README.md

metadata

language:
  - el
license: apache-2.0
base_model: openai/whisper-base
tags:
  - generated_from_trainer
datasets:
  - AMoustakis/test-dataset
metrics:
  - wer
model-index:
  - name: Whisper Base Greek
    results:
      - task:
          name: Automatic Speech Recognition
          type: automatic-speech-recognition
        dataset:
          name: Test Dataset for greek language
          type: AMoustakis/test-dataset
          args: 'split: train'
        metrics:
          - name: Wer
            type: wer
            value: 61.77777777777778

Whisper Base Greek

This model is a fine-tuned version of openai/whisper-base on the Test Dataset for greek language dataset. It achieves the following results on the evaluation set:

• Loss: 0.2675
• Wer: 61.7778

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: 0.001
• train_batch_size: 3
• eval_batch_size: 1
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 5
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.9014	1.0	4	0.5733	74.5185
0.4829	2.0	8	0.4158	64.4444
0.5963	3.0	12	0.3257	65.0370
0.3399	4.0	16	0.2857	61.7778
0.4436	5.0	20	0.2675	61.7778

Framework versions

• Transformers 4.40.2
• Pytorch 2.2.1+cu121
• Datasets 2.19.1
• Tokenizers 0.19.1