ssc-ttj-mms-model-initadapt-bl

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: 0.1886
• Cer: 0.0713
• Wer: 0.3995

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: 4
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 8
• 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
0.3941	1.1114	200	0.2399	0.0845	0.4563
0.3327	2.2228	400	0.2173	0.0776	0.4271
0.3033	3.3343	600	0.2103	0.0771	0.4290
0.288	4.4457	800	0.2085	0.0751	0.4172
0.2646	5.5571	1000	0.1984	0.0740	0.4130
0.2616	6.6685	1200	0.1974	0.0726	0.4052
0.2385	7.7799	1400	0.1950	0.0737	0.4014
0.2265	8.8914	1600	0.1892	0.0715	0.4003
0.2392	10.0	1800	0.1886	0.0713	0.3995

Framework versions

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