golesheed/whisper-0-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: 0.2651
• Wer: 9.8186

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
0.5952	0.13	30	0.3084	14.4155
0.3011	0.25	60	0.2771	17.2921
0.278	0.38	90	0.2652	12.2835
0.2685	0.51	120	0.2496	16.0334
0.2858	0.63	150	0.2387	11.5359
0.2544	0.76	180	0.2352	12.7537
0.2445	0.89	210	0.2288	10.5050
0.2361	1.01	240	0.2276	12.7537
0.1265	1.14	270	0.2309	12.4792
0.1338	1.27	300	0.2316	12.7041
0.1392	1.39	330	0.2285	10.8437
0.1415	1.52	360	0.2284	11.8630
0.1283	1.65	390	0.2266	10.9430
0.1311	1.77	420	0.2288	12.4880
0.1222	1.9	450	0.2201	10.8145
0.1168	2.03	480	0.2257	13.6386
0.0552	2.15	510	0.2346	12.0908
0.0613	2.28	540	0.2244	13.8138
0.0569	2.41	570	0.2306	10.9197
0.0587	2.53	600	0.2332	9.7515
0.0558	2.66	630	0.2352	11.8075
0.0601	2.78	660	0.2295	10.7590
0.0536	2.91	690	0.2294	10.9021
0.051	3.04	720	0.2353	11.0394
0.0244	3.16	750	0.2439	10.4845
0.0218	3.29	780	0.2483	11.0511
0.0218	3.42	810	0.2434	10.5517
0.0222	3.54	840	0.2510	9.8741
0.0209	3.67	870	0.2436	10.8466
0.0219	3.8	900	0.2476	10.4465
0.0228	3.92	930	0.2433	11.1767
0.0149	4.05	960	0.2499	10.1808
0.0079	4.18	990	0.2625	10.4290
0.0083	4.3	1020	0.2650	9.9909
0.0085	4.43	1050	0.2641	10.1194
0.0085	4.56	1080	0.2637	10.5750
0.0077	4.68	1110	0.2649	10.0055
0.0077	4.81	1140	0.2654	9.9734
0.0085	4.94	1170	0.2651	9.8186

Framework versions

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