Modelpublic

orpe42/deberta_MP_dynamic

sourceHugging Facemitupdated 28d agoView on Hugging Face

0likes237downloads

Model Card

debertaMPdynamic

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: 0.0460
—Macro F1: 0.5491
—Micro F1: 0.6332
—Macro Precision: 0.6797
—Macro Recall: 0.4702
—Micro Precision: 0.7566
—Micro Recall: 0.5443
—Exact Match Ratio: 0.0853
—Macro Roc Auc: 0.9217

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
—trainbatchsize: 16
—evalbatchsize: 16
—seed: 42
—gradientaccumulationsteps: 2
—totaltrainbatch_size: 32
—optimizer: Use OptimizerNames.ADAMWTORCHFUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
—lrschedulertype: cosine
—lrschedulerwarmup_steps: 0.1
—num_epochs: 1000

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
0.1277	12.8312	500	0.0921	0.0091	0.1081	0.0156	0.0179	0.3306	0.0646	0.0128	0.5595
0.0672	25.6494	1000	0.0563	0.0003	0.0011	0.0228	0.0002	0.2821	0.0005	0.0105	0.7364
0.0590	38.4675	1500	0.0497	0.0317	0.0880	0.2114	0.0184	0.8787	0.0463	0.0159	0.8042
0.0541	51.2857	2000	0.0467	0.0586	0.1344	0.4023	0.0351	0.8968	0.0727	0.0201	0.8336
0.0503	64.1039	2500	0.0431	0.1210	0.2378	0.4635	0.0770	0.9086	0.1368	0.0434	0.8610
0.0485	76.9351	3000	0.0411	0.1879	0.3432	0.5437	0.1270	0.8781	0.2133	0.0552	0.8747
0.0453	89.7532	3500	0.0399	0.2265	0.3507	0.6134	0.1561	0.8861	0.2186	0.0607	0.8866
0.0428	102.5714	4000	0.0389	0.2439	0.4006	0.6789	0.1669	0.8647	0.2607	0.0613	0.8924
0.0416	115.3896	4500	0.0382	0.2945	0.4504	0.6668	0.2131	0.8446	0.3071	0.0648	0.8994
0.0403	128.2078	5000	0.0378	0.2957	0.4441	0.7214	0.2092	0.8545	0.3000	0.0640	0.9031
0.0385	141.0260	5500	0.0374	0.3255	0.4808	0.7189	0.2375	0.8413	0.3366	0.0708	0.9070
0.0364	153.8571	6000	0.0373	0.3372	0.4788	0.6982	0.2455	0.8504	0.3332	0.0727	0.9100
0.0358	166.6753	6500	0.0378	0.3380	0.4618	0.7820	0.2453	0.8533	0.3166	0.0722	0.9109
0.0341	179.4935	7000	0.0370	0.3645	0.4935	0.7500	0.2676	0.8496	0.3478	0.0741	0.9134
0.0326	192.3117	7500	0.0370	0.4061	0.5497	0.7122	0.3129	0.8121	0.4155	0.0756	0.9142
0.0319	205.1299	8000	0.0375	0.3608	0.4971	0.7664	0.2623	0.8473	0.3518	0.0700	0.9157
0.0304	217.9610	8500	0.0370	0.4163	0.5357	0.7252	0.3145	0.8237	0.3969	0.0768	0.9170
0.0292	230.7792	9000	0.0372	0.4394	0.5623	0.7222	0.3432	0.8028	0.4326	0.0767	0.9181
0.0286	243.5974	9500	0.0372	0.4332	0.5652	0.7344	0.3332	0.8016	0.4364	0.0759	0.9190
0.0283	256.4156	10000	0.0369	0.4454	0.5773	0.7243	0.3506	0.7932	0.4538	0.0786	0.9192
0.0270	269.2338	10500	0.0380	0.4462	0.5684	0.7241	0.3516	0.8103	0.4377	0.0835	0.9197
0.0262	282.0519	11000	0.0376	0.4597	0.5714	0.7021	0.3642	0.7994	0.4446	0.0804	0.9193
0.0253	294.8831	11500	0.0381	0.4566	0.5690	0.7149	0.3629	0.8004	0.4413	0.0788	0.9198
0.0248	307.7013	12000	0.0373	0.4716	0.5719	0.7170	0.3711	0.8071	0.4428	0.0805	0.9212
0.0235	320.5195	12500	0.0380	0.4580	0.5696	0.7129	0.3608	0.8021	0.4416	0.0800	0.9207
0.0230	333.3377	13000	0.0380	0.4732	0.5796	0.7057	0.3743	0.7990	0.4547	0.0827	0.9212
0.0228	346.1558	13500	0.0389	0.4825	0.5891	0.7074	0.3882	0.7885	0.4702	0.0823	0.9228
0.0219	358.9870	14000	0.0390	0.4777	0.5784	0.7050	0.3808	0.7981	0.4535	0.0827	0.9208
0.0211	371.8052	14500	0.0393	0.5023	0.5989	0.6941	0.4141	0.7777	0.4869	0.0841	0.9220
0.0207	384.6234	15000	0.0389	0.4754	0.5739	0.7123	0.3749	0.8047	0.4460	0.0832	0.9220
0.0200	397.4416	15500	0.0398	0.4993	0.6024	0.6898	0.4123	0.7781	0.4915	0.0832	0.9223
0.0198	410.2597	16000	0.0403	0.5005	0.5996	0.6980	0.4098	0.7842	0.4854	0.0826	0.9220
0.0188	423.0779	16500	0.0402	0.4989	0.5954	0.6914	0.4096	0.7885	0.4783	0.0859	0.9214
0.0187	435.9091	17000	0.0392	0.4915	0.5868	0.7066	0.3938	0.7973	0.4642	0.0828	0.9215
0.0181	448.7273	17500	0.0407	0.5164	0.6169	0.6801	0.4354	0.7694	0.5149	0.0861	0.9225
0.0177	461.5455	18000	0.0405	0.5083	0.6147	0.6914	0.4208	0.7728	0.5102	0.0859	0.9230
0.0177	474.3636	18500	0.0407	0.5028	0.5945	0.6982	0.4100	0.7928	0.4755	0.0850	0.9217
0.0167	487.1818	19000	0.0421	0.5046	0.5969	0.6934	0.4155	0.7861	0.4811	0.0843	0.9214
0.0164	500.0	19500	0.0437	0.5301	0.6169	0.6628	0.4552	0.7582	0.5201	0.0832	0.9229
0.0157	512.8312	20000	0.0396	0.5049	0.5967	0.6950	0.4112	0.7897	0.4795	0.0872	0.9209
0.0150	525.6494	20500	0.0414	0.5114	0.6051	0.6891	0.4214	0.7791	0.4946	0.0845	0.9224
0.0155	538.4675	21000	0.0411	0.5140	0.6055	0.6952	0.4206	0.7819	0.4940	0.0865	0.9214
0.0154	551.2857	21500	0.0413	0.5216	0.6159	0.6867	0.4347	0.7736	0.5116	0.0887	0.9220
0.0141	564.1039	22000	0.0418	0.5253	0.6149	0.6914	0.4386	0.7743	0.5100	0.0905	0.9222
0.0142	576.9351	22500	0.0431	0.5272	0.6267	0.6795	0.4470	0.7609	0.5328	0.0891	0.9215
0.0138	589.7532	23000	0.0420	0.5185	0.6045	0.6966	0.4271	0.7862	0.4910	0.0890	0.9210
0.0141	602.5714	23500	0.0411	0.5141	0.6043	0.6975	0.4188	0.7877	0.4902	0.0865	0.9201
0.0138	615.3896	24000	0.0433	0.5247	0.6192	0.6851	0.4405	0.7751	0.5156	0.0886	0.9215
0.0131	628.2078	24500	0.0437	0.5319	0.6213	0.6824	0.4505	0.7681	0.5216	0.0878	0.9208
0.0126	641.0260	25000	0.0439	0.5319	0.6219	0.6819	0.4481	0.7715	0.5208	0.0897	0.9211
0.0128	653.8571	25500	0.0433	0.5307	0.6164	0.6887	0.4431	0.7782	0.5103	0.0897	0.9216
0.0123	666.6753	26000	0.0435	0.5227	0.6141	0.6925	0.4346	0.7823	0.5055	0.0914	0.9211
0.0124	679.4935	26500	0.0443	0.5349	0.6286	0.6809	0.4527	0.7667	0.5326	0.0887	0.9207
0.0123	692.3117	27000	0.0441	0.5356	0.6235	0.6787	0.4549	0.7683	0.5247	0.0911	0.9216
0.0118	705.1299	27500	0.0434	0.5217	0.6114	0.6959	0.4297	0.7850	0.5007	0.0891	0.9197
0.0116	717.9610	28000	0.0449	0.5358	0.6243	0.6733	0.4564	0.7662	0.5267	0.0891	0.9211
0.0118	730.7792	28500	0.0444	0.5324	0.6245	0.6849	0.4477	0.7745	0.5231	0.0919	0.9216
0.0112	743.5974	29000	0.0454	0.5323	0.6196	0.6862	0.4474	0.7767	0.5153	0.0919	0.9204
0.0110	756.4156	29500	0.0456	0.5375	0.6291	0.6823	0.4570	0.7698	0.5318	0.0942	0.9213
0.0107	769.2338	30000	0.0452	0.5349	0.6259	0.6827	0.4512	0.7692	0.5276	0.0922	0.9209
0.0105	782.0519	30500	0.0453	0.5369	0.6267	0.6767	0.4560	0.7688	0.5290	0.0925	0.9216
0.0105	794.8831	31000	0.0455	0.5398	0.6309	0.6764	0.4600	0.7655	0.5366	0.0925	0.9211
0.0106	807.7013	31500	0.0458	0.5431	0.6304	0.6785	0.4641	0.7670	0.5351	0.0948	0.9210
0.0103	820.5195	32000	0.0461	0.5415	0.6285	0.6712	0.4638	0.7655	0.5331	0.0937	0.9210
0.0103	833.3377	32500	0.0456	0.5416	0.6287	0.6856	0.4588	0.7710	0.5308	0.0957	0.9211
0.0103	846.1558	33000	0.0462	0.5448	0.6335	0.6743	0.4675	0.7649	0.5406	0.0920	0.9211
0.0099	858.9870	33500	0.0458	0.5397	0.6263	0.6832	0.4566	0.7738	0.5261	0.0966	0.9213
0.0097	871.8052	34000	0.0467	0.5434	0.6320	0.6747	0.4669	0.7640	0.5389	0.0952	0.9212
0.0100	884.6234	34500	0.0460	0.5446	0.6318	0.6777	0.4656	0.7657	0.5377	0.0945	0.9211
0.0098	897.4416	35000	0.0462	0.5421	0.6308	0.6771	0.4627	0.7675	0.5354	0.0947	0.9211
0.0099	910.2597	35500	0.0462	0.5412	0.6297	0.6792	0.4604	0.7674	0.5339	0.0948	0.9210
0.0099	923.0779	36000	0.0459	0.5398	0.6285	0.6812	0.4580	0.7707	0.5307	0.0959	0.9208
0.0096	935.9091	36500	0.0463	0.5427	0.6304	0.6780	0.4628	0.7679	0.5346	0.0962	0.9210
0.0098	948.7273	37000	0.0464	0.5441	0.6308	0.6782	0.4649	0.7674	0.5355	0.0948	0.9210
0.0099	961.5455	37500	0.0461	0.5425	0.6299	0.6787	0.4626	0.7688	0.5335	0.0957	0.9209
0.0096	974.3636	38000	0.0463	0.5433	0.6305	0.6782	0.4639	0.7678	0.5349	0.0957	0.9211
0.0097	987.1818	38500	0.0462	0.5436	0.6309	0.6784	0.4642	0.7674	0.5356	0.0954	0.9211
0.0097	1000.0	39000	0.0462	0.5432	0.6305	0.6783	0.4637	0.7676	0.5350	0.0956	0.9211

Framework versions

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