ssc-ukv-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: 0.9252
• Cer: 0.1584
• Wer: 0.4522

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
0.3411	0.5312	200	0.8065	0.1925	0.5386
0.3151	1.0611	400	0.7449	0.1757	0.4866
0.2706	1.5923	600	0.7683	0.1778	0.5014
0.2013	2.1222	800	0.8294	0.1689	0.4758
0.1876	2.6534	1000	0.8056	0.1640	0.4668
0.1272	3.1833	1200	0.8668	0.1651	0.4687
0.1163	3.7145	1400	0.8754	0.1605	0.4574
0.0709	4.2444	1600	0.9090	0.1613	0.4561
0.0717	4.7756	1800	0.9252	0.1584	0.4522

Framework versions

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