sbuedenb/beetle-gpn-wide-reduced

wide-big-reduced

This model is a GPN trained on the sbuedenb/big_beetle_dataset dataset. It achieves the following results on the evaluation set:

• Loss: 1.1583

Model description

This model deviates from the default GPN in the following parameters:

first_kernel_size=9 (actually default, just to be explicit about kernel size)
rest_kernel_size=9
dilation_max=81
dilation_cycle=5
dilation_base=3
num_hidden_layers=20

--config_overrides "first_kernel_size=9,rest_kernel_size=9,dilation_max=81,dilation_cycle=5,dilation_base=3,num_hidden_layers=20"

Intended uses & limitations

This model is meant for DNA analysis of the Cucujiformia infraorder of insects.

Training and evaluation data

The dataset was created from 12 NCBI reference genomes from Cucujiformia.

Training procedure

240000 steps with linear LR 30000 steps with cosine LR

Training hyperparameters

The following hyperparameters were used for first 240000 steps training:

• learning_rate: 0.001
• train_batch_size: 256
• eval_batch_size: 256
• seed: 42
• distributed_type: multi-GPU
• num_devices: 4
• total_train_batch_size: 1024
• total_eval_batch_size: 1024
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• training_steps: 240000

The following hyperparameters were used for last 30000 steps training:

• learning_rate: 0.001
• train_batch_size: 256
• eval_batch_size: 256
• seed: 42
• distributed_type: multi-GPU
• num_devices: 4
• total_train_batch_size: 1024
• total_eval_batch_size: 1024
• 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
• training_steps: 270000

Training results

Training Loss	Epoch	Step	Validation Loss
1.2423	0.0083	1000	1.2538
1.2322	0.0167	2000	1.2448
1.2251	0.025	3000	1.2364
1.2212	0.0333	4000	1.2350
1.218	0.0417	5000	1.2322
1.2153	0.05	6000	1.2298
1.2134	0.0583	7000	1.2292
1.209	0.0667	8000	1.2252
1.2051	0.075	9000	1.2245
1.2016	0.0833	10000	1.2215
1.1981	0.0917	11000	1.2191
1.1953	0.1	12000	1.2172
1.1918	0.1083	13000	1.2158
1.1897	0.1167	14000	1.2146
1.1874	0.125	15000	1.2132
1.1841	0.1333	16000	1.2110
1.1817	0.1417	17000	1.2104
1.1798	0.15	18000	1.2098
1.178	0.1583	19000	1.2083
1.1754	0.1667	20000	1.2071
1.1747	0.175	21000	1.2062
1.1729	0.1833	22000	1.2055
1.1714	0.1917	23000	1.2050
1.1696	0.2	24000	1.2042
1.1682	0.2083	25000	1.2026
1.1668	0.2167	26000	1.2026
1.165	0.225	27000	1.2007
1.1645	0.2333	28000	1.2004
1.1638	0.2417	29000	1.2001
1.1613	0.25	30000	1.2006
1.1615	0.2583	31000	1.1983
1.1606	0.2667	32000	1.1979
1.1598	0.275	33000	1.1991
1.1585	0.2833	34000	1.1971
1.1575	0.2917	35000	1.1976
1.1563	0.3	36000	1.1964
1.1569	0.3083	37000	1.1954
1.1553	0.3167	38000	1.1953
1.1545	0.325	39000	1.1946
1.1537	0.3333	40000	1.1941
1.1523	0.3417	41000	1.1930
1.1522	0.35	42000	1.1930
1.1508	0.3583	43000	1.1926
1.15	0.3667	44000	1.1912
1.1489	0.375	45000	1.1920
1.1482	0.3833	46000	1.1915
1.1486	0.3917	47000	1.1917
1.1467	0.4	48000	1.1895
1.1474	0.4083	49000	1.1915
1.1463	0.4167	50000	1.1894
1.1457	0.425	51000	1.1898
1.1453	0.4333	52000	1.1888
1.1441	0.4417	53000	1.1887
1.1437	0.45	54000	1.1863
1.1425	0.4583	55000	1.1880
1.143	0.4667	56000	1.1873
1.1415	0.475	57000	1.1869
1.1419	0.4833	58000	1.1865
1.1415	0.4917	59000	1.1876
1.1403	0.5	60000	1.1859
1.1394	0.5083	61000	1.1855
1.14	0.5167	62000	1.1870
1.1388	0.525	63000	1.1859
1.1387	0.5333	64000	1.1852
1.1365	0.5417	65000	1.1840
1.1376	0.55	66000	1.1848
1.1369	0.5583	67000	1.1851
1.1362	0.5667	68000	1.1845
1.1364	0.575	69000	1.1835
1.1365	0.5833	70000	1.1833
1.1347	0.5917	71000	1.1832
1.1355	0.6	72000	1.1831
1.1342	0.6083	73000	1.1824
1.134	0.6167	74000	1.1824
1.1339	0.625	75000	1.1819
1.1323	0.6333	76000	1.1820
1.1322	0.6417	77000	1.1816
1.1325	0.65	78000	1.1820
1.1304	0.6583	79000	1.1810
1.1313	0.6667	80000	1.1811
1.1305	0.675	81000	1.1814
1.1309	0.6833	82000	1.1813
1.1294	0.6917	83000	1.1796
1.1299	0.7	84000	1.1816
1.1299	0.7083	85000	1.1796
1.1292	0.7167	86000	1.1793
1.1287	0.725	87000	1.1800
1.1278	0.7333	88000	1.1796
1.1277	0.7417	89000	1.1786
1.1265	0.75	90000	1.1796
1.128	0.7583	91000	1.1787
1.1266	0.7667	92000	1.1787
1.1271	0.775	93000	1.1789
1.1274	0.7833	94000	1.1784
1.1254	0.7917	95000	1.1779
1.1255	0.8	96000	1.1781
1.1252	0.8083	97000	1.1780
1.1255	0.8167	98000	1.1779
1.1249	0.825	99000	1.1776
1.1231	0.8333	100000	1.1767
1.1254	0.8417	101000	1.1770
1.1239	0.85	102000	1.1779
1.1227	0.8583	103000	1.1780
1.1233	0.8667	104000	1.1761
1.1242	0.875	105000	1.1768
1.1226	0.8833	106000	1.1760
1.1228	0.8917	107000	1.1763
1.1223	0.9	108000	1.1757
1.1224	0.9083	109000	1.1763
1.1221	0.9167	110000	1.1760
1.1209	0.925	111000	1.1760
1.1204	0.9333	112000	1.1758
1.1209	0.9417	113000	1.1753
1.1199	0.95	114000	1.1758
1.1203	0.9583	115000	1.1755
1.1196	0.9667	116000	1.1752
1.1207	0.975	117000	1.1754
1.1193	0.9833	118000	1.1747
1.1196	0.9917	119000	1.1748
1.1198	1.0	120000	1.1752
1.1184	0.0056	121000	1.1748
1.1198	0.0111	122000	1.1729
1.1183	0.0167	123000	1.1739
1.1186	0.0222	124000	1.1744
1.1185	0.0278	125000	1.1740
1.1168	0.0333	126000	1.1742
1.1184	0.0389	127000	1.1743
1.1166	0.0444	128000	1.1732
1.1161	0.05	129000	1.1726
1.1162	0.0556	130000	1.1730
1.1161	0.0611	131000	1.1746
1.1178	0.0667	132000	1.1736
1.115	0.0722	133000	1.1742
1.1157	0.0778	134000	1.1734
1.1159	0.0833	135000	1.1723
1.1153	0.0889	136000	1.1725
1.1154	0.0944	137000	1.1736
1.1143	0.1	138000	1.1717
1.1145	0.1056	139000	1.1725
1.1138	0.1111	140000	1.1728
1.1136	0.1167	141000	1.1721
1.1143	0.1222	142000	1.1719
1.1143	0.1278	143000	1.1721
1.1145	0.1333	144000	1.1707
1.1133	0.1389	145000	1.1729
1.1127	0.1444	146000	1.1715
1.1125	0.15	147000	1.1714
1.1126	0.1556	148000	1.1722
1.1132	0.1611	149000	1.1710
1.1114	0.1667	150000	1.1703
1.1134	0.1722	151000	1.1713
1.112	0.1778	152000	1.1713
1.1124	0.1833	153000	1.1717
1.112	0.1889	154000	1.1707
1.1131	0.1944	155000	1.1713
1.1114	0.2	156000	1.1699
1.1132	0.2056	157000	1.1713
1.1116	0.2111	158000	1.1712
1.1121	0.2167	159000	1.1705
1.1117	0.2222	160000	1.1699
1.1099	0.2278	161000	1.1693
1.1108	0.2333	162000	1.1702
1.1102	0.2389	163000	1.1699
1.1099	0.2444	164000	1.1697
1.1098	0.25	165000	1.1700
1.1091	0.2556	166000	1.1698
1.1117	0.2611	167000	1.1700
1.1094	0.2667	168000	1.1700
1.1099	0.2722	169000	1.1691
1.1098	0.2778	170000	1.1690
1.1097	0.2833	171000	1.1709
1.11	0.2889	172000	1.1707
1.1095	0.2944	173000	1.1690
1.108	0.3	174000	1.1687
1.108	0.3056	175000	1.1696
1.1087	0.3111	176000	1.1688
1.1074	0.3167	177000	1.1689
1.1085	0.3222	178000	1.1699
1.1089	0.3278	179000	1.1697
1.1082	0.3333	180000	1.1691
1.1073	0.0042	181000	1.1684
1.1087	0.0083	182000	1.1677
1.1075	0.0125	183000	1.1673
1.108	0.0167	184000	1.1678
1.1082	0.0208	185000	1.1687
1.1063	0.025	186000	1.1689
1.1073	0.0292	187000	1.1678
1.1073	0.0333	188000	1.1675
1.1062	0.0375	189000	1.1678
1.1062	0.0417	190000	1.1686
1.1054	0.0458	191000	1.1679
1.1079	0.05	192000	1.1679
1.1061	0.0542	193000	1.1680
1.106	0.0583	194000	1.1685
1.1061	0.0625	195000	1.1672
1.1064	0.0667	196000	1.1672
1.1064	0.0708	197000	1.1673
1.1047	0.075	198000	1.1671
1.1055	0.0792	199000	1.1674
1.1049	0.0833	200000	1.1678
1.1053	0.0875	201000	1.1675
1.1046	0.0917	202000	1.1666
1.1051	0.0958	203000	1.1672
1.1058	0.1	204000	1.1659
1.1044	0.1042	205000	1.1667
1.1035	0.1083	206000	1.1677
1.1038	0.1125	207000	1.1671
1.1042	0.1167	208000	1.1657
1.1044	0.1208	209000	1.1664
1.1028	0.125	210000	1.1663
1.1046	0.1292	211000	1.1654
1.1031	0.1333	212000	1.1662
1.104	0.1375	213000	1.1673
1.1045	0.1417	214000	1.1666
1.1045	0.1458	215000	1.1665
1.1027	0.15	216000	1.1648
1.1047	0.1542	217000	1.1660
1.1035	0.1583	218000	1.1659
1.1038	0.1625	219000	1.1672
1.1037	0.1667	220000	1.1665
1.1026	0.1708	221000	1.1653
1.1035	0.175	222000	1.1655
1.1033	0.1792	223000	1.1648
1.1018	0.1833	224000	1.1650
1.1024	0.1875	225000	1.1654
1.1017	0.1917	226000	1.1651
1.1036	0.1958	227000	1.1655
1.1021	0.2	228000	1.1667
1.102	0.2042	229000	1.1660
1.1029	0.2083	230000	1.1664
1.102	0.2125	231000	1.1661
1.1026	0.2167	232000	1.1651
1.1019	0.2208	233000	1.1644
1.1014	0.225	234000	1.1651
1.1011	0.2292	235000	1.1658
1.1018	0.2333	236000	1.1658
1.1008	0.2375	237000	1.1653
1.1018	0.2417	238000	1.1651
1.1025	0.2458	239000	1.1652
1.1008	0.25	240000	1.1654
1.0963	0.0037	241000	1.1616
1.0963	0.0074	242000	1.1611
1.0926	0.0111	243000	1.1597
1.0941	0.0148	244000	1.1594
1.094	0.0185	245000	1.1591
1.0917	0.0222	246000	1.1596
1.0915	0.0259	247000	1.1591
1.0922	0.0296	248000	1.1589
1.0913	0.0333	249000	1.1586
1.091	0.0370	250000	1.1584
1.0896	0.0407	251000	1.1583
1.0931	0.0444	252000	1.1586
1.0914	0.0481	253000	1.1582
1.09	0.0518	254000	1.1592
1.0907	0.0556	255000	1.1584
1.0898	0.0593	256000	1.1585
1.0917	0.0630	257000	1.1577
1.0898	0.0667	258000	1.1583
1.0894	0.0704	259000	1.1576
1.0888	0.0741	260000	1.1591
1.0906	0.0778	261000	1.1578
1.0895	0.0815	262000	1.1574
1.0902	0.0852	263000	1.1580
1.0906	0.0889	264000	1.1579
1.088	0.0926	265000	1.1584
1.0896	0.0963	266000	1.1584
1.0894	0.1000	267000	1.1581
1.0895	0.1037	268000	1.1577
1.0894	0.1074	269000	1.1574
1.0891	0.1111	270000	1.1573

Framework versions

• Transformers 4.51.3
• Pytorch 2.7.0+cu126
• Datasets 3.6.0.dev0
• Tokenizers 0.21.1