Neural Subgraph Isomorphism Counting复现记录

train.py：
与输入参数如下：

--model
RGIN
--predict_net
DIAMNet
--predict_net_mem_init
mean
--predict_net_mem_len
4
--predict_net_recurrent_steps
3
--gpu_id
-1
--batch_size
512
--max_npv
8
--max_npe
8
--max_npvl
8
--max_npel
8
--max_ngv
64
--max_nge
256
--max_ngvl
16
--max_ngel
16
--pattern_dir
../data/debug/patterns
--graph_dir
../data/debug/graphs
--metadata_dir
../data/debug/metadata
--save_data_dir
../data/debug
--save_model_dir
../dumps/debug/RGIN-DIAMNet

2020/12/19 16:28:17: [ data_type: train best mean loss: 401.561 (epoch: 050) ]
2020/12/19 16:28:17: [ data_type: dev best mean loss: 236.555 (epoch: 073) ]
2020/12/19 16:28:17: [ data_type: test best mean loss: 267.581 (epoch: 045) ]

evaluate.py: