ssc-tob-mms-model-mix-adapt-max-lowlr

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: 0.7360
• Cer: 0.1766
• Wer: 0.6035

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: 6
• 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
1.1871	0.7648	200	0.8279	0.1866	0.6352
0.7886	1.5277	400	0.7769	0.1887	0.6419
0.7329	2.2906	600	0.7502	0.1785	0.6124
0.6881	3.0535	800	0.7380	0.1798	0.6157
0.647	3.8184	1000	0.7332	0.1798	0.6102
0.673	4.5813	1200	0.7311	0.1805	0.6145
0.5852	5.3442	1400	0.7352	0.1780	0.6076
0.6142	6.1071	1600	0.7315	0.1774	0.6054
0.6056	6.8719	1800	0.7371	0.1778	0.6061
0.6021	7.6348	2000	0.7350	0.1772	0.6072
0.5971	8.3977	2200	0.7379	0.1787	0.6102
0.5732	9.1606	2400	0.7342	0.1762	0.6020
0.5988	9.9254	2600	0.7360	0.1766	0.6035

Framework versions

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