mms-1b-all-bemgen-combined-m25f100-62-DAT-0.8

This model is a fine-tuned version of facebook/mms-1b-all on the BEMGEN - BEM dataset. It achieves the following results on the evaluation set:

• Loss: 0.2775
• Cer: 0.0767

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: 8
• eval_batch_size: 4
• seed: 62
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• 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: 30.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Cer
8.2097	0.6711	100	2.8926	0.9993
2.5698	1.3423	200	0.5546	0.2055
1.4278	2.0134	300	0.3571	0.1014
1.2464	2.6846	400	0.3270	0.0947
1.2138	3.3557	500	0.3035	0.0865
1.1864	4.0268	600	0.2977	0.0842
1.1346	4.6980	700	0.2831	0.0781
1.2265	5.3691	800	0.2776	0.0768
1.2146	6.0403	900	0.2766	0.0755
1.1592	6.7114	1000	0.2715	0.0764
1.1769	7.3826	1100	0.2655	0.0742
1.1731	8.0537	1200	0.2720	0.0751
1.1623	8.7248	1300	0.2697	0.0750
1.1975	9.3960	1400	0.2682	0.0750

Framework versions

• Transformers 4.53.0.dev0
• Pytorch 2.6.0+cu124
• Datasets 3.6.0
• Tokenizers 0.21.0