whisper-base.en-speech-commands-h

This model is a fine-tuned version of openai/whisper-base.en on the speech_commands dataset. It achieves the following results on the evaluation set:

• Loss: 1.3313
• Accuracy: 0.7923

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: 0.0005
• train_batch_size: 96
• eval_batch_size: 96
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 20
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
0.3859	1.0	412	1.3474	0.7707
0.2732	2.0	824	1.2471	0.7599
0.2373	3.0	1236	1.2114	0.7729
0.1694	4.0	1648	1.1600	0.7914
0.1495	5.0	2060	1.1535	0.7914
0.1931	6.0	2472	1.1446	0.7860
0.1329	7.0	2884	1.3313	0.7923
0.0731	8.0	3296	1.2812	0.7860
0.0702	9.0	3708	1.2134	0.7873
0.0828	10.0	4120	1.6292	0.7887
0.08	11.0	4532	1.4677	0.7797
0.0481	12.0	4944	1.3770	0.7909

Framework versions

• Transformers 4.43.3
• Pytorch 2.2.2+cu121
• Datasets 2.18.0
• Tokenizers 0.19.1