golesheed/whisper-5-dutch

Whisper Large V2

This model is a fine-tuned version of openai/whisper-large-v2 on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.2801
• Wer: 9.8223

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: 3e-05
• train_batch_size: 12
• 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: 20
• num_epochs: 5

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.5984	0.09	30	0.3378	18.2664
0.3397	0.19	60	0.3041	17.6296
0.3062	0.28	90	0.2931	16.5683
0.2995	0.38	120	0.2821	12.3463
0.2783	0.47	150	0.2799	12.5440
0.2946	0.57	180	0.2626	12.3870
0.2615	0.66	210	0.2652	11.8054
0.2456	0.76	240	0.2654	12.1282
0.2446	0.85	270	0.2579	12.3172
0.2836	0.95	300	0.2494	12.5149
0.1951	1.04	330	0.2531	13.0267
0.1233	1.14	360	0.2517	11.3402
0.1219	1.23	390	0.2512	11.8926
0.144	1.33	420	0.2510	11.6775
0.1268	1.42	450	0.2509	11.9624
0.1366	1.52	480	0.2496	13.8350
0.134	1.61	510	0.2467	11.1279
0.139	1.71	540	0.2465	12.2823
0.1449	1.8	570	0.2428	13.7158
0.1257	1.9	600	0.2400	12.9452
0.1319	1.99	630	0.2335	13.2738
0.066	2.09	660	0.2525	13.3291
0.0517	2.18	690	0.2553	12.6225
0.0595	2.28	720	0.2530	13.2273
0.0638	2.37	750	0.2533	10.6452
0.0612	2.47	780	0.2565	12.9191
0.0633	2.56	810	0.2512	10.7935
0.056	2.66	840	0.2549	13.8263
0.0614	2.75	870	0.2526	11.1803
0.062	2.85	900	0.2555	10.2585
0.0586	2.94	930	0.2456	10.5900
0.0462	3.04	960	0.2538	10.5754
0.0225	3.13	990	0.2629	11.0320
0.0244	3.23	1020	0.2659	10.0695
0.0246	3.32	1050	0.2651	9.8863
0.0217	3.42	1080	0.2652	10.0550
0.0222	3.51	1110	0.2625	10.7063
0.023	3.61	1140	0.2660	10.7470
0.0236	3.7	1170	0.2651	9.9183
0.0214	3.8	1200	0.2642	9.9386
0.0208	3.89	1230	0.2635	9.9619
0.021	3.99	1260	0.2674	9.9794
0.0127	4.08	1290	0.2700	9.6391
0.0096	4.18	1320	0.2783	9.6217
0.0082	4.27	1350	0.2822	9.6799
0.009	4.37	1380	0.2799	9.7729
0.0077	4.46	1410	0.2806	9.9154
0.007	4.56	1440	0.2803	9.7526
0.0078	4.65	1470	0.2805	9.7264
0.0083	4.75	1500	0.2804	9.8427
0.0085	4.84	1530	0.2807	9.8543
0.0081	4.94	1560	0.2801	9.8223

Framework versions

• Transformers 4.38.0.dev0
• Pytorch 2.1.0+cu121
• Datasets 2.17.0
• Tokenizers 0.15.0