ssc-qxp-mms-model-mix-adapt-max2

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: 0.3115
• Cer: 0.0961
• Wer: 0.5349

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: 0.001
• 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: 40
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Cer	Wer
4.2051	0.9975	200	4.1710	0.8497	1.0
2.7502	1.9925	400	2.8237	0.8696	1.0009
2.5117	2.9875	600	2.5394	0.8131	1.0037
2.3454	3.9825	800	2.3081	0.7919	1.0018
1.9502	4.9776	1000	1.7710	0.6450	0.9972
1.3568	5.9726	1200	1.1673	0.3991	0.9614
0.9912	6.9676	1400	0.8145	0.2613	0.8621
0.8234	7.9626	1600	0.7160	0.2247	0.8346
0.6972	8.9576	1800	0.6836	0.1997	0.7767
0.6133	9.9526	2000	0.5995	0.1721	0.7491
0.5654	10.9476	2200	0.5727	0.1684	0.6967
0.5331	11.9426	2400	0.5843	0.1603	0.7022
0.4988	12.9377	2600	0.5104	0.1494	0.6746
0.4952	13.9327	2800	0.4855	0.1416	0.6489
0.4607	14.9277	3000	0.4461	0.1365	0.6553
0.43	15.9227	3200	0.4764	0.1399	0.6415
0.3792	16.9177	3400	0.5336	0.1255	0.5938
0.3817	17.9127	3600	0.4101	0.1323	0.6287
0.3681	18.9077	3800	0.3829	0.1188	0.6094
0.3436	19.9027	4000	0.4131	0.1190	0.5956
0.3244	20.8978	4200	0.3923	0.1233	0.6029
0.3224	21.8928	4400	0.3739	0.1194	0.5956
0.2962	22.8878	4600	0.3631	0.1139	0.5965
0.303	23.8828	4800	0.3541	0.1123	0.5818
0.2881	24.8778	5000	0.3626	0.1132	0.5680
0.2844	25.8728	5200	0.3890	0.1185	0.5892
0.269	26.8678	5400	0.3426	0.1072	0.5763
0.2573	27.8628	5600	0.3284	0.1068	0.5653
0.2377	28.8579	5800	0.3249	0.1023	0.5524
0.2397	29.8529	6000	0.3283	0.1028	0.5515
0.2306	30.8479	6200	0.3960	0.1082	0.5680
0.23	31.8429	6400	0.3104	0.0999	0.5395
0.2207	32.8379	6600	0.3318	0.1063	0.5662
0.2102	33.8329	6800	0.3183	0.1025	0.5653
0.1997	34.8279	7000	0.3179	0.0987	0.5285
0.2022	35.8229	7200	0.3345	0.0953	0.5303
0.193	36.8180	7400	0.3143	0.0977	0.5432
0.1839	37.8130	7600	0.3121	0.0978	0.5423
0.1788	38.8080	7800	0.3107	0.0973	0.5322
0.1807	39.8030	8000	0.3115	0.0961	0.5349

Framework versions

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