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: 0.0073
• Wer: 4.1441

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: 100
• training_steps: 1000

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.3134	3.125	50	0.1611	13.9040
0.0807	6.25	100	0.0474	7.5329
0.0419	9.375	150	0.0497	8.4043
0.03	12.5	200	0.0239	5.5577
0.0217	15.625	250	0.0185	6.6421
0.0198	18.75	300	0.0152	4.5895
0.0137	21.875	350	0.0109	4.2796
0.011	25.0	400	0.0092	4.9380
0.0107	28.125	450	0.0087	4.8412
0.01	31.25	500	0.0082	4.2796
0.0089	34.375	550	0.0082	4.3184
0.0087	37.5	600	0.0083	4.9961
0.0083	40.625	650	0.0077	4.8025
0.0085	43.75	700	0.0076	4.3571
0.0082	46.875	750	0.0077	4.3765
0.0085	50.0	800	0.0075	4.2603
0.0082	53.125	850	0.0074	4.3184
0.0077	56.25	900	0.0073	4.2409
0.0067	59.375	950	0.0073	4.3184
0.0074	62.5	1000	0.0073	4.1441

Framework versions

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