metadata

library_name: transformers
license: cc-by-nc-4.0
base_model: facebook/mms-1b-all
tags:
  - generated_from_trainer
datasets:
  - fleurs
metrics:
  - wer
model-index:
  - name: mms-1b-all-lg-CV-Fleurs_filtered-100hrs-v1
    results:
      - task:
          name: Automatic Speech Recognition
          type: automatic-speech-recognition
        dataset:
          name: fleurs
          type: fleurs
          config: lg_ug
          split: test
          args: lg_ug
        metrics:
          - name: Wer
            type: wer
            value: 0.4098153547133139

mms-1b-all-lg-CV-Fleurs_filtered-100hrs-v1

This model is a fine-tuned version of facebook/mms-1b-all on the fleurs dataset. It achieves the following results on the evaluation set:

Loss: 0.2897
Wer: 0.4098
Cer: 0.0743

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: 0.001
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
0.3203	1.0	7125	0.3178	0.4156	0.0762
0.2149	2.0	14250	0.3008	0.4194	0.0759
0.2093	3.0	21375	0.3015	0.4017	0.0743
0.2064	4.0	28500	0.3043	0.4114	0.0745
0.2042	5.0	35625	0.2955	0.4069	0.0753
0.2022	6.0	42750	0.3009	0.4088	0.0750
0.1989	7.0	49875	0.3088	0.4092	0.0756
0.1983	8.0	57000	0.2980	0.4081	0.0754
0.1969	9.0	64125	0.2951	0.4040	0.0741
0.1957	10.0	71250	0.2899	0.4039	0.0745
0.1945	11.0	78375	0.2896	0.4083	0.0744
0.1936	12.0	85500	0.2931	0.4038	0.0743
0.1929	13.0	92625	0.2897	0.4098	0.0743

Framework versions

Transformers 4.46.2
Pytorch 2.1.0+cu118
Datasets 3.1.0
Tokenizers 0.20.3