1954	11.0	78375	0.3596	0.4580	0.0955
0.1861	12.0	85500	0.3626	0.4573	0.0952
0.1746	13.0	92625	0.3656	0.4972	0.1014
0.1656	14.0	99750	0.3712	0.4566	0.0918
0.1572	15.0	106875	0.3874	0.4569	0.0933
0.149	16.0	114000	0.3919	0.4809	0.0966
0.142	17.0	121125	0.3837	0.4424	0.0907
0.1332	18.0	128250	0.3843	0.4635	0.0941
0.1271	19.0	135375	0.4080	0.4560	0.0942
0.1203	20.0	142500	0.4209	0.4673	0.0929
0.1136	21.0	149625	0.4188	0.4632	0.0934
0.1084	22.0	156750	0.4369	0.4588	0.0930
0.1029	23.0	163875	0.4553	0.4735	0.0944
0.0993	24.0	171000	0.4547	0.4654	0.0941
0.0943	25.0	178125	0.4775	0.4561	0.0925
0.0902	26.0	185250	0.5074	0.4649	0.0935
0.0867	27.0	192375	0.5073	0.4509	0.0912
0.0833	28.0	199500	0.5150	0.4749	0.0953
0.0799	29.0	206625	0.5624	0.4725	0.0944
0.0771	30.0	213750	0.5769	0.4552	0.0918
0.0739	31.0	220875	0.5697	0.4533	0.0917
0.0704	32.0	228000	0.5693	0.4557	0.0926

Framework versions

Transformers 4.46.2
Pytorch 2.1.0+cu118
Datasets 3.1.0
Tokenizers 0.20.3