ytchx1999 / pyg-ogb-tricks Goto Github PK

DGL-ogbn-proteins/models.py 408 line.

Because each h in the variable layer_out has a different size, the weighted sum function will report an error.

Is the uploaded code version wrong?

Thanks for your leaderboard submission. Currently, the GAT paper is referenced, but there are many more tricks that you have used to achieve a good performance. To be qualified for the leaderboard, we require writing a specialized technical report that describes your approach. Please submit again once this is ready.

Thank you!

好奇为什么给交叉熵加上常数后效果会好？这个有什么理论依据吗？

Traceback (most recent call last):
File "node2vec_proteins.py", line 4, in
from torch_geometric.nn import Node2Vec
File "/home/waqar.ali/.local/lib/python3.8/site-packages/torch_geometric/init.py", line 6, in
import torch_geometric.transforms
File "/home/waqar.ali/.local/lib/python3.8/site-packages/torch_geometric/transforms/init.py", line 38, in
from .delaunay import Delaunay
File "/home/waqar.ali/.local/lib/python3.8/site-packages/torch_geometric/transforms/delaunay.py", line 1, in
import scipy.spatial
File "/home/waqar.ali/.conda/envs/Affan/lib/python3.8/site-packages/scipy/spatial/init.py", line 105, in
from ._kdtree import *
File "/home/waqar.ali/.conda/envs/Affan/lib/python3.8/site-packages/scipy/spatial/_kdtree.py", line 5, in
from ._ckdtree import cKDTree, cKDTreeNode
ImportError: /lib/x86_64-linux-gnu/libstdc++.so.6: version `GLIBCXX_3.4.29' not found (required by /home/waqar.ali/.conda/envs/Affan/lib/python3.8/site-packages/scipy/spatial/_ckdtree.cpython-38-x86_64-linux-gnu.so)