VOOZH about

URL: https://huggingface.co/abasseyfresh/whisper-large-v3-igbo

⇱ abasseyfresh/whisper-large-v3-igbo · Hugging Face

whisper-large-v3-igbo

This model is a fine-tuned version of openai/whisper-large-v3 on an unknown dataset. It achieves the following results on the evaluation set:

  • Loss: 0.6021
  • Wer: 51.26

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: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
  • lr_scheduler_type: cosine
  • lr_scheduler_warmup_steps: 500
  • num_epochs: 15

Training results

Training Loss Epoch Step Validation Loss Wer
2.0089 1.0661 500 0.9678 77.77
1.3785 2.1322 1000 0.7134 60.19
1.1897 3.1983 1500 0.6503 55.73
1.0523 4.2644 2000 0.6214 52.91
0.9796 5.3305 2500 0.6098 52.23
0.9334 6.3966 3000 0.6042 52.11
0.9120 7.4627 3500 0.6021 51.26
0.8714 8.5288 4000 0.6032 51.65
0.8599 9.5949 4500 0.6021 51.29
0.8597 10.6610 5000 0.6031 51.43
0.8562 11.7271 5500 0.6034 51.78
0.8612 12.7932 6000 0.6037 51.69
0.8595 13.8593 6500 0.6038 51.57
0.8415 14.9254 7000 0.6038 51.48

Framework versions

  • Transformers 5.0.0
  • Pytorch 2.11.0+cu128
  • Datasets 4.0.0
  • Tokenizers 0.22.2
Downloads last month
346
Safetensors
Model size
2B params
Tensor type
BF16
·

Model tree for abasseyfresh/whisper-large-v3-igbo

Finetuned
(867)
this model

Dataset used to train abasseyfresh/whisper-large-v3-igbo