ssc-qxp-mms-model-mix-adapt-max-longcv2

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: 0.1744
• Cer: 0.0639
• Wer: 0.4265

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: 8
• 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
0.7351	0.9524	200	0.3521	0.1190	0.6562
0.2168	1.9048	400	0.1954	0.0808	0.4807
0.187	2.8571	600	0.1856	0.0777	0.4623
0.1725	3.8095	800	0.1876	0.0786	0.4779
0.1661	4.7619	1000	0.1985	0.0779	0.4577
0.162	5.7143	1200	0.1795	0.0722	0.4623
0.1491	6.6667	1400	0.1709	0.0712	0.4540
0.1417	7.6190	1600	0.1825	0.0732	0.4632
0.14	8.5714	1800	0.1672	0.0695	0.4550
0.1235	9.5238	2000	0.1890	0.0738	0.4531
0.1204	10.4762	2200	0.1660	0.0658	0.4568
0.1192	11.4286	2400	0.1698	0.0681	0.4485
0.117	12.3810	2600	0.1735	0.0709	0.4421
0.1119	13.3333	2800	0.1782	0.0707	0.4403
0.1047	14.2857	3000	0.1737	0.0696	0.4476
0.1068	15.2381	3200	0.1709	0.0675	0.4458
0.0906	16.1905	3400	0.1748	0.0685	0.4393
0.0959	17.1429	3600	0.1700	0.0673	0.4439
0.0958	18.0952	3800	0.1826	0.0696	0.4393
0.0956	19.0476	4000	0.1691	0.0666	0.4384
0.0915	20.0	4200	0.1759	0.0677	0.4393
0.0881	20.9524	4400	0.1732	0.0677	0.4421
0.0864	21.9048	4600	0.1759	0.0678	0.4403
0.0836	22.8571	4800	0.1757	0.0661	0.4329
0.081	23.8095	5000	0.1752	0.0640	0.4311
0.0764	24.7619	5200	0.1710	0.0640	0.4366
0.0735	25.7143	5400	0.1732	0.0637	0.4357
0.0714	26.6667	5600	0.1760	0.0638	0.4320
0.0668	27.6190	5800	0.1775	0.0644	0.4329
0.0654	28.5714	6000	0.1735	0.0635	0.4311
0.0647	29.5238	6200	0.1744	0.0639	0.4265

Framework versions

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