golesheed/whisper-1-dutch

Whisper Large V2

This model is a fine-tuned version of openai/whisper-large-v2 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.2995
• Wer: 10.8891

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: 16
• 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
0.5753	0.13	30	0.3392	14.8670
0.2975	0.25	60	0.3042	19.4403
0.2893	0.38	90	0.3097	38.1677
0.2771	0.51	120	0.2772	13.8256
0.2656	0.63	150	0.2709	15.7969
0.2518	0.76	180	0.2602	14.6440
0.2419	0.89	210	0.2567	16.0404
0.2391	1.01	240	0.2599	13.9781
0.1319	1.14	270	0.2546	13.2594
0.128	1.27	300	0.2591	18.8448
0.1299	1.39	330	0.2599	11.7135
0.1261	1.52	360	0.2587	13.1303
0.1329	1.65	390	0.2541	12.4850
0.1303	1.77	420	0.2501	11.9980
0.115	1.9	450	0.2506	13.2799
0.1189	2.03	480	0.2486	11.0270
0.0577	2.15	510	0.2554	12.2532
0.0566	2.28	540	0.2587	11.3702
0.0573	2.41	570	0.2651	10.9390
0.0533	2.53	600	0.2590	11.1473
0.0519	2.66	630	0.2636	10.8363
0.056	2.78	660	0.2577	11.3732
0.062	2.91	690	0.2537	11.9833
0.0447	3.04	720	0.2675	11.7017
0.0232	3.16	750	0.2757	11.7927
0.0237	3.29	780	0.2751	12.5378
0.0216	3.42	810	0.2791	12.0244
0.0195	3.54	840	0.2787	10.9390
0.022	3.67	870	0.2749	12.4967
0.0213	3.8	900	0.2757	11.9012
0.0201	3.92	930	0.2735	11.5052
0.0146	4.05	960	0.2854	11.3673
0.0096	4.18	990	0.2956	11.7399
0.008	4.3	1020	0.2955	11.8073
0.008	4.43	1050	0.2963	11.8044
0.0087	4.56	1080	0.2971	11.6519
0.0076	4.68	1110	0.2971	11.7017
0.0073	4.81	1140	0.2982	10.8246
0.0068	4.94	1170	0.2995	10.8891

Framework versions

• Transformers 4.38.0.dev0
• Pytorch 2.1.0+cu121
• Datasets 2.14.6
• Tokenizers 0.15.0