mms-1b-all-swagen-male-5hrs-42

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

• Loss: 0.3874
• Wer: 0.2498

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	Wer
6.9421	0.4854	100	3.2542	1.0074
2.7173	0.9709	200	2.2981	0.9449
1.9015	1.4563	300	1.4669	0.8143
1.08	1.9417	400	0.8388	0.4645
0.7296	2.4272	500	0.6773	0.3854
0.6061	2.9126	600	0.5788	0.3479
0.5115	3.3981	700	0.5366	0.3287
0.471	3.8835	800	0.4893	0.3109
0.4547	4.3689	900	0.4581	0.2972
0.4147	4.8544	1000	0.4429	0.2900
0.4097	5.3398	1100	0.4415	0.2828
0.3895	5.8252	1200	0.4283	0.2716
0.3815	6.3107	1300	0.4205	0.2779
0.3699	6.7961	1400	0.4121	0.2642
0.3338	7.2816	1500	0.4084	0.2627
0.3466	7.7670	1600	0.3896	0.2550
0.3482	8.2524	1700	0.3992	0.2558
0.3235	8.7379	1800	0.4046	0.2541
0.3077	9.2233	1900	0.3888	0.2455
0.3101	9.7087	2000	0.3875	0.2498
0.3019	10.1942	2100	0.3791	0.2466
0.2943	10.6796	2200	0.3807	0.2414
0.315	11.1650	2300	0.3802	0.2419
0.2931	11.6505	2400	0.3831	0.2386

Framework versions

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