ssc-led-mms-model

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

• Loss: 0.4620
• Cer: 0.1094
• Wer: 0.3057

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: 12
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED 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: 10
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Cer	Wer
2.1523	0.2080	200	1.1494	0.3505	0.9158
1.5516	0.4160	400	0.7710	0.1949	0.5516
1.4143	0.6240	600	0.7063	0.1745	0.5122
1.3084	0.8320	800	0.6080	0.1513	0.4280
1.1718	1.0395	1000	0.6245	0.1469	0.4409
1.1952	1.2475	1200	0.5926	0.1415	0.4116
1.1589	1.4555	1400	0.5521	0.1373	0.3912
1.1722	1.6635	1600	0.5385	0.1328	0.3748
1.1743	1.8716	1800	0.5272	0.1305	0.3714
1.1037	2.0790	2000	0.5284	0.1296	0.3676
1.1259	2.2871	2200	0.5284	0.1281	0.3654
1.1265	2.4951	2400	0.5120	0.1264	0.3555
1.0877	2.7031	2600	0.5108	0.1229	0.3493
1.0929	2.9111	2800	0.5011	0.1232	0.3480
1.0686	3.1186	3000	0.4911	0.1208	0.3389
1.0356	3.3266	3200	0.4911	0.1192	0.3331
0.9567	3.5346	3400	0.4872	0.1190	0.3376
1.1178	3.7426	3600	0.4824	0.1189	0.3322
1.0506	3.9506	3800	0.4804	0.1183	0.3391
1.0343	4.1581	4000	0.4853	0.1202	0.3422
0.9977	4.3661	4200	0.4819	0.1174	0.3226
1.0162	4.5741	4400	0.4835	0.1148	0.3170
1.1006	4.7821	4600	0.4745	0.1172	0.3296
0.9562	4.9901	4800	0.4720	0.1170	0.3233
0.9926	5.1976	5000	0.4808	0.1141	0.3150
0.9625	5.4056	5200	0.4870	0.1143	0.3178
1.0531	5.6136	5400	0.4759	0.1140	0.3177
1.0173	5.8216	5600	0.4701	0.1159	0.3232
0.9915	6.0291	5800	0.4674	0.1129	0.3169
0.9667	6.2371	6000	0.4724	0.1117	0.3093
0.9627	6.4451	6200	0.4630	0.1123	0.3108
0.9505	6.6531	6400	0.4755	0.1119	0.3093
1.0117	6.8612	6600	0.4642	0.1119	0.3075
1.0087	7.0686	6800	0.4668	0.1115	0.3068
0.9271	7.2767	7000	0.4766	0.1116	0.3118
0.949	7.4847	7200	0.4631	0.1130	0.3150
0.8933	7.6927	7400	0.4639	0.1113	0.3107
0.9838	7.9007	7600	0.4662	0.1117	0.3080
0.9619	8.1082	7800	0.4757	0.1105	0.3063
0.9731	8.3162	8000	0.4655	0.1098	0.3040
0.9118	8.5242	8200	0.4722	0.1103	0.3070
0.9808	8.7322	8400	0.4699	0.1105	0.3082
0.8915	8.9402	8600	0.4681	0.1099	0.3059
0.923	9.1477	8800	0.4702	0.1100	0.3043
0.9328	9.3557	9000	0.4674	0.1099	0.3052
0.932	9.5637	9200	0.4656	0.1099	0.3056
0.9293	9.7717	9400	0.4623	0.1093	0.3058
0.9046	9.9797	9600	0.4620	0.1094	0.3057

Framework versions

• Transformers 4.57.2
• Pytorch 2.9.1+cu128
• Datasets 3.6.0
• Tokenizers 0.22.0