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: 0.0000
• 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: 200
• training_steps: 4000

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.0073	50.0	100	0.0008	0.0
0.0003	100.0	200	0.0002	0.0
0.0001	150.0	300	0.0001	0.0
0.0001	200.0	400	0.0001	0.0
0.0001	250.0	500	0.0001	0.0
0.0	300.0	600	0.0000	0.0
0.0	350.0	700	0.0000	0.0
0.0	400.0	800	0.0000	0.0
0.0	450.0	900	0.0000	0.0
0.0	500.0	1000	0.0000	0.0
0.0	550.0	1100	0.0000	0.0
0.0	600.0	1200	0.0000	0.0
0.0	650.0	1300	0.0000	0.0
0.0	700.0	1400	0.0000	0.0
0.0	750.0	1500	0.0000	0.0
0.0	800.0	1600	0.0000	0.0
0.0	850.0	1700	0.0000	0.0
0.0	900.0	1800	0.0000	0.0
0.0	950.0	1900	0.0000	0.0
0.0	1000.0	2000	0.0000	0.0
0.0	1050.0	2100	0.0000	0.0
0.0	1100.0	2200	0.0000	0.0
0.0	1150.0	2300	0.0000	0.0
0.0	1200.0	2400	0.0000	0.0
0.0	1250.0	2500	0.0000	0.0
0.0	1300.0	2600	0.0000	0.0
0.0	1350.0	2700	0.0000	0.0
0.0	1400.0	2800	0.0000	0.0
0.0	1450.0	2900	0.0000	0.0
0.0	1500.0	3000	0.0000	0.0
0.0	1550.0	3100	0.0000	0.0
0.0	1600.0	3200	0.0000	0.0
0.0	1650.0	3300	0.0000	0.0
0.0	1700.0	3400	0.0000	0.0
0.0	1750.0	3500	0.0000	0.0
0.0	1800.0	3600	0.0000	0.0
0.0	1850.0	3700	0.0000	0.0
0.0	1900.0	3800	0.0000	0.0
0.0	1950.0	3900	0.0000	0.0
0.0	2000.0	4000	0.0000	0.0

Framework versions

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