sarpba/whisper-base-cv16-hu-v5

Whisper Tiny Hu v5

This model is a fine-tuned version of openai/whisper-tiny on the Common Voice 16.0 dataset. It achieves the following results on the evaluation set:

• Loss: 0.1835
• Wer Ortho: 14.8079
• Wer: 13.5339

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: 3.75e-05
• train_batch_size: 32
• eval_batch_size: 16
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: constant_with_warmup
• lr_scheduler_warmup_steps: 500
• training_steps: 10000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer Ortho	Wer
0.4291	0.67	1000	0.4821	47.5702	44.3878
0.271	1.34	2000	0.3431	35.7913	33.0685
0.2015	2.01	3000	0.2665	28.8089	26.0777
0.1559	2.68	4000	0.2355	24.7712	22.3006
0.0934	3.35	5000	0.2089	21.6879	19.7658
0.0542	4.02	6000	0.1921	18.6950	16.7003
0.061	4.69	7000	0.1895	17.2558	15.6122
0.0356	5.35	8000	0.1866	16.5302	14.9867
0.0225	6.02	9000	0.1815	15.8708	14.4115
0.0318	6.69	10000	0.1835	14.8079	13.5339

Framework versions

• Transformers 4.37.2
• Pytorch 2.1.0+cu121
• Datasets 2.16.1
• Tokenizers 0.15.0