whisper-test

This model is a fine-tuned version of facebook/wav2vec2-base-960h on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.0795
• Wer: 0.1633

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: 0.0001
• 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
• lr_scheduler_warmup_steps: 1000
• num_epochs: 30
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
1.9037	1.05	500	0.1117	0.2023
0.2943	2.11	1000	0.1147	0.1998
0.2635	3.16	1500	0.1119	0.1984
0.2392	4.21	2000	0.1225	0.2093
0.2019	5.26	2500	0.1917	0.2344
0.1904	6.32	3000	0.1247	0.2069
0.1656	7.37	3500	0.1811	0.2229
0.1597	8.42	4000	0.1950	0.2215
0.1431	9.47	4500	0.1479	0.2163
0.1403	10.53	5000	0.2155	0.2191
0.1238	11.58	5500	0.1103	0.1942
0.1284	12.63	6000	0.1485	0.2018
0.1072	13.68	6500	0.0882	0.1766
0.1014	14.74	7000	0.1009	0.1780
0.1023	15.79	7500	0.0934	0.1737
0.0934	16.84	8000	0.0992	0.1735
0.0916	17.89	8500	0.0986	0.1733
0.0851	18.95	9000	0.0906	0.1710
0.0786	20.0	9500	0.0819	0.1687
0.0773	21.05	10000	0.0827	0.1683
0.0752	22.11	10500	0.0824	0.1679
0.0643	23.16	11000	0.0869	0.1672
0.062	24.21	11500	0.0827	0.1661
0.058	25.26	12000	0.0841	0.1666
0.0551	26.32	12500	0.0828	0.1659
0.0508	27.37	13000	0.0771	0.1643
0.0557	28.42	13500	0.0793	0.1636
0.048	29.47	14000	0.0795	0.1633

Framework versions

• Transformers 4.17.0
• Pytorch 2.5.1+cu121
• Datasets 1.18.3
• Tokenizers 0.20.3