Edit model card

Incremental Swahili Luganda

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

  • Loss: 0.3450
  • Wer: 30.7579

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: 8
  • eval_batch_size: 8
  • seed: 42
  • gradient_accumulation_steps: 2
  • total_train_batch_size: 16
  • 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
0.1454 0.1129 500 0.3666 32.6860
0.1537 0.2258 1000 0.3721 32.9290
0.1471 0.3388 1500 0.3665 32.9660
0.1397 0.4517 2000 0.3626 32.0067
0.1501 0.5646 2500 0.3562 32.2413
0.1381 0.6775 3000 0.3510 30.8636
0.14 0.7904 3500 0.3476 30.9122
0.135 0.9033 4000 0.3450 30.7579

Framework versions

  • Transformers 4.40.0
  • Pytorch 2.2.2+cu118
  • Datasets 2.19.0
  • Tokenizers 0.19.1
Downloads last month
1
Safetensors
Model size
242M params
Tensor type
F32
·

Finetuned from

Dataset used to train mn720/inctraining5

Evaluation results