exo-5

This model is a fine-tuned version of openai/whisper-tiny on the PolyAI/minds14 dataset. It achieves the following results on the evaluation set:

• Loss: 0.1314
• Wer Ortho: 0.1507
• Wer: 0.0908

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: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.12
• num_epochs: 15
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer Ortho	Wer
0.0443	1.0	57	0.0811	0.1175	0.0680
0.0262	2.0	114	0.1065	0.1454	0.0977
0.0409	3.0	171	0.1275	0.1074	0.0748
0.0204	4.0	228	0.1301	0.1371	0.1057
0.0095	5.0	285	0.1293	0.1982	0.1605
0.0071	6.0	342	0.1422	0.1822	0.1365
0.0011	7.0	399	0.1329	0.1448	0.0982
0.0049	8.0	456	0.1294	0.1359	0.0788
0.0012	9.0	513	0.1296	0.1478	0.0891
0.0002	10.0	570	0.1305	0.1484	0.0891
0.0013	11.0	627	0.1298	0.1490	0.0897
0.0002	12.0	684	0.1309	0.1490	0.0891
0.0004	13.0	741	0.1311	0.1513	0.0914
0.0001	14.0	798	0.1313	0.1507	0.0908
0.0001	15.0	855	0.1314	0.1507	0.0908

Framework versions

• Transformers 4.55.3
• Pytorch 2.8.0+cu128
• Datasets 3.6.0
• Tokenizers 0.21.2