ssc-bas-mms-model-mix-adapt-max3

This model is a fine-tuned version of facebook/mms-1b-all on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.1415
• Cer: 0.0973
• Wer: 0.3790

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: 0.0005
• train_batch_size: 8
• eval_batch_size: 6
• 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
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Cer	Wer
0.4901	0.7055	200	0.1873	0.1106	0.4235
0.3613	1.4092	400	0.1625	0.1066	0.4147
0.3119	2.1129	600	0.1547	0.1074	0.4204
0.3175	2.8183	800	0.1467	0.1034	0.4090
0.2935	3.5220	1000	0.1487	0.1027	0.4014
0.2797	4.2257	1200	0.1433	0.1014	0.3972
0.2612	4.9312	1400	0.1384	0.1008	0.3914
0.2594	5.6349	1600	0.1383	0.1004	0.3869
0.2461	6.3386	1800	0.1391	0.1019	0.3975
0.2402	7.0423	2000	0.1411	0.1023	0.4005
0.2389	7.7478	2200	0.1395	0.0998	0.3863
0.2315	8.4515	2400	0.1358	0.1008	0.3935
0.2142	9.1552	2600	0.1372	0.0990	0.3845
0.2099	9.8607	2800	0.1393	0.0986	0.3860
0.215	10.5644	3000	0.1346	0.0994	0.3881
0.2065	11.2681	3200	0.1366	0.1002	0.3887
0.2107	11.9735	3400	0.1341	0.0990	0.3875
0.1844	12.6772	3600	0.1394	0.0984	0.3799
0.186	13.3810	3800	0.1346	0.0980	0.3820
0.1754	14.0847	4000	0.1355	0.0982	0.3808
0.1758	14.7901	4200	0.1349	0.0975	0.3790
0.1785	15.4938	4400	0.1393	0.0980	0.3796
0.1764	16.1975	4600	0.1349	0.0984	0.3817
0.1715	16.9030	4800	0.1322	0.0973	0.3778
0.1625	17.6067	5000	0.1359	0.0991	0.3866
0.1625	18.3104	5200	0.1348	0.0989	0.3838
0.1683	19.0141	5400	0.1371	0.0971	0.3790
0.1507	19.7196	5600	0.1337	0.0965	0.3769
0.1465	20.4233	5800	0.1377	0.0973	0.3793
0.1353	21.1270	6000	0.1366	0.0977	0.3775
0.1471	21.8325	6200	0.1411	0.0975	0.3799
0.1449	22.5362	6400	0.1400	0.0976	0.3820
0.1258	23.2399	6600	0.1396	0.0977	0.3778
0.1364	23.9453	6800	0.1419	0.0977	0.3763
0.1435	24.6490	7000	0.1403	0.0970	0.3763
0.1279	25.3527	7200	0.1410	0.0975	0.3760
0.1378	26.0564	7400	0.1395	0.0966	0.3742
0.1253	26.7619	7600	0.1447	0.0979	0.3784
0.1322	27.4656	7800	0.1417	0.0970	0.3793
0.1208	28.1693	8000	0.1429	0.0980	0.3799
0.1205	28.8748	8200	0.1413	0.0975	0.3787
0.1232	29.5785	8400	0.1415	0.0973	0.3790

Framework versions

• Transformers 4.52.1
• Pytorch 2.9.1+cu128
• Datasets 3.6.0
• Tokenizers 0.21.4