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: 0.1315
• Macro F1: 0.6518
• Micro F1: 0.6962
• Macro Precision: 0.6229
• Macro Recall: 0.6869
• Micro Precision: 0.6614
• Micro Recall: 0.7348
• Exact Match Ratio: 0.0764
• Macro Roc Auc: 0.9035

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-05
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 4
• 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: 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
0.1612	12.8312	500	0.1435	0.2058	0.3965	0.3632	0.1788	0.5793	0.3014	0.0187	0.7830
0.1494	25.6494	1000	0.1391	0.4377	0.5473	0.4825	0.4386	0.5611	0.5341	0.0284	0.8402
0.1443	38.4675	1500	0.1366	0.4759	0.5803	0.5350	0.4824	0.5938	0.5674	0.0315	0.8641
0.1404	51.2857	2000	0.1334	0.5131	0.6173	0.5879	0.4986	0.6492	0.5884	0.0518	0.8832
0.1376	64.1039	2500	0.1319	0.5475	0.6361	0.5572	0.5796	0.5942	0.6843	0.0328	0.8915
0.1357	76.9351	3000	0.1314	0.5660	0.6440	0.5519	0.6105	0.6023	0.6918	0.0337	0.8963
0.1328	89.7532	3500	0.1298	0.5918	0.6626	0.5663	0.6415	0.6128	0.7212	0.0395	0.9035
0.1311	102.5714	4000	0.1303	0.5968	0.6590	0.5597	0.6617	0.5945	0.7394	0.0365	0.9051
0.1300	115.3896	4500	0.1315	0.5922	0.6528	0.5358	0.6861	0.5751	0.7549	0.0296	0.9083
0.1282	128.2078	5000	0.1306	0.6106	0.6689	0.5691	0.6768	0.6095	0.7409	0.0435	0.9077
0.1265	141.0260	5500	0.1292	0.6195	0.6787	0.5987	0.6552	0.6361	0.7274	0.0520	0.9085
0.1248	153.8571	6000	0.1297	0.6180	0.6755	0.5711	0.6894	0.6052	0.7644	0.0407	0.9113
0.1238	166.6753	6500	0.1302	0.6154	0.6782	0.5973	0.6578	0.6469	0.7127	0.0539	0.9109
0.1230	179.4935	7000	0.1296	0.6288	0.6838	0.5932	0.6778	0.6421	0.7312	0.0584	0.9075
0.1221	192.3117	7500	0.1299	0.6275	0.6830	0.5998	0.6745	0.6389	0.7337	0.0582	0.9077
0.1205	205.1299	8000	0.1299	0.6316	0.6839	0.6201	0.6586	0.6585	0.7114	0.0594	0.9066
0.1201	217.9610	8500	0.1304	0.6306	0.6809	0.5783	0.7029	0.6141	0.7639	0.0447	0.9086
0.1189	230.7792	9000	0.1303	0.6330	0.6851	0.5977	0.6803	0.6427	0.7335	0.0541	0.9086
0.1184	243.5974	9500	0.1316	0.6303	0.6822	0.5795	0.7018	0.6196	0.7588	0.0451	0.9073
0.1184	256.4156	10000	0.1301	0.6362	0.6855	0.5955	0.6933	0.6377	0.7410	0.0545	0.9089
0.1180	269.2338	10500	0.1306	0.6373	0.6873	0.5997	0.6902	0.6405	0.7415	0.0557	0.9056
0.1175	282.0519	11000	0.1305	0.6390	0.6854	0.5882	0.7044	0.6273	0.7553	0.0479	0.9106
0.1171	294.8831	11500	0.1300	0.6392	0.6908	0.6201	0.6708	0.6598	0.7250	0.0636	0.9062
0.1169	307.7013	12000	0.1304	0.6398	0.6876	0.5901	0.7043	0.6285	0.7591	0.0512	0.9104
0.1154	320.5195	12500	0.1305	0.6388	0.6898	0.6170	0.6742	0.6647	0.7168	0.0690	0.9059
0.1151	333.3377	13000	0.1306	0.6392	0.6867	0.5990	0.6946	0.6395	0.7414	0.0573	0.9092
0.1156	346.1558	13500	0.1312	0.6369	0.6882	0.6061	0.6851	0.6482	0.7334	0.0609	0.9078
0.1147	358.9870	14000	0.1313	0.6378	0.6882	0.5958	0.6942	0.6426	0.7409	0.0596	0.9089
0.1140	371.8052	14500	0.1311	0.6397	0.6896	0.6088	0.6822	0.6550	0.7279	0.0651	0.9066
0.1138	384.6234	15000	0.1319	0.6412	0.6891	0.5995	0.6953	0.6434	0.7417	0.0581	0.9080
0.1136	397.4416	15500	0.1307	0.6428	0.6906	0.6087	0.6887	0.6462	0.7415	0.0622	0.9068
0.1131	410.2597	16000	0.1318	0.6399	0.6890	0.5955	0.6970	0.6421	0.7433	0.0631	0.9067
0.1129	423.0779	16500	0.1317	0.6407	0.6904	0.6019	0.6954	0.6497	0.7366	0.0634	0.9086
0.1126	435.9091	17000	0.1304	0.6415	0.6920	0.6218	0.6707	0.6677	0.7181	0.0683	0.9064
0.1123	448.7273	17500	0.1317	0.6412	0.6895	0.5953	0.6999	0.6399	0.7475	0.0593	0.9070
0.1121	461.5455	18000	0.1326	0.6414	0.6904	0.6008	0.6966	0.6467	0.7404	0.0609	0.9063
0.1122	474.3636	18500	0.1322	0.6414	0.6893	0.6037	0.6917	0.6451	0.7401	0.0607	0.9046
0.1114	487.1818	19000	0.1313	0.6419	0.6922	0.6065	0.6880	0.6523	0.7372	0.0644	0.9076
0.1110	500.0	19500	0.1310	0.6430	0.6938	0.6304	0.6629	0.6796	0.7086	0.0772	0.9034
0.1110	512.8312	20000	0.1318	0.6436	0.6930	0.6151	0.6836	0.6614	0.7278	0.0675	0.9045
0.1107	525.6494	20500	0.1313	0.6413	0.6919	0.6100	0.6843	0.6545	0.7337	0.0669	0.9056
0.1107	538.4675	21000	0.1318	0.6461	0.6936	0.6102	0.6923	0.6549	0.7370	0.0657	0.9051
0.1105	551.2857	21500	0.1321	0.6410	0.6896	0.5957	0.7011	0.6399	0.7478	0.0589	0.9059
0.1099	564.1039	22000	0.1327	0.6458	0.6926	0.6015	0.7041	0.6440	0.7491	0.0598	0.9052
0.1100	576.9351	22500	0.1325	0.6437	0.6922	0.5991	0.7011	0.6438	0.7485	0.0626	0.9061
0.1095	589.7532	23000	0.1325	0.6428	0.6930	0.6055	0.6924	0.6533	0.7378	0.0643	0.9053
0.1098	602.5714	23500	0.1313	0.6470	0.6956	0.6209	0.6807	0.6654	0.7288	0.0717	0.9035
0.1097	615.3896	24000	0.1324	0.6482	0.6959	0.6072	0.6981	0.6530	0.7449	0.0671	0.9055
0.1091	628.2078	24500	0.1324	0.6468	0.6933	0.6113	0.6933	0.6532	0.7386	0.0662	0.9035
0.1091	641.0260	25000	0.1322	0.6464	0.6944	0.6059	0.6966	0.6508	0.7442	0.0692	0.9044
0.1088	653.8571	25500	0.1319	0.6517	0.6977	0.6230	0.6863	0.6682	0.7299	0.0750	0.9036
0.1086	666.6753	26000	0.1315	0.6478	0.6970	0.6229	0.6807	0.6678	0.7290	0.0739	0.9033
0.1087	679.4935	26500	0.1328	0.6493	0.6945	0.6060	0.7041	0.6459	0.7510	0.0636	0.9055
0.1086	692.3117	27000	0.1329	0.6450	0.6933	0.6032	0.6972	0.6472	0.7464	0.0655	0.9020
0.1085	705.1299	27500	0.1328	0.6452	0.6938	0.6142	0.6871	0.6594	0.7321	0.0714	0.9024
0.1081	717.9610	28000	0.1326	0.6491	0.6960	0.6132	0.6933	0.6596	0.7366	0.0709	0.9027
0.1083	730.7792	28500	0.1325	0.6484	0.6951	0.6119	0.6933	0.6552	0.7403	0.0695	0.9044
0.1078	743.5974	29000	0.1326	0.6487	0.6955	0.6167	0.6884	0.6605	0.7344	0.0700	0.9032
0.1080	756.4156	29500	0.1322	0.6467	0.6948	0.6169	0.6845	0.6591	0.7346	0.0733	0.9018
0.1078	769.2338	30000	0.1325	0.6472	0.6948	0.6095	0.6933	0.6564	0.7380	0.0696	0.9034
0.1075	782.0519	30500	0.1326	0.6484	0.6962	0.6102	0.6944	0.6545	0.7436	0.0696	0.9019
0.1078	794.8831	31000	0.1327	0.6497	0.6962	0.6132	0.6941	0.6570	0.7404	0.0705	0.9035
0.1076	807.7013	31500	0.1328	0.6484	0.6958	0.6128	0.6925	0.6577	0.7386	0.0696	0.9050
0.1074	820.5195	32000	0.1328	0.6475	0.6958	0.6098	0.6942	0.6574	0.7390	0.0722	0.9041
0.1074	833.3377	32500	0.1321	0.6516	0.6982	0.6202	0.6906	0.6636	0.7367	0.0739	0.9040
0.1074	846.1558	33000	0.1326	0.6486	0.6967	0.6117	0.6946	0.6572	0.7413	0.0714	0.9037
0.1072	858.9870	33500	0.1321	0.6503	0.6975	0.6215	0.6855	0.6669	0.7310	0.0755	0.9031
0.1070	871.8052	34000	0.1327	0.6488	0.6961	0.6117	0.6946	0.6576	0.7394	0.0714	0.9043
0.1072	884.6234	34500	0.1328	0.6501	0.6969	0.6118	0.6965	0.6571	0.7417	0.0722	0.9033
0.1071	897.4416	35000	0.1326	0.6507	0.6974	0.6137	0.6957	0.6593	0.7403	0.0726	0.9034
0.1071	910.2597	35500	0.1323	0.6517	0.6984	0.6225	0.6876	0.6666	0.7333	0.0751	0.9027
0.1071	923.0779	36000	0.1326	0.6513	0.6979	0.6154	0.6949	0.6599	0.7406	0.0722	0.9035
0.1072	935.9091	36500	0.1326	0.6511	0.6976	0.6150	0.6950	0.6598	0.7400	0.0726	0.9034
0.1070	948.7273	37000	0.1325	0.6513	0.6980	0.6170	0.6929	0.6621	0.7381	0.0744	0.9033
0.1071	961.5455	37500	0.1326	0.6509	0.6975	0.6147	0.6950	0.6596	0.7400	0.0722	0.9037
0.1068	974.3636	38000	0.1325	0.6513	0.6980	0.6164	0.6938	0.6615	0.7388	0.0733	0.9035
0.1072	987.1818	38500	0.1325	0.6511	0.6980	0.6154	0.6946	0.6608	0.7396	0.0736	0.9036
0.1071	1000.0	39000	0.1325	0.6513	0.6979	0.6157	0.6945	0.6608	0.7395	0.0732	0.9035

Framework versions

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