sarpba/whisper-small-cv18-hu-cleaned

Whisper Small Hu CV18

This model is a fine-tuned version of openai/whisper-small on the Common Voice 18.0 dataset. It achieves the following results on the evaluation set:

• Loss: 0.6647
• Wer Ortho: 32.6950
• Wer: 25.9818

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: 2.5e-05
• train_batch_size: 64
• eval_batch_size: 32
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• num_epochs: 5
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer Ortho	Wer
0.2902	0.1723	250	0.6498	43.8115	38.8390
0.235	0.3446	500	0.6526	43.4572	37.5766
0.1898	0.5169	750	0.6127	40.0623	34.8072
0.177	0.6892	1000	0.5866	39.0613	33.7390
0.1461	0.8615	1250	0.5848	37.2397	31.7552
0.0834	1.0338	1500	0.5942	36.6963	30.8611
0.0783	1.2061	1750	0.5973	36.2904	29.8850
0.0757	1.3784	2000	0.6127	36.8602	30.0451
0.0737	1.5507	2250	0.5925	35.8202	29.6189
0.0724	1.7229	2500	0.5721	34.6654	29.0627
0.0707	1.8952	2750	0.5818	34.6326	28.4650
0.0292	2.0675	3000	0.5917	34.4423	28.0841
0.0288	2.2398	3250	0.6147	34.3465	28.0210
0.0283	2.4121	3500	0.6279	34.7310	28.1572
0.0319	2.5844	3750	0.6122	33.9229	27.3514
0.0292	2.7567	4000	0.5988	33.6947	27.7600
0.0262	2.9290	4250	0.6170	33.8876	27.3716
0.0093	3.1013	4500	0.6297	32.9862	26.4131
0.0094	3.2736	4750	0.6167	32.2336	26.3790
0.0086	3.4459	5000	0.6430	32.9068	26.3904
0.0094	3.6182	5250	0.6432	32.9749	26.4358
0.0088	3.7905	5500	0.6330	32.8438	26.3551
0.0082	3.9628	5750	0.6530	33.4325	26.6212
0.0035	4.1351	6000	0.6549	32.8589	26.1924
0.003	4.3074	6250	0.6625	32.7757	25.9175
0.003	4.4797	6500	0.6684	32.6950	25.8393
0.0032	4.6520	6750	0.6622	32.3912	25.7687
0.0027	4.8243	7000	0.6667	32.6585	25.8847
0.0028	4.9966	7250	0.6647	32.6950	25.9818

Framework versions

• Transformers 4.44.0
• Pytorch 2.3.0+cu121
• Datasets 2.21.0
• Tokenizers 0.19.1