README.md

metadata

library_name: transformers
language:
  - en
license: cc-by-sa-4.0
base_model: openai/whisper-large
tags:
  - generated_from_trainer
datasets:
  - sage-bergerson/edacc_processed
metrics:
  - wer
model-index:
  - name: Whisper Large EdAcc V2
    results:
      - task:
          name: Automatic Speech Recognition
          type: automatic-speech-recognition
        dataset:
          name: EdAcc
          type: sage-bergerson/edacc_processed
          args: 'config: en, split: train'
        metrics:
          - name: Wer
            type: wer
            value: 0.5855270257403117

Whisper Large EdAcc V2

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

• Loss: 0.6378
• Wer: 0.5855

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: 5e-06
• train_batch_size: 32
• eval_batch_size: 16
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• training_steps: 1000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
1.1515	0.3247	100	0.7869	0.3055
0.6272	0.6494	200	0.6171	0.4607
0.5614	0.9740	300	0.5925	0.6110
0.43	1.2987	400	0.5868	0.5105
0.4576	1.6234	500	0.5844	0.6095
0.4727	1.9481	600	0.5784	0.6796
0.3274	2.2727	700	0.6094	0.5416
0.2862	2.5974	800	0.6027	0.5609
0.2908	2.9221	900	0.6107	0.4607
0.2221	3.2468	1000	0.6378	0.5855

Framework versions

• Transformers 4.44.2
• Pytorch 2.4.0
• Datasets 2.21.0
• Tokenizers 0.19.1