Skip to main content

Table 2 Dataset description

From: Scalable training of graph convolutional neural networks for fast and accurate predictions of HOMO-LUMO gap in molecules

Dataset

Graph size

Data file size (GB)

#Graphs

#Nodes

#Edges

#Avg\(^1\)

CSV

Pickle

ADIOS

PCQM4Mv2

3.6 M

105.8 M

214.6 M

29.4

0.16

34

22

AISD HOMO-LUMO

10.5 M

550.6 M

1.1 B

52.4

0.88

94

60

  1. 1Average number of nodes (atoms) per graph