deberta-misconception-classifier

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: 0.2595
• Macro F1: 0.6012
• Weighted F1: 0.7862
• Accuracy: 0.7823
• Map@3: 0.8846

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: 1e-05
• 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
• num_epochs: 10
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Macro F1	Weighted F1	Accuracy	Map@3
1.3548	0.2422	500	1.0357	0.2067	0.4193	0.4221	0.5941
0.9062	0.4845	1000	0.7145	0.3536	0.6222	0.6183	0.7672
0.5924	0.7267	1500	0.4780	0.4251	0.7250	0.7368	0.8460
0.4113	0.9690	2000	0.4354	0.4210	0.7139	0.7354	0.8430
0.2906	1.2112	2500	0.3885	0.4757	0.7373	0.7559	0.8635
0.3248	1.4535	3000	0.3100	0.5215	0.7591	0.7589	0.8651
0.264	1.6957	3500	0.3245	0.5371	0.7838	0.7864	0.8852
0.3461	1.9380	4000	0.2863	0.5582	0.8036	0.8136	0.8988
0.202	2.1802	4500	0.2697	0.5758	0.8058	0.8147	0.9013
0.1641	2.4225	5000	0.2837	0.6015	0.8224	0.8245	0.9062
0.1642	2.6647	5500	0.2991	0.5559	0.8113	0.8139	0.9009
0.1857	2.9070	6000	0.2518	0.5931	0.8051	0.8109	0.8995
0.1322	3.1492	6500	0.2595	0.6012	0.7862	0.7823	0.8846

Framework versions

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