base_sami_22k_cont_pt_ftpseudo_wr15

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

• Loss: 279.1175
• Wer: 0.5077
• Cer: 0.1722

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: 0.15
• num_epochs: 60.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
2898.1974	1.0	3628	315.9604	0.6330	0.1893
757.7022	2.0	7256	279.5220	0.5065	0.1721
662.3705	3.0	10884	301.2926	0.5251	0.1764
647.4321	4.0	14512	309.8589	0.5494	0.2027
659.7851	5.0	18140	309.3199	0.5862	0.2070
692.9787	6.0	21768	373.3183	0.6487	0.2412
739.3389	7.0	25396	379.1962	0.6765	0.2556
786.0181	8.0	29024	414.1455	0.7022	0.2729
827.0145	9.0	32652	441.2162	0.7871	0.3400
849.8929	10.0	36280	422.3855	0.7259	0.2963
830.547	11.0	39908	426.6281	0.7506	0.3046
803.7191	12.0	43536	418.4753	0.7314	0.2984

Framework versions

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