golesheed/whisper-9-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: 0.2872
• Wer: 10.3543

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
0.6091	0.09	30	0.3548	15.0266
0.3073	0.19	60	0.3203	13.7016
0.3171	0.28	90	0.3049	12.6189
0.29	0.38	120	0.3033	13.9760
0.2907	0.47	150	0.2824	12.9750
0.2748	0.57	180	0.2737	13.1413
0.2637	0.66	210	0.2655	15.0149
0.2672	0.76	240	0.2629	15.7094
0.2483	0.85	270	0.2616	13.7483
0.2531	0.95	300	0.2603	13.5732
0.1988	1.04	330	0.2713	12.3417
0.1271	1.14	360	0.2644	12.3942
0.1309	1.23	390	0.2612	12.6218
0.1506	1.33	420	0.2633	17.3204
0.1365	1.42	450	0.2621	13.2551
0.1379	1.52	480	0.2636	13.2901
0.1325	1.61	510	0.2550	12.8845
0.129	1.71	540	0.2575	14.0139
0.1334	1.8	570	0.2513	12.2104
0.1418	1.9	600	0.2484	12.2541
0.1438	1.99	630	0.2457	12.0119
0.0651	2.09	660	0.2646	12.3358
0.0649	2.18	690	0.2684	10.6286
0.0638	2.28	720	0.2645	11.6121
0.0651	2.37	750	0.2616	11.4020
0.0656	2.47	780	0.2574	11.4457
0.0643	2.56	810	0.2592	11.7113
0.0682	2.66	840	0.2597	11.5625
0.0583	2.75	870	0.2571	12.9020
0.0608	2.85	900	0.2574	14.3991
0.064	2.94	930	0.2535	10.6023
0.0429	3.04	960	0.2648	10.9788
0.0264	3.13	990	0.2710	10.3514
0.0251	3.23	1020	0.2688	10.4302
0.0244	3.32	1050	0.2709	9.9778
0.0251	3.42	1080	0.2732	10.1733
0.0245	3.51	1110	0.2720	11.1043
0.0246	3.61	1140	0.2765	10.8446
0.0254	3.7	1170	0.2709	10.7658
0.0234	3.8	1200	0.2663	10.3485
0.022	3.89	1230	0.2649	11.4370
0.0237	3.99	1260	0.2688	11.0138
0.011	4.08	1290	0.2791	10.3076
0.0107	4.18	1320	0.2839	10.4798
0.0087	4.27	1350	0.2871	10.4856
0.0081	4.37	1380	0.2894	10.3280
0.0094	4.46	1410	0.2872	10.2259
0.0083	4.56	1440	0.2887	10.2288
0.0104	4.65	1470	0.2856	10.2638
0.009	4.75	1500	0.2855	10.3339
0.0068	4.84	1530	0.2865	10.4010
0.0082	4.94	1560	0.2872	10.3543

Framework versions

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