leffff/graphormer-pyg
Microsoft Graphormer (https://arxiv.org/abs/2106.05234) rewritten in Pytorch-Geometric
PythonMIT
Issues
- 0
Hidden state
#31 opened by feyhong1112 - 0
Any new improvements
#30 opened by Aron-04Wang - 3
softmax in GraphormerAttentionHead
#14 opened by JingweiQu - 1
- 1
Pre-trained
#28 opened by Assassinswhh - 0
Pre-train
#27 opened by Assassinswhh - 2
Time-cost
#22 opened by Lookfor2201 - 3
Activation Functions - GELU?
#23 opened by luke-a-thompson - 1
The code will be very slow.
#24 opened by StefanIsSmart - 1
heterogeneous graphs
#25 opened by Xiaobaijav - 4
Add [VNode]
#19 opened by leffff - 3
exeuction speed
#17 opened by andreYoo - 5
实现出来的效果可以吗
#20 opened by caojiangege - 2
Potential Bug in CentralityEncoding
#16 opened by AjSiciliano - 1
作者能否发一下邮箱,有些问题请教
#21 opened by caojiangege - 0
Add Flash Attention 2.0
#18 opened by leffff - 0
Optimize for loop in SpatialEncoding
#15 opened by leffff - 8
Optimize for loop in SpatialEncoding
#5 opened by leffff - 3
RuntimeError: Cannot re-initialize CUDA in forked subprocess. To use CUDA with multiprocessing, you must use the 'spawn' start method
#10 opened by lizhixi212 - 7
- 1
Difficulty of porting pre-trained Graphormer weights to this torch geometric implementation
#13 opened by seyonechithrananda - 1
Rewrite Edge Encoding as in original paper
#11 opened by leffff - 1
- 1
Optimize for loop in CentralityEncoding
#6 opened by leffff - 1
- 0
Optimize for loop in EdgeEncoding
#4 opened by leffff - 1
- 1