mms-1b-bemgen-female-model

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: 0.1921
• Wer: 0.3440

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: 4
• eval_batch_size: 4
• seed: 42
• 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
• num_epochs: 30.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
6.6524	0.1028	100	0.6406	0.6924
0.4626	0.2055	200	0.2750	0.4520
0.3384	0.3083	300	0.2556	0.4587
0.3391	0.4111	400	0.2391	0.4119
0.2928	0.5139	500	0.2326	0.4009
0.3126	0.6166	600	0.2275	0.4031
0.3305	0.7194	700	0.2208	0.3978
0.3043	0.8222	800	0.2123	0.3804
0.2989	0.9250	900	0.2135	0.3793
0.2911	1.0277	1000	0.2117	0.3890
0.2994	1.1305	1100	0.2066	0.3868
0.2849	1.2333	1200	0.2113	0.3932
0.2864	1.3361	1300	0.2018	0.3713
0.2611	1.4388	1400	0.2005	0.4004
0.2995	1.5416	1500	0.1996	0.3729
0.2787	1.6444	1600	0.2015	0.3647
0.2444	1.7472	1700	0.1988	0.3603
0.2734	1.8499	1800	0.1950	0.3590
0.2794	1.9527	1900	0.1953	0.3546
0.2708	2.0555	2000	0.1934	0.3610
0.2545	2.1583	2100	0.1953	0.3616
0.2529	2.2610	2200	0.1940	0.3559
0.2628	2.3638	2300	0.1926	0.3585
0.2788	2.4666	2400	0.1925	0.3510
0.2473	2.5694	2500	0.1878	0.3499
0.2595	2.6721	2600	0.1911	0.3705
0.2516	2.7749	2700	0.1883	0.3480
0.2445	2.8777	2800	0.1921	0.3442

Framework versions

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