base_sami_22k_cont_pt_ftpseudo_wr20

This model was trained from scratch on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 271.5626
• Wer: 0.4858
• Cer: 0.1540

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: 0.0005
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH 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: 0.2
• num_epochs: 60.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
2875.4683	1.0	3628	287.1173	0.5699	0.1733
758.2716	2.0	7256	272.0594	0.4882	0.1551
644.8381	3.0	10884	290.3244	0.5062	0.1729
609.5405	4.0	14512	257.9622	0.5174	0.1867
606.3339	5.0	18140	296.2997	0.5668	0.2280
619.5554	6.0	21768	324.2591	0.5901	0.2188
646.4168	7.0	25396	341.2834	0.5953	0.2440
667.3437	8.0	29024	361.0744	0.6283	0.2255
693.7482	9.0	32652	364.4579	0.6597	0.2685
733.3208	10.0	36280	409.5368	0.7001	0.2802
780.5633	11.0	39908	410.3908	0.7154	0.2849
813.6104	12.0	43536	419.5198	0.7614	0.3089

Framework versions

• Transformers 4.48.3
• Pytorch 2.5.1
• Datasets 3.2.0
• Tokenizers 0.21.0