mms-1b-all-bemgen-baseline-52

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: 0.2075
• Wer: 0.4022

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: 52
• 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	Wer
5.9375	0.5115	100	0.5113	0.6293
0.5664	1.0205	200	0.2328	0.4386
0.4652	1.5320	300	0.2238	0.4194
0.4601	2.0409	400	0.2172	0.4217
0.4309	2.5524	500	0.2128	0.4084
0.4337	3.0614	600	0.2101	0.4055
0.4026	3.5729	700	0.2125	0.4065
0.4023	4.0818	800	0.2075	0.4022
0.3906	4.5934	900	0.2086	0.4074
0.3997	5.1023	1000	0.2068	0.4031
0.3779	5.6138	1100	0.2066	0.3981
0.3857	6.1228	1200	0.2074	0.3962
0.3685	6.6343	1300	0.2047	0.3954
0.3737	7.1432	1400	0.2047	0.3951
0.3672	7.6547	1500	0.2061	0.3935
0.359	8.1637	1600	0.2069	0.3958
0.3536	8.6752	1700	0.2047	0.3960
0.3542	9.1841	1800	0.2039	0.3973
0.3571	9.6957	1900	0.2003	0.3838
0.3425	10.2046	2000	0.2040	0.3919
0.3449	10.7161	2100	0.2041	0.3926
0.3389	11.2251	2200	0.2020	0.3854
0.3493	11.7366	2300	0.2030	0.3910

Framework versions

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