Whisper Small

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

  • Loss: 1.0151
  • Wer: 37.2345

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: 16
  • seed: 42
  • gradient_accumulation_steps: 2
  • 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: 500
  • num_epochs: 40
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss Wer
0.1977 8.1301 500 0.7429 41.9683
0.0059 16.2602 1000 0.9167 38.4064
0.001 24.3902 1500 0.9849 37.3501
0.0007 32.5203 2000 1.0151 37.2345

Framework versions

  • Transformers 4.44.2
  • Pytorch 2.1.0+cu118
  • Datasets 2.21.0
  • Tokenizers 0.19.1
Downloads last month
21
Safetensors
Model size
242M params
Tensor type
F32
·
Inference Examples
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social visibility and check back later, or deploy to Inference Endpoints (dedicated) instead.

Model tree for KasuleTrevor/whisper-lingala-small-test-20

Finetuned
(2095)
this model

Dataset used to train KasuleTrevor/whisper-lingala-small-test-20

Evaluation results