README.md

metadata

language:
  - ne
license: apache-2.0
tags:
  - whisper-event
  - generated_from_trainer
datasets:
  - mozilla-foundation/common_voice_11_0
metrics:
  - wer
model-index:
  - name: Whisper Large-V2 Nepali
    results:
      - task:
          name: Automatic Speech Recognition
          type: automatic-speech-recognition
        dataset:
          name: mozilla-foundation/common_voice_11_0 ne-NP
          type: mozilla-foundation/common_voice_11_0
          config: ne-NP
          split: test
          args: ne-NP
        metrics:
          - name: Wer
            type: wer
            value: 9.75609756097561

Whisper Large-V2 Nepali

This model is a fine-tuned version of DrishtiSharma/whisper-large-v2-hindi-3k-steps on the Common Voice 11.0 dataset. It achieves the following results on the evaluation set:

• Loss: 0.8851
• Wer: 9.7561

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: 8
• 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: 100
• training_steps: 200
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.0	200.0	200	0.8851	9.7561

Framework versions

• Transformers 4.26.0.dev0
• Pytorch 1.13.0+cu116
• Datasets 2.8.1.dev0
• Tokenizers 0.13.2