Table 3 The loss values of different models for conditional generation task in test dataset respect to training rounds.

From: Screening of multi deep learning-based de novo molecular generation models and their application for specific target molecular generation

Epoch

Model

2

4

6

8

10

Mamba

0.120

0.104

0.095

0.090

0.089

T5MolGe

0.119

0.102

0.091

0.084

0.082

GPT

0.152

0.122

0.108

0.100

0.098

GPT-RoPE

0.119

0.103

0.094

0.088

0.086

GPT-Deep

0.154

0.125

0.112

0.104

0.103

GPT-GEGLU

0.136

0.114

0.101

0.093

0.090

Back to article page