Hanjun Dai

Hi, In general this code is building the symbolic computation graph first, and then run the forward/backward. This is similar to tensorflow's logic. Regarding the two functions you are talking...

Hi, There was a (possibly out-dated) document: https://www.cc.gatech.edu/~hdai8/graphnn/html/annotated.html Basically you add new operators to the computation graph through the 'af' function: https://github.com/Hanjun-Dai/graphnn/blob/bdf51e66231d51bc2b9a560b2be255bc642d4a03/include/nn/factor_graph.h#L334 like af(computation_graph, {list of inputs}, other arguments). But...

It is trained jointly. For example, in MVC, up to [this line](https://github.com/Hanjun-Dai/graph_comb_opt/blob/master/code/s2v_mvc/mvc_lib/src/lib/qnet.cpp#L65) was for embedding the graph, and after that we feed the embedding into another mlp to calculate the...

Hi there, Have you built the .so file successfully?

which version of python are you using? I'm using py2 in this repo

go to clib folder under each experiment, and copy the Makefile.example to Makefile (with necessary adaptation to your system config). After that simply run make command.

Yes there's no pretrained model dump in this repo

Hi there, From the error message it seems you are getting 'out of memory'. I'm not sure whether that's the reason of segfault.

Hello, sorry for the late reply -- I didn't see the 'test.py', which file are you working with? Basically this function will load the model with best validation error: https://github.com/Hanjun-Dai/graph_comb_opt/blob/master/code/s2v_mvc/evaluate.py#L14...

I checked on my side and it works well. Probably that is relevant to your own key? You can also clone the two repositories separately, without using recursive clone.