README.md

metadata

license: apache-2.0
base_model: openai/whisper-large-v3
tags:
  - generated_from_trainer
datasets:
  - common_voice_18_0
metrics:
  - wer
model-index:
  - name: whisper-large-v3-pt-3000h-3
    results:
      - task:
          name: Automatic Speech Recognition
          type: automatic-speech-recognition
        dataset:
          name: common_voice_18_0
          type: common_voice_18_0
          config: pt
          split: None
          args: pt
        metrics:
          - name: Wer
            type: wer
            value: 0.10174567584881486

whisper-large-v3-pt-3000h-3

This model is a fine-tuned version of openai/whisper-large-v3 on the common_voice_18_0 dataset. It achieves the following results on the evaluation set:

• Loss: 0.1478
• Wer: 0.1017

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: 1
• eval_batch_size: 1
• seed: 42
• gradient_accumulation_steps: 32
• total_train_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 1000
• num_epochs: 2.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.13	0.9998	691	0.1486	0.1037
0.0844	1.9998	1382	0.1478	0.1017

Framework versions

• Transformers 4.44.0.dev0
• Pytorch 2.4.0+cu124
• Datasets 2.18.1.dev0
• Tokenizers 0.19.1