ssc-pne-xlsr300m-model-mix-adapt-max-lowlr

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

• Loss: 0.8773
• Cer: 0.1637
• Wer: 0.4353

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: 0.0003
• train_batch_size: 2
• eval_batch_size: 6
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 4
• 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_steps: 100
• num_epochs: 5
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Cer	Wer
0.2572	0.4415	200	0.8300	0.2183	0.5500
0.2792	0.8830	400	0.8513	0.2168	0.5158
0.2176	1.3245	600	0.7886	0.1994	0.5137
0.2369	1.7660	800	0.7914	0.1844	0.4914
0.1462	2.2075	1000	0.8252	0.1717	0.4565
0.1398	2.6490	1200	0.8196	0.1760	0.4576
0.1279	3.0905	1400	0.8324	0.1742	0.4575
0.0904	3.5320	1600	0.8069	0.1667	0.4446
0.0944	3.9735	1800	0.8357	0.1679	0.4395
0.0636	4.4150	2000	0.8764	0.1652	0.4379
0.0669	4.8565	2200	0.8773	0.1637	0.4353

Framework versions

• Transformers 4.57.2
• Pytorch 2.9.1+cu128
• Datasets 3.6.0
• Tokenizers 0.22.0