Question about the demo in Interpretable Multi-Step Planing

[GuliGuli-Boom]

这篇工作相当精彩，有一些问题想请教

我尝试运行api_for_multisetp.py

目标SMILES：CC(CC1=CC=C2OCOC2=C1)NCC(O)C1=CC=C(O)C(O)=C1 默认参数：expansion_topk=500，iterations=10

但是并没有得到理想的路径

日志的最后是： [('CC(=O)Oc1cc(C(O)CN(C(=O)c2ccccc2)C(C)Cc2ccc3c(c2)OCO3)ccc1OS(C)(=O)=O', 13.081154823303223, ['LGM|Select Leaving Group with Index 154 and Energy 12.65', 'IT|Initial Energy:24.02', 'RCP|Add Bonds: between 34 and 35 with Bond Type 1.0 and Cost -0.86', 'H number change -1 cost of atom 35: -10.09'], [])] [('CC(=O)Oc1cc(C(O)CN(C(=O)c2ccccc2)C(C)Cc2ccc3c(c2)OCO3)ccc1OC(=O)OCc1ccccc1', 13.151131629943848, ['LGM|Select Leaving Group with Index 104 and Energy 12.71', 'IT|Initial Energy:23.87', 'RCP|Add Bonds: between 34 and 35 with Bond Type 1.0 and Cost -0.63', 'H number change -1 cost of atom 35: -10.09'], [])] root : INFO Final search status | success value | iter: 44.424156188964844 | inf | 500 root : INFO Synthesis path for CC(CC1=CC=C2OCOC2=C1)NCC(O)C1=CC=C(O)C(O)=C1 not found. Please try increasing the number of iterations. None

另一个令我疑惑的地方在于，日志中提到的所有物质，例如： CC(=O)Oc1cc(C(O)CN(C(=O)c2ccccc2)C(C)Cc2ccc3c(c2)OCO3)ccc1OC(=O)OCc1ccccc1 CC(=O)Oc1cc(C(O)CN(C(=O)c2ccccc2)C(C)Cc2ccc3c(c2)OCO3)ccc1OS(C)(=O)=O C=CCOC(=O)N(CC(OC(=O)OCC(Cl)(Cl)Cl)c1ccc(OCOC)c(OCOCc2ccccc2)c1)C(C)Cc1ccc2c(c1)OCO2 都包含目标SMILES的结构

这看起来更像是通过分解更加复杂的结构得到目标SMILES，但是这与我对逆合成规划的理解并不一致，并且demo中的结果也显示目标SMILES会被逐步分解为更加简单的结构。

是我的参数没有正确设置吗？非常感谢您的解答！