Zhaocheng Zhu
Zhaocheng Zhu
Good idea. A small problem is that ensuring the entity set are observed in the splits might not be as scalable as previous stream-style random split. We will discuss on...
I just checked entity prediction. It works on `FB15k` and `WN18`. So it means their test splits don't have any unobserved entities. Then we will change `edge_split()` to satisfy that.
I would suggest using vertex embedding alone, or concating the two embeddings.
Only 2nd order is used. We found that 2nd order is consistently better than 1st order.
Pip doesn't install pybind11 search script to cmake. If I remember correctly, the conda version of pybind11 will do the correct job if your cmake is also installed by conda....
Sorry, we are still working on this feature.
All the solvers in GraphVite have only been compiled for common dimensions, for maximal GPU optimization, like 32, 64, 128, etc. You can use 32 instead. 10 is also too...
`pthread` library is missing. It looks like your C++ compilation tools are not setup properly. Have you installed the dependencies in `conda/requirements.txt`? Or you can try `conda install gxx_linux-64`.
There shouldn't be large memory allocation. Maybe it is caused by `pickle`? We will check the memory usage later.
It's an illegal memory access in GPUs. Really weird. Could you provide any graph dataset that can reproduce this error?