mms-1b-all-bemgen-combined-m50f100-42-DAT-2e-1

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: 0.3343
• Cer: 0.0892

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: 42
• 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
1.9663	0.5618	100	3.0571	1.0
0.7101	1.1236	200	1.8286	0.6320
0.5324	1.6854	300	0.4045	0.1061
0.5969	2.2472	400	0.3343	0.0892
0.595	2.8090	500	0.3095	0.0822
0.588	3.3708	600	0.3278	0.0889
0.6406	3.9326	700	0.3181	0.0857
0.6429	4.4944	800	0.3297	0.0888

Framework versions

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