shuix007 / hmgnn Goto Github PK

Hi, when I try to run training with main.py with given parameters, the following error occurs:

terminate called after throwing an instance of 'dmlc::Error'
what(): [09:48:37] /opt/dgl/src/runtime/tensordispatch.cc:23: Check failed: !available_: The tensor adapter can only load once.
Stack trace:
[bt] (0) /opt/conda/lib/python3.7/site-packages/dgl/libdgl.so(dmlc::LogMessageFatal::~LogMessageFatal()+0x4f) [0x7f5cad0ad30f]
[bt] (1) /opt/conda/lib/python3.7/site-packages/dgl/libdgl.so(dgl::runtime::TensorDispatcher::Load(char const*)+0x11a) [0x7f5cad3e992a]
[bt] (2) /opt/conda/lib/python3.7/site-packages/dgl/libdgl.so(DGLLoadTensorAdapter+0x26) [0x7f5cad3bac76]
[bt] (3) /opt/conda/lib/python3.7/lib-dynload/../../libffi.so.8(+0x6a4a) [0x7f5d5c0a3a4a]
[bt] (4) /opt/conda/lib/python3.7/lib-dynload/../../libffi.so.8(+0x5fea) [0x7f5d5c0a2fea]
[bt] (5) /opt/conda/lib/python3.7/lib-dynload/_ctypes.cpython-37m-x86_64-linux-gnu.so(_ctypes_callproc+0x2f4) [0x7f5d5c0b97b4]
[bt] (6) /opt/conda/lib/python3.7/lib-dynload/_ctypes.cpython-37m-x86_64-linux-gnu.so(+0x1102c) [0x7f5d5c0ba02c]
[bt] (7) python3(_PyObject_FastCallKeywords+0x3cb) [0x5562b3145dcb]
[bt] (8) python3(+0x191f79) [0x5562b3146f79]

Hi, I read your paper. Thank you, it looks so helpful for me.
But one part of code is difficult to me.
In your code, it's look for single task predict.
But, how can I used it for the multi-task train and predict?

Hi, I read your paper. Thank you, it looks so generic to extend to the higher orders.
But one part is hard to understand for me.

In figure 2 of your paper, the input module concatenates one discrete feature and one continuous feature of order p.

Two orders were considered. Atom & Bond

How atom can have a continuous feature?
How bond can have a discrete feature?