Hungarians/whisper-base-cv17-hu

Whisper Base Hu CV17

This model is a fine-tuned version of openai/whisper-small on the Common Voice 17.0 dataset. It achieves the following results on the evaluation set:

• Loss: 0.1148
• Wer Ortho: 8.9576
• Wer: 8.1322

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
• training_steps: 5000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer Ortho	Wer
0.5156	0.3298	250	0.5620	52.2010	49.0898
0.3906	0.6596	500	0.4262	43.7131	40.2668
0.3276	0.9894	750	0.3243	33.3915	30.2728
0.2161	1.3193	1000	0.2639	27.8152	24.9778
0.2009	1.6491	1250	0.2314	24.8705	21.9923
0.1835	1.9789	1500	0.1938	21.4922	18.8260
0.0973	2.3087	1750	0.1748	18.8396	16.0243
0.0963	2.6385	2000	0.1600	17.0240	14.8651
0.0913	2.9683	2250	0.1414	14.0853	12.1198
0.046	3.2982	2500	0.1374	13.2000	11.4468
0.0447	3.6280	2750	0.1306	12.5677	10.9191
0.0409	3.9578	3000	0.1216	11.1436	9.8251
0.0173	4.2876	3250	0.1205	10.4812	9.2292
0.0165	4.6174	3500	0.1180	10.2343	9.0898
0.0152	4.9472	3750	0.1149	9.6200	8.5562
0.0061	5.2770	4000	0.1149	9.1021	8.1589
0.0056	5.6069	4250	0.1144	9.2406	8.2864
0.006	5.9367	4500	0.1138	9.0630	8.1559
0.0036	6.2665	4750	0.1148	9.0148	8.1737
0.0033	6.5963	5000	0.1148	8.9576	8.1322

Framework versions

• Transformers 4.41.1
• Pytorch 2.2.1+cu121
• Datasets 2.19.1
• Tokenizers 0.19.1