Whisper Small Hi - Sanchit Gandhi

This model is a fine-tuned version of openai/whisper-small on the Common Voice 13.0 dataset in Flax. It is trained using the Transformers Flax examples script, and achieves the following results on the evaluation set:

• Loss: 0.02091
• Wer: 33.96828

The training run can be reproduced in approximately 25 minutes by executing the script run.sh.

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 1e-04
• train_batch_size: 64
• eval_batch_size: 64
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• num_train_epochs: 10

Training results

See Tensorboard logs for details.