cantillation/whisper-medium-he-teamim-silsuless

he

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

• Loss: 0.0478
• Wer: 7.1707

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: 4e-05
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• training_steps: 1500
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
1.3738	0.02	50	0.1757	6.6641
0.1077	0.03	100	0.1403	33.0086
0.1118	0.05	150	0.1070	40.2962
0.0848	0.07	200	0.1144	16.7966
0.0879	0.08	250	0.1032	21.4341
0.0786	0.1	300	0.0971	7.0538
0.0665	0.11	350	0.0940	15.3157
0.0631	0.13	400	0.0868	14.6142
0.0591	0.15	450	0.0965	27.0460
0.0574	0.16	500	0.0807	19.0959
0.0536	0.18	550	0.0804	12.4318
0.0443	0.2	600	0.0791	3.9751
0.0455	0.21	650	0.0881	12.9774
0.0435	0.23	700	0.0772	6.0016
0.0439	0.25	750	0.0750	13.9906
0.0317	0.26	800	0.0645	9.4310
0.0352	0.28	850	0.0664	5.9626
0.0328	0.29	900	0.0633	8.4178
0.0287	0.31	950	0.0658	15.5105
0.0266	0.33	1000	0.0570	10.2494
0.0209	0.34	1050	0.0613	7.4825
0.0272	0.36	1100	0.0599	6.8589
0.024	0.38	1150	0.0583	10.8730
0.0203	0.39	1200	0.0538	8.0670
0.0187	0.41	1250	0.0508	5.4949
0.0165	0.43	1300	0.0507	6.8200
0.0106	0.44	1350	0.0504	7.3266
0.0203	0.46	1400	0.0477	7.7163
0.0142	0.48	1450	0.0480	7.5994
0.0194	0.49	1500	0.0478	7.1707

Framework versions

• Transformers 4.37.0.dev0
• Pytorch 2.1.2
• Datasets 2.16.1
• Tokenizers 0.15.0