whisper-small-se

This model is a fine-tuned version of openai/whisper-small on the Common Voice 11.0 dataset.

Model description

The model was initially trained on 680 000 hours of audio with corresponding transcripts from the internet, 65% of which was in english audio and 83 % of which had english transcripts.

The model was then further trained for 4000 iterations, 500 of which as warm-up, on Swedish data from Common_voice 11.0. Achieving a WER of 19.865.

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: 16
• 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: 500
• training_steps: 500
• mixed_precision_training: Native AMP

Training results

Model Plot

View Training Plots

Framework versions

• Transformers 4.26.0.dev0
• Pytorch 1.12.1+cu113
• Datasets 2.7.1
• Tokenizers 0.13.2