datdo2717/whisper-small-20-11

Whisper Small Ori vi

This model is a fine-tuned version of openai/whisper-small on the Common Voice 11.0 dataset. It achieves the following results on the evaluation set:

• Loss: 0.3950
• Wer: 17.6577

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: 16
• eval_batch_size: 8
• seed: 42
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 200
• training_steps: 1300
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
1.1125	0.0667	30	0.9877	30.4667
0.9337	0.1333	60	0.7582	18.3338
0.7185	0.2	90	0.4716	16.3129
0.493	0.2667	120	0.4382	16.1893
0.4328	0.3333	150	0.4298	15.6223
0.4127	0.4	180	0.4208	16.8726
0.3865	0.4667	210	0.4171	20.0422
0.419	0.5333	240	0.4141	17.0835
0.4141	0.6	270	0.4157	15.8258
0.464	0.6667	300	0.4077	16.9235
0.4303	0.7333	330	0.4043	18.4865
0.4418	0.8	360	0.4050	16.7999
0.4786	0.8667	390	0.3981	15.1352
0.4238	0.9333	420	0.3953	17.0907
0.3986	1.0	450	0.3926	16.7054
0.2304	1.0667	480	0.3948	16.3928
0.2583	1.1333	510	0.3943	16.6327
0.2385	1.2	540	0.3997	15.1425
0.2126	1.2667	570	0.3985	15.0552
0.2259	1.3333	600	0.3970	16.5964
0.2237	1.4	630	0.3964	16.5382
0.2344	1.4667	660	0.3983	17.9485
0.2068	1.5333	690	0.3974	17.9703
0.2535	1.6	720	0.3950	17.6577

Framework versions

• Transformers 4.46.3
• Pytorch 2.4.0
• Datasets 3.1.0
• Tokenizers 0.20.0