deberta-misconception

This model is a fine-tuned version of microsoft/deberta-v3-large on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.1066
• Macro F1: 0.5639
• Weighted F1: 0.7517
• Accuracy: 0.7177

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: 2e-05
• train_batch_size: 32
• eval_batch_size: 32
• seed: 42
• 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: 10
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Macro F1	Weighted F1	Accuracy
0.4896	0.4840	500	0.4428	0.2597	0.1155	0.2245
0.2755	0.9681	1000	0.2203	0.4467	0.6258	0.5809
0.1658	1.4521	1500	0.1576	0.5330	0.7263	0.6850
0.1688	1.9361	2000	0.1388	0.5112	0.6329	0.5902
0.0482	2.4201	2500	0.1152	0.5605	0.7041	0.6700
0.0269	2.9042	3000	0.1368	0.5653	0.6868	0.6480
0.1069	3.3882	3500	0.1131	0.5633	0.7404	0.7054
0.0304	3.8722	4000	0.1527	0.5592	0.7287	0.6965
0.0577	4.3562	4500	0.1066	0.5639	0.7517	0.7177

Framework versions

• Transformers 4.53.3
• Pytorch 2.6.0+cu124
• Datasets 4.0.0
• Tokenizers 0.21.2