metadata

language:
  - bem
license: apache-2.0
tags:
  - hf-asr-leaderboard
  - generated_from_trainer
datasets:
  - BembaSpeech
metrics:
  - wer
model-index:
  - name: Whisper Small Bem - Claytone Sikasote
    results:
      - task:
          name: Automatic Speech Recognition
          type: automatic-speech-recognition
        dataset:
          name: BembaSpeech
          type: BembaSpeech
          config: null
          split: None
          args: 'config: Bem, split: test'
        metrics:
          - name: Wer
            type: wer
            value: 39.667599423386754

Whisper Small Bem - Claytone Sikasote

This model is a fine-tuned version of openai/whisper-small on the BembaSpeech dataset. It achieves the following results on the evaluation set:

Loss: 0.4251
Wer: 39.6676

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: 1e-05
train_batch_size: 16
eval_batch_size: 8
seed: 42
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
lr_scheduler_warmup_steps: 500
training_steps: 4000
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.4394	1.34	1000	0.4104	45.0013
0.2845	2.68	2000	0.3709	39.4217
0.1073	4.03	3000	0.3963	38.4720
0.0565	5.37	4000	0.4251	39.6676

Framework versions

Transformers 4.26.0.dev0
Pytorch 1.12.1+cu113
Datasets 2.7.1
Tokenizers 0.13.2