Edit model card

Whisper Small Sr Yodas

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

  • Loss: 0.2688
  • Wer Ortho: 0.3334
  • Wer: 0.2450

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: 50
  • training_steps: 4000
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss Wer Ortho Wer
1.0469 0.24 500 0.4020 0.5071 0.4270
0.9924 0.49 1000 0.3401 0.4082 0.3183
0.865 0.73 1500 0.3047 0.3644 0.2776
0.8443 0.98 2000 0.2893 0.3623 0.2735
0.7377 1.22 2500 0.2817 0.3472 0.2591
0.6851 1.46 3000 0.2728 0.3348 0.2466
0.7286 1.71 3500 0.2702 0.3325 0.2444
0.7215 1.95 4000 0.2688 0.3334 0.2450

Framework versions

  • Transformers 4.39.3
  • Pytorch 2.0.1+cu117
  • Datasets 2.18.0
  • Tokenizers 0.15.1
Downloads last month
2
Safetensors
Model size
72.6M params
Tensor type
F32
·
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.

Finetuned from

Dataset used to train Sagicc/whisper-base-sr-yodas

Evaluation results