nurzhanit/whisper-omg-2

Whisper Small Hi - Sanchit Gandhi

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

• Loss: 0.0001
• Wer: 0.0

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: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 50
• training_steps: 1000

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.0417	0.8475	50	0.0308	2.8184
0.0233	1.6949	100	0.0178	2.0473
0.0153	2.5424	150	0.0104	1.5953
0.0121	3.3898	200	0.0056	1.3560
0.0055	4.2373	250	0.0039	0.8508
0.0061	5.0847	300	0.0022	0.3457
0.0026	5.9322	350	0.0012	0.2127
0.0018	6.7797	400	0.0008	0.0798
0.0015	7.6271	450	0.0006	0.1595
0.0009	8.4746	500	0.0005	0.0266
0.0004	9.3220	550	0.0003	0.0266
0.0003	10.1695	600	0.0002	0.0
0.0002	11.0169	650	0.0003	0.0266
0.0002	11.8644	700	0.0002	0.0266
0.0002	12.7119	750	0.0002	0.0
0.0002	13.5593	800	0.0002	0.0
0.0002	14.4068	850	0.0002	0.0
0.0002	15.2542	900	0.0002	0.0
0.0001	16.1017	950	0.0001	0.0
0.0002	16.9492	1000	0.0001	0.0

Framework versions

• Transformers 4.40.0
• Pytorch 2.5.0+cu124
• Datasets 3.0.2
• Tokenizers 0.19.1