README.md

metadata

license: apache-2.0
tags:
  - generated_from_trainer
metrics:
  - wer
model-index:
  - name: openai/whisper-medium
    results:
      - task:
          type: automatic-speech-recognition
          name: Automatic Speech Recognition
        dataset:
          name: rishabhjain16/infer_cmu_9h
          type: rishabhjain16/infer_cmu_9h
          config: en
          split: test
        metrics:
          - type: wer
            value: 16.57
            name: WER
      - task:
          type: automatic-speech-recognition
          name: Automatic Speech Recognition
        dataset:
          name: rishabhjain16/infer_pfs
          type: rishabhjain16/infer_pfs
          config: en
          split: test
        metrics:
          - type: wer
            value: 3.15
            name: WER
      - task:
          type: automatic-speech-recognition
          name: Automatic Speech Recognition
        dataset:
          name: rishabhjain16/infer_myst
          type: rishabhjain16/infer_myst
          config: en
          split: test
        metrics:
          - type: wer
            value: 16.18
            name: WER
      - task:
          type: automatic-speech-recognition
          name: Automatic Speech Recognition
        dataset:
          name: rishabhjain16/libritts_dev_clean
          type: rishabhjain16/libritts_dev_clean
          config: en
          split: test
        metrics:
          - type: wer
            value: 5.33
            name: WER

openai/whisper-medium

This model is a fine-tuned version of openai/whisper-medium on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.1594
• Wer: 21.8343

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: 32
• eval_batch_size: 32
• 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: 4000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.0269	5.0	500	0.1069	118.0302
0.0049	10.01	1000	0.1263	135.2788
0.0009	15.01	1500	0.1355	94.5731
0.0001	20.01	2000	0.1413	7.5188
0.0001	25.01	2500	0.1515	7.2508
0.0001	30.02	3000	0.1568	24.8493
0.0	35.02	3500	0.1588	22.1470
0.0	40.02	4000	0.1594	21.8343

Framework versions

• Transformers 4.27.0.dev0
• Pytorch 1.13.1+cu117
• Datasets 2.9.1.dev0
• Tokenizers 0.13.2