ModernBERT-large_nli

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

• Loss: 2.6038
• Accuracy: 0.5787
• Precision Macro: 0.5794
• Recall Macro: 0.5790
• F1 Macro: 0.5792
• F1 Weighted: 0.5788

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: 5e-05
• train_batch_size: 64
• eval_batch_size: 64
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 128
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 20
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Precision Macro	Recall Macro	F1 Macro	F1 Weighted
2.1283	1.0	143	1.0136	0.4807	0.4674	0.4835	0.4509	0.4492
1.8848	2.0	286	0.9818	0.5202	0.5745	0.5219	0.5042	0.5038
1.7416	3.0	429	1.1233	0.3220	0.2102	0.3259	0.2190	0.2174
2.2168	4.0	572	1.1135	0.3277	0.1092	0.3333	0.1646	0.1618
2.2099	5.0	715	1.1089	0.3277	0.1092	0.3333	0.1646	0.1618
2.2191	6.0	858	1.1231	0.3282	0.4426	0.3338	0.1655	0.1627
2.2027	7.0	1001	1.0931	0.3774	0.2508	0.3801	0.3016	0.2993
2.1846	8.0	1144	1.0723	0.4013	0.3861	0.3995	0.3692	0.3705
2.1232	9.0	1287	1.0461	0.4244	0.4225	0.4248	0.4203	0.4202
2.0586	10.0	1430	1.0345	0.4510	0.4495	0.4494	0.4210	0.4220
2.0578	11.0	1573	1.0390	0.4523	0.4797	0.4511	0.4522	0.4525
2.0289	12.0	1716	1.0626	0.4665	0.5296	0.4668	0.4391	0.4389
1.5688	13.0	1859	0.8686	0.6084	0.6082	0.6089	0.6064	0.6061
1.2262	14.0	2002	0.9452	0.5973	0.5972	0.5978	0.5961	0.5958
0.6694	15.0	2145	1.2849	0.5809	0.5809	0.5817	0.5802	0.5798
0.2152	16.0	2288	1.9241	0.5752	0.5760	0.5753	0.5755	0.5753
0.043	17.0	2431	2.3196	0.5672	0.5685	0.5673	0.5675	0.5672
0.0074	18.0	2574	2.5393	0.5734	0.5747	0.5736	0.5740	0.5737
0.0015	19.0	2717	2.5970	0.5769	0.5780	0.5772	0.5776	0.5772
0.002	20.0	2860	2.6038	0.5787	0.5794	0.5790	0.5792	0.5788

Framework versions

• Transformers 4.55.0
• Pytorch 2.7.0+cu126
• Datasets 4.0.0
• Tokenizers 0.21.4