ssc-aln-mms-model-mix-adapt-max2

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

• Loss: 1.0381
• Cer: 0.2006
• Wer: 0.5092

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: 0.001
• train_batch_size: 1
• eval_batch_size: 6
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 2
• 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_steps: 100
• num_epochs: 5
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Cer	Wer
0.9764	0.4145	200	1.0542	0.2188	0.5582
0.9583	0.8290	400	1.1293	0.2147	0.5609
0.9433	1.2425	600	1.0775	0.2154	0.5456
0.9256	1.6570	800	1.1419	0.2108	0.5370
0.8733	2.0705	1000	1.1627	0.2086	0.5344
0.8905	2.4850	1200	1.0674	0.2070	0.5301
0.8976	2.8995	1400	1.0538	0.2067	0.5246
0.8322	3.3130	1600	1.0242	0.2070	0.5285
0.83	3.7275	1800	1.0482	0.2035	0.5182
0.7651	4.1409	2000	1.0452	0.2043	0.5177
0.8179	4.5554	2200	1.0240	0.2032	0.5141
0.7546	4.9699	2400	1.0381	0.2006	0.5092

Framework versions

• Transformers 4.52.1
• Pytorch 2.9.1+cu128
• Datasets 3.6.0
• Tokenizers 0.21.4