pangolin-guard-base

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

• Loss: 0.1999
• Accuracy: 0.916
• F1: 0.9160
• Precision: 0.9165
• Recall: 0.9159

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: 32
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 2

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1	Precision	Recall
0.9557	0.0355	5	0.7064	0.507	0.3537	0.7510	0.5099
0.5695	0.0709	10	0.5353	0.738	0.7329	0.7558	0.7372
0.4891	0.1064	15	0.6198	0.671	0.6457	0.7437	0.6726
0.4992	0.1418	20	0.4391	0.803	0.7975	0.8375	0.8020
0.3817	0.1773	25	0.3565	0.838	0.8376	0.8410	0.8377
0.3696	0.2128	30	0.4334	0.801	0.7974	0.8259	0.8018
0.4089	0.2482	35	0.3024	0.864	0.8632	0.8715	0.8636
0.3693	0.2837	40	0.3085	0.861	0.8607	0.8636	0.8607
0.3878	0.3191	45	0.3034	0.864	0.8636	0.8672	0.8637
0.3265	0.3546	50	0.2968	0.862	0.8602	0.8794	0.8614
0.348	0.3901	55	0.2663	0.882	0.8818	0.8837	0.8818
0.2369	0.4255	60	0.2684	0.88	0.8795	0.8857	0.8796
0.2555	0.4610	65	0.3063	0.861	0.8604	0.8678	0.8614
0.3215	0.4965	70	0.2620	0.879	0.8781	0.8895	0.8785
0.2521	0.5319	75	0.2532	0.884	0.8840	0.8840	0.8840
0.2939	0.5674	80	0.2879	0.875	0.8737	0.8894	0.8744
0.2491	0.6028	85	0.2917	0.871	0.8704	0.8788	0.8714
0.2554	0.6383	90	0.2685	0.88	0.8789	0.8925	0.8795
0.2865	0.6738	95	0.2633	0.88	0.8796	0.8854	0.8804
0.2562	0.7092	100	0.2398	0.889	0.8880	0.9021	0.8885
0.2324	0.7447	105	0.2068	0.91	0.9099	0.9112	0.9098
0.2391	0.7801	110	0.2184	0.901	0.9009	0.9027	0.9008
0.2033	0.8156	115	0.2198	0.893	0.8929	0.8942	0.8928
0.2586	0.8511	120	0.2055	0.908	0.9079	0.9096	0.9078
0.2612	0.8865	125	0.2140	0.912	0.9120	0.9125	0.9121
0.1926	0.9220	130	0.2478	0.885	0.8836	0.9025	0.8844
0.2383	0.9574	135	0.2345	0.903	0.9029	0.9056	0.9032
0.2876	0.9929	140	0.2399	0.881	0.8797	0.8962	0.8804
0.2427	1.0284	145	0.2249	0.901	0.9009	0.9019	0.9009
0.2552	1.0638	150	0.2659	0.883	0.8826	0.8898	0.8834
0.1926	1.0993	155	0.2553	0.885	0.8835	0.9046	0.8843
0.1991	1.1348	160	0.2033	0.918	0.9180	0.9181	0.9179
0.1366	1.1702	165	0.1991	0.91	0.9100	0.9104	0.9099
0.1614	1.2057	170	0.2287	0.912	0.9117	0.9160	0.9117
0.1377	1.2411	175	0.2702	0.903	0.9029	0.9053	0.9032
0.119	1.2766	180	0.2001	0.916	0.9160	0.9163	0.9159
0.1446	1.3121	185	0.2489	0.898	0.8978	0.9018	0.8983
0.2386	1.3475	190	0.2541	0.902	0.9018	0.9053	0.9023
0.1573	1.3830	195	0.2960	0.882	0.8804	0.9018	0.8813
0.242	1.4184	200	0.2013	0.915	0.9149	0.9163	0.9148
0.1685	1.4539	205	0.2665	0.895	0.8947	0.9000	0.8953
0.1708	1.4894	210	0.1989	0.921	0.9209	0.9228	0.9208
0.1474	1.5248	215	0.1988	0.916	0.9159	0.9181	0.9158
0.1352	1.5603	220	0.2026	0.92	0.9200	0.9200	0.9200
0.111	1.5957	225	0.2200	0.912	0.9120	0.9129	0.9121
0.1404	1.6312	230	0.1968	0.913	0.9129	0.9141	0.9128
0.1236	1.6667	235	0.2025	0.914	0.9138	0.9166	0.9138
0.1532	1.7021	240	0.2201	0.909	0.9090	0.9098	0.9091
0.1586	1.7376	245	0.2300	0.904	0.9039	0.9057	0.9042
0.1292	1.7730	250	0.1980	0.915	0.9149	0.9165	0.9148
0.1608	1.8085	255	0.2010	0.917	0.9169	0.9192	0.9168
0.0933	1.8440	260	0.1997	0.914	0.9139	0.9147	0.9139
0.1497	1.8794	265	0.2026	0.914	0.9140	0.9143	0.9139
0.0911	1.9149	270	0.1999	0.915	0.9150	0.9155	0.9149
0.111	1.9504	275	0.2001	0.914	0.9140	0.9146	0.9139
0.134	1.9858	280	0.1999	0.916	0.9160	0.9165	0.9159

Framework versions

• Transformers 4.55.4
• Pytorch 2.8.0+cu126
• Datasets 4.0.0
• Tokenizers 0.21.4