JerryLearnCode/4320-whisper-small-eng

Whisper Small Finetune - IERG4320 Project

This model is a fine-tuned version of openai/whisper-small on the Common Voice 13 dataset. It achieves the following results on the evaluation set:

• Loss: 0.5674
• Wer Ortho: 22.0544
• Wer: 18.2934

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: 64
• eval_batch_size: 64
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: constant_with_warmup
• lr_scheduler_warmup_steps: 50
• training_steps: 1000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer Ortho	Wer
0.094	2.5974	200	0.4300	20.8774	17.1318
0.0105	5.1948	400	0.5000	21.6635	17.7999
0.0024	7.7922	600	0.5250	21.7294	17.9615
0.0015	10.3896	800	0.5528	23.3630	19.5205
0.0011	12.9870	1000	0.5674	22.0544	18.2934

Framework versions

• Transformers 4.46.3
• Pytorch 2.5.1
• Datasets 3.1.0
• Tokenizers 0.20.4