mms-1b-bigcgen-combined-5hrs-model

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

• Loss: inf
• Wer: 0.5720

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: 5e-05
• train_batch_size: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 8
• optimizer: Use 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
• training_steps: 2500
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
19.6398	0.3049	100	inf	1.0
6.659	0.6098	200	inf	0.9999
3.4985	0.9146	300	inf	0.7555
2.2439	1.2195	400	inf	0.6923
2.0278	1.5244	500	inf	0.6787
1.7863	1.8293	600	inf	0.6540
1.7217	2.1341	700	inf	0.6249
1.7965	2.4390	800	inf	0.6127
1.6945	2.7439	900	inf	0.6080
1.8135	3.0488	1000	inf	0.6049
1.7788	3.3537	1100	inf	0.5962
1.7022	3.6585	1200	inf	0.5921
1.5461	3.9634	1300	inf	0.5893
1.7282	4.2683	1400	inf	0.5866
1.6337	4.5732	1500	inf	0.5829
1.6981	4.8780	1600	inf	0.5805
1.62	5.1829	1700	inf	0.5764
1.7366	5.4878	1800	inf	0.5782
1.5943	5.7927	1900	inf	0.5757
1.6907	6.0976	2000	inf	0.5773
1.698	6.4024	2100	inf	0.5738
1.6108	6.7073	2200	inf	0.5740
1.6195	7.0122	2300	inf	0.5733
1.6189	7.3171	2400	inf	0.5731
1.7298	7.6220	2500	inf	0.5724

Framework versions

• Transformers 4.47.1
• Pytorch 2.5.1+cu124
• Datasets 3.2.0
• Tokenizers 0.21.0