ssc-qxp-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: 2.4935
• Cer: 0.8858
• Wer: 1.0

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: 10
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Cer	Wer
4.7441	0.6700	200	4.5692	0.8499	1.0
3.1454	1.3384	400	2.8762	0.8834	1.0
2.8759	2.0067	600	2.7991	0.8939	1.0
2.7927	2.6767	800	2.7170	0.8714	1.0
2.6427	3.3451	1000	2.5878	0.8401	1.0046
2.6826	4.0134	1200	2.6477	0.9101	1.0
2.6617	4.6834	1400	2.5153	0.8783	1.0
2.5807	5.3518	1600	2.4798	0.8144	1.0055
2.5083	6.0201	1800	2.4124	0.8721	1.0
2.5154	6.6901	2000	2.3597	0.8803	1.0
2.7958	7.3585	2200	2.7983	0.9183	1.0
2.7276	8.0268	2400	2.5911	0.9095	1.0
2.649	8.6968	2600	2.5195	0.8957	1.0
2.5892	9.3652	2800	2.4935	0.8858	1.0

Framework versions

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