golesheed/whisper-8-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: 0.2674
• Wer: 8.9178

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	0.3391	13.4234
0.3844	0.19	60	0.2936	16.0882
0.3245	0.28	90	0.2801	12.7436
0.2967	0.38	120	0.2602	12.8549
0.2526	0.47	150	0.2604	17.7364
0.2889	0.57	180	0.2466	13.2940
0.2378	0.66	210	0.2506	15.9919
0.237	0.76	240	0.2500	17.4176
0.2769	0.85	270	0.2340	15.0956
0.2579	0.95	300	0.2365	13.3482
0.1979	1.04	330	0.2461	15.3333
0.1336	1.14	360	0.2416	13.3331
0.1415	1.23	390	0.2380	14.3918
0.1307	1.33	420	0.2397	11.2879
0.1489	1.42	450	0.2389	11.0954
0.1311	1.52	480	0.2378	14.1783
0.1256	1.61	510	0.2333	12.2895
0.1283	1.71	540	0.2318	10.5901
0.1418	1.8	570	0.2317	14.6084
0.1346	1.9	600	0.2284	12.2564
0.1357	1.99	630	0.2212	10.5029
0.0641	2.09	660	0.2369	11.4894
0.0587	2.18	690	0.2383	9.7690
0.0585	2.28	720	0.2378	11.6037
0.0601	2.37	750	0.2409	11.6609
0.0645	2.47	780	0.2397	10.4397
0.0648	2.56	810	0.2430	10.2984
0.0616	2.66	840	0.2421	10.3946
0.0668	2.75	870	0.2351	13.2489
0.0553	2.85	900	0.2343	10.6563
0.0576	2.94	930	0.2359	10.2262
0.0468	3.04	960	0.2433	10.1329
0.0253	3.13	990	0.2496	10.0638
0.025	3.23	1020	0.2480	11.0864
0.0232	3.32	1050	0.2550	9.9916
0.0252	3.42	1080	0.2531	9.3269
0.0254	3.51	1110	0.2472	9.0381
0.0225	3.61	1140	0.2549	9.2908
0.0218	3.7	1170	0.2496	9.5404
0.0242	3.8	1200	0.2432	9.9284
0.0223	3.89	1230	0.2462	10.8277
0.0204	3.99	1260	0.2522	9.6637
0.0115	4.08	1290	0.2585	8.8426
0.0094	4.18	1320	0.2622	9.4923
0.0092	4.27	1350	0.2638	10.6773
0.009	4.37	1380	0.2640	10.0999
0.009	4.46	1410	0.2664	10.0036
0.0087	4.56	1440	0.2666	9.9705
0.0075	4.65	1470	0.2672	9.8622
0.0077	4.75	1500	0.2658	9.1254
0.0069	4.84	1530	0.2667	9.0442
0.0081	4.94	1560	0.2674	8.9178

Framework versions

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