tomtom5/whisper-small-he

Whisper Small He - Tom Apt

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

• Loss: 0.5687
• Wer Ortho: 41.6562
• Wer: 40.9912

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: 16
• 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: 500
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer Ortho	Wer
0.8098	0.1121	25	0.8606	45.0808	44.3578
0.705	0.2242	50	0.8148	45.3982	44.7695
0.7774	0.3363	75	0.7506	44.4756	43.8231
0.711	0.4484	100	0.6873	44.7265	44.0616
0.5994	0.5605	125	0.6026	44.1287	44.0182
0.538	0.6726	150	0.5826	43.7080	43.3247
0.5181	0.7848	175	0.5767	42.6895	42.5300
0.5217	0.8969	200	0.5695	43.1692	42.3566
0.4944	1.0090	225	0.5614	41.3905	41.0707
0.3309	1.1211	250	0.5733	44.2394	43.7942
0.3116	1.2332	275	0.5608	42.3574	41.6269
0.3164	1.3453	300	0.5646	44.0254	43.2596
0.3101	1.4574	325	0.5589	42.2762	41.7714
0.3246	1.5695	350	0.5572	42.4386	42.2049
0.3127	1.6816	375	0.5610	41.5898	41.1068
0.2999	1.7937	400	0.5602	40.9329	40.8611
0.308	1.9058	425	0.5531	42.1064	41.8726
0.2709	2.0179	450	0.5553	41.4127	40.7311
0.171	2.1300	475	0.5640	44.5863	44.4444
0.1592	2.2422	500	0.5687	41.6562	40.9912

Framework versions

• Transformers 4.48.3
• Pytorch 2.5.1+cu124
• Datasets 3.3.2
• Tokenizers 0.21.0