deberta_MP_dynamic

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: 0.0168
• Macro F1: 0.4727
• Micro F1: 0.6043
• Macro Precision: 0.7187
• Macro Recall: 0.3780
• Micro Precision: 0.7985
• Micro Recall: 0.4860
• Exact Match Ratio: 0.1168
• Macro Roc Auc: 0.9238

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: 16
• eval_batch_size: 16
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• 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: cosine
• lr_scheduler_warmup_steps: 0.1
• num_epochs: 600

Training results

Training Loss	Epoch	Step	Validation Loss	Macro F1	Micro F1	Macro Precision	Macro Recall	Micro Precision	Micro Recall	Exact Match Ratio	Macro Roc Auc
0.0357	13.1579	500	0.0258	0.0000	0.0000	0.0045	0.0000	0.0189	0.0000	0.0057	0.6533
0.0273	26.3158	1000	0.0210	0.0371	0.1188	0.1858	0.0233	0.8534	0.0638	0.0143	0.7992
0.0247	39.4737	1500	0.0196	0.0819	0.1937	0.4000	0.0522	0.8955	0.1086	0.0306	0.8427
0.0227	52.6316	2000	0.0188	0.1215	0.2689	0.4948	0.0794	0.8786	0.1587	0.0454	0.8638
0.0210	65.7895	2500	0.0178	0.1623	0.3199	0.5564	0.1077	0.8790	0.1955	0.0558	0.8793
0.0202	78.9474	3000	0.0172	0.2124	0.3941	0.6246	0.1489	0.8656	0.2551	0.0757	0.8898
0.0190	92.1053	3500	0.0167	0.2465	0.4484	0.6255	0.1757	0.8485	0.3047	0.0807	0.8969
0.0180	105.2632	4000	0.0168	0.2283	0.4148	0.6575	0.1596	0.8725	0.2721	0.0770	0.9012
0.0168	118.4211	4500	0.0165	0.2929	0.4806	0.6900	0.2143	0.8461	0.3356	0.0889	0.9026
0.0162	131.5789	5000	0.0165	0.2999	0.4973	0.6551	0.2217	0.8266	0.3557	0.0876	0.9059
0.0152	144.7368	5500	0.0163	0.3163	0.4897	0.6524	0.2296	0.8396	0.3456	0.0856	0.9082
0.0147	157.8947	6000	0.0167	0.3460	0.5503	0.6838	0.2689	0.7918	0.4217	0.0912	0.9085
0.0142	171.0526	6500	0.0166	0.3465	0.5168	0.6859	0.2588	0.8239	0.3765	0.0891	0.9107
0.0135	184.2105	7000	0.0166	0.3614	0.5270	0.7102	0.2749	0.8203	0.3882	0.0892	0.9109
0.0129	197.3684	7500	0.0161	0.3506	0.5254	0.7256	0.2600	0.8246	0.3855	0.0877	0.9110
0.0128	210.5263	8000	0.0165	0.3648	0.5367	0.7236	0.2756	0.8168	0.3996	0.0927	0.9128
0.0122	223.6842	8500	0.0161	0.3644	0.5436	0.7168	0.2734	0.8240	0.4056	0.0911	0.9138
0.0116	236.8421	9000	0.0165	0.3697	0.5508	0.6876	0.2817	0.8151	0.4160	0.0893	0.9141
0.0115	250.0	9500	0.0163	0.3962	0.5709	0.7120	0.3049	0.7996	0.4439	0.0914	0.9144
0.0108	263.1579	10000	0.0170	0.4181	0.5795	0.7015	0.3288	0.7812	0.4606	0.0901	0.9152
0.0106	276.3158	10500	0.0170	0.4134	0.5827	0.7131	0.3265	0.7829	0.4641	0.0886	0.9160
0.0103	289.4737	11000	0.0166	0.4160	0.5826	0.7024	0.3269	0.7866	0.4627	0.0909	0.9155
0.0099	302.6316	11500	0.0169	0.3875	0.5526	0.7278	0.2940	0.8180	0.4172	0.0932	0.9145
0.0095	315.7895	12000	0.0168	0.3892	0.5476	0.7462	0.2940	0.8131	0.4128	0.0893	0.9153
0.0093	328.9474	12500	0.0170	0.4017	0.5594	0.7369	0.3042	0.8139	0.4261	0.0954	0.9156
0.0094	342.1053	13000	0.0170	0.4152	0.5597	0.7313	0.3187	0.8032	0.4295	0.0896	0.9159
0.0089	355.2632	13500	0.0170	0.4377	0.5887	0.7235	0.3461	0.7898	0.4692	0.0952	0.9179
0.0090	368.4211	14000	0.0173	0.4290	0.5792	0.7208	0.3343	0.7924	0.4564	0.0927	0.9158
0.0087	381.5789	14500	0.0172	0.4378	0.5847	0.7241	0.3440	0.7901	0.4641	0.0937	0.9161
0.0081	394.7368	15000	0.0169	0.4186	0.5693	0.7369	0.3201	0.8040	0.4407	0.0938	0.9163
0.0081	407.8947	15500	0.0170	0.4349	0.5868	0.7210	0.3411	0.7900	0.4668	0.0924	0.9169
0.0080	421.0526	16000	0.0175	0.4544	0.5936	0.7102	0.3612	0.7842	0.4776	0.0951	0.9170
0.0079	434.2105	16500	0.0175	0.4580	0.5989	0.6976	0.3679	0.7806	0.4859	0.0952	0.9174
0.0077	447.3684	17000	0.0173	0.4487	0.5907	0.7144	0.3527	0.7909	0.4714	0.0962	0.9170
0.0076	460.5263	17500	0.0179	0.4580	0.5988	0.7039	0.3659	0.7815	0.4853	0.0958	0.9171
0.0074	473.6842	18000	0.0176	0.4447	0.5866	0.7201	0.3461	0.7948	0.4648	0.0954	0.9167
0.0073	486.8421	18500	0.0174	0.4468	0.5899	0.7157	0.3494	0.7948	0.4690	0.0952	0.9177
0.0070	500.0	19000	0.0176	0.4511	0.5919	0.7119	0.3552	0.7917	0.4726	0.0974	0.9177
0.0073	513.1579	19500	0.0174	0.4522	0.5915	0.7144	0.3541	0.7922	0.4719	0.0969	0.9176
0.0070	526.3158	20000	0.0176	0.4547	0.5957	0.7105	0.3594	0.7886	0.4786	0.0981	0.9173
0.0070	539.4737	20500	0.0176	0.4531	0.5924	0.7084	0.3567	0.7917	0.4733	0.0982	0.9174
0.0072	552.6316	21000	0.0176	0.4560	0.5958	0.7118	0.3598	0.7905	0.4781	0.0983	0.9174
0.0068	565.7895	21500	0.0177	0.4581	0.5965	0.7081	0.3621	0.7882	0.4799	0.0969	0.9173
0.0070	578.9474	22000	0.0178	0.4607	0.5986	0.7038	0.3655	0.7863	0.4833	0.0978	0.9174
0.0070	592.1053	22500	0.0177	0.4578	0.5963	0.7063	0.3619	0.7882	0.4795	0.0976	0.9173
0.0071	600.0	22800	0.0177	0.4577	0.5964	0.7051	0.3619	0.7885	0.4796	0.0976	0.9173

Framework versions

• Transformers 5.12.1
• Pytorch 2.5.1+cu121
• Datasets 5.0.1
• Tokenizers 0.22.2