deberta-v3-large-ft-icar-a-v1.3

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: 2.1576
• Accuracy: 0.8760
• Precision: 0.8372
• Recall: 0.7144
• F1: 0.7350

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
2.4886	1.0	871	0.7499	0.7963	0.6199	0.6086	0.6017
1.8992	2.0	1742	0.9652	0.8469	0.6961	0.6813	0.6877
1.548	3.0	2613	1.1634	0.8484	0.8299	0.6871	0.6986
1.2239	4.0	3484	1.1407	0.8729	0.8258	0.7136	0.7342
0.8945	5.0	4355	1.1494	0.8530	0.7620	0.7112	0.7293
0.7381	6.0	5226	1.3039	0.8499	0.7615	0.7118	0.7206
0.5129	7.0	6097	1.4266	0.8652	0.8201	0.7121	0.7253
0.4255	8.0	6968	1.5332	0.8591	0.7777	0.7069	0.7206
0.279	9.0	7839	1.5381	0.8606	0.8104	0.7047	0.7180
0.2978	10.0	8710	1.6327	0.8622	0.7995	0.7076	0.7218
0.13	11.0	9581	1.7521	0.8729	0.8319	0.7116	0.7297
0.0562	12.0	10452	2.0256	0.8622	0.8134	0.7106	0.7191
0.0537	13.0	11323	1.9061	0.8683	0.8211	0.7148	0.7269
0.071	14.0	12194	2.1575	0.8714	0.8270	0.7179	0.7305
0.0502	15.0	13065	2.1730	0.8591	0.8223	0.7047	0.7136
0.1456	16.0	13936	2.2140	0.8729	0.8318	0.7145	0.7325
0.1122	17.0	14807	2.3640	0.8637	0.8189	0.7070	0.7220
0.0238	18.0	15678	2.3073	0.8683	0.8297	0.7064	0.7222
0.0762	19.0	16549	2.1576	0.8760	0.8372	0.7144	0.7350

Framework versions

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