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

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: 0.6787
• Cer: 0.2539
• Wer: 0.8033

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: 15
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Cer	Wer
3.0113	0.9524	200	2.9124	0.8802	1.0138
2.793	1.9048	400	2.6125	0.8607	1.0221
2.6415	2.8571	600	2.5472	0.8418	1.0074
2.4118	3.8095	800	2.4415	0.7797	1.0028
2.3343	4.7619	1000	2.3311	0.8075	1.0083
2.2079	5.7143	1200	2.1933	0.7873	0.9917
1.9533	6.6667	1400	1.7645	0.6747	0.9798
1.6268	7.6190	1600	1.4405	0.5509	0.9504
1.3779	8.5714	1800	1.1779	0.4539	0.9164
1.1936	9.5238	2000	1.0616	0.3850	0.8980
1.0331	10.4762	2200	0.9177	0.3650	0.8704
0.9716	11.4286	2400	0.8287	0.3159	0.8428
0.8937	12.3810	2600	0.7543	0.2861	0.8373
0.8052	13.3333	2800	0.6951	0.2653	0.8125
0.757	14.2857	3000	0.6787	0.2539	0.8033

Framework versions

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