ssc-cgg-mms-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: 0.3659
• Cer: 0.1318
• Wer: 0.5826

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
0.4125	0.4520	200	0.3951	0.1396	0.6133
0.4094	0.9040	400	0.3917	0.1356	0.5983
0.3807	1.3548	600	0.3838	0.1343	0.5959
0.3611	1.8068	800	0.3819	0.1346	0.5998
0.3419	2.2576	1000	0.3781	0.1353	0.5925
0.3907	2.7096	1200	0.3777	0.1332	0.5871
0.3712	3.1605	1400	0.3729	0.1323	0.5857
0.3658	3.6124	1600	0.3704	0.1330	0.5871
0.4114	4.0633	1800	0.3680	0.1320	0.5836
0.3342	4.5153	2000	0.3677	0.1317	0.5856
0.3695	4.9672	2200	0.3659	0.1318	0.5826

Framework versions

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