sarpba/whisper-base-cv18-hu-cleaned

Whisper Base Hu CV18

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

• Loss: 0.9673
• Wer Ortho: 43.0437
• Wer: 36.6182

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: 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
0.4737	0.1723	250	0.9877	61.4013	58.1242
0.3609	0.3446	500	0.9450	56.6701	52.1893
0.2894	0.5169	750	0.9007	53.2487	48.7805
0.2652	0.6892	1000	0.8833	52.5755	47.6039
0.225	0.8615	1250	0.8573	49.4478	45.3212
0.1414	1.0338	1500	0.8702	48.7835	43.8570
0.1361	1.2061	1750	0.8836	48.6700	43.0146
0.1297	1.3784	2000	0.8782	48.1922	42.8923
0.1285	1.5507	2250	0.8695	47.0576	41.8178
0.1241	1.7229	2500	0.8498	46.3403	40.9186
0.1198	1.8952	2750	0.8658	46.1928	40.0787
0.0552	2.0675	3000	0.8843	45.9684	39.5667
0.0608	2.2398	3250	0.8747	45.2347	39.5112
0.0576	2.4121	3500	0.8752	45.0557	39.7067
0.0629	2.5844	3750	0.8949	45.2297	39.2073
0.0613	2.7567	4000	0.9124	45.4137	39.0811
0.0542	2.9290	4250	0.8890	44.1443	38.4127
0.0248	3.1013	4500	0.9102	44.2388	37.9159
0.0253	3.2736	4750	0.9119	43.5908	37.3130
0.0248	3.4459	5000	0.9342	44.2325	37.8515
0.0238	3.6182	5250	0.9300	44.0018	37.6712
0.0241	3.7905	5500	0.9281	43.6614	37.3710
0.0231	3.9628	5750	0.9352	43.6715	37.7469
0.0101	4.1351	6000	0.9549	43.3387	37.2046
0.0083	4.3074	6250	0.9580	43.2795	36.9889
0.0082	4.4797	6500	0.9571	43.2152	37.0759
0.0087	4.6520	6750	0.9592	42.6290	36.3912
0.0077	4.8243	7000	0.9675	42.9139	36.5437
0.0076	4.9966	7250	0.9673	43.0437	36.6182

Framework versions

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