metadata

license: apache-2.0
tags:
  - automatic-speech-recognition
  - google/fleurs
  - generated_from_trainer
  - hf-asr-leaderboard
  - ps
  - Pashto
datasets:
  - fleurs
metrics:
  - wer
model-index:
  - name: facebook/wav2vec2-xls-r-300m
    results:
      - task:
          name: Automatic Speech Recognition
          type: automatic-speech-recognition
        dataset:
          name: GOOGLE/FLEURS - PS_AF
          type: fleurs
          config: ps_af
          split: test
          args: 'Config: ps_af, Training split: train+validation, Eval split: test'
        metrics:
          - name: Wer
            type: wer
            value: 0.5137278308321964

facebook/wav2vec2-xls-r-300m

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

Loss: 0.9154
Wer: 0.5137
Cer: 0.1966

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: 7.5e-05
train_batch_size: 8
eval_batch_size: 8
seed: 42
gradient_accumulation_steps: 4
total_train_batch_size: 32
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
lr_scheduler_warmup_steps: 2000
num_epochs: 50.0
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
5.0767	6.33	500	4.8783	1.0	1.0
3.1156	12.66	1000	3.0990	1.0	1.0
1.3506	18.99	1500	1.1056	0.7031	0.2889
0.9997	25.32	2000	0.9191	0.5944	0.2301
0.7838	31.65	2500	0.8952	0.5556	0.2152
0.6665	37.97	3000	0.8908	0.5252	0.2017
0.6265	44.3	3500	0.9063	0.5133	0.1954

Framework versions

Transformers 4.26.0.dev0
Pytorch 1.13.0+cu117
Datasets 2.7.1.dev0
Tokenizers 0.13.2