deberta-v3-large-ft-icar-a-v0.10

This model is a fine-tuned version of microsoft/deberta-v3-large on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.9480
• Accuracy: 0.9173
• Precision: 0.9147
• Recall: 0.9119
• F1: 0.9126

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: 3e-06
• train_batch_size: 1
• eval_batch_size: 1
• seed: 42
• gradient_accumulation_steps: 3
• total_train_batch_size: 3
• 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
• num_epochs: 50

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Precision	Recall	F1
2.5523	1.0	871	0.6494	0.8683	0.8773	0.8534	0.8630
1.8228	2.0	1742	0.7068	0.8897	0.9008	0.8642	0.8751
1.4126	3.0	2613	0.6934	0.9005	0.8904	0.8965	0.8927
1.0964	4.0	3484	0.6747	0.9158	0.9111	0.9062	0.9082
0.8857	5.0	4355	0.8318	0.8989	0.8926	0.8980	0.8939
0.6865	6.0	5226	0.8226	0.9066	0.9050	0.8958	0.8996
0.5458	7.0	6097	0.7935	0.9158	0.9105	0.9089	0.9090
0.5311	8.0	6968	0.8565	0.9051	0.9062	0.8928	0.8982
0.3778	9.0	7839	0.8764	0.9158	0.9096	0.9096	0.9092
0.3385	10.0	8710	0.9250	0.9005	0.8790	0.9024	0.8897
0.226	11.0	9581	0.9597	0.9066	0.9104	0.8892	0.8979
0.2004	12.0	10452	1.0016	0.9127	0.9131	0.9083	0.9101
0.1238	13.0	11323	0.9480	0.9173	0.9147	0.9119	0.9126
0.138	14.0	12194	0.9453	0.9127	0.9080	0.9023	0.9045
0.1447	15.0	13065	1.0570	0.8943	0.8884	0.8996	0.8917
0.149	16.0	13936	1.1390	0.8989	0.8953	0.8846	0.8889
0.1106	17.0	14807	1.1037	0.9051	0.8952	0.8976	0.8960
0.0848	18.0	15678	1.0446	0.9066	0.9045	0.9071	0.9048
0.0556	19.0	16549	1.1846	0.9035	0.9052	0.8954	0.8997

Framework versions

• Transformers 4.52.4
• Pytorch 2.6.0+cu124
• Datasets 4.0.0
• Tokenizers 0.21.2