Sorry to bother you, but, where can I obtain the smile_list?

In your paper, you mentioned the validity of the graph is checked using RDKit.
What specific function did you use? Is it on the graph, or on the SMILES format converted from the graph?
Thanks in advance.

Hi,
I try to run nevae and find something confusing.

1. In the model.py file, the get_lossfunc function only contains KL and likelihood losses, but the paper states that there are four losses including KL, likelihood, log(pλn), log(pλm). Should I need to implement log(pλn) and log(pλm) by myself?

2. In the model.py file, the implementation of masked_gen function is a little bit complex, so I can't understand it clearly. Could you give me some related papers or information to read?

3. In the utils.py file, the code of line 135 is edgelist which appends same all edges of compound graph many times in the loop of G.nodes(). I think that line135 should be put outside the scope of loop of G.nodes() to avoid duplicate contents. If I make mistakes, please let me know.

Thank you for your patience to read my questions. I'll appreciate it if anyone can answer my questions.

Hi,
I've seen the paper, Nevae, paper showed that the C_u(k) aggregates from t_u(atom type) , x_u（coordinates）and weight(bond type),while codes(layer.py, input_layer function) showed that c_x aggregates from adj, weight, and degree feature. confusing me.
sorry to bother you. hope your reply. thank you.

How can I get the training data? It seems there doesn't have introduction about data format or data examples for running the program. I'll appreciate for your help.

lots of error! do you test the code before opensource?

Hi,

I'm sorry to bother you. I attended the conference held at Boston University this summer and found your works interesting to me. I tried to make the codes work to better understand your paper but I failed to run them due to the errors. Is there a clean version of the codes available now? Thank you!

Hello, I'm trying to run the project. And I find some problems when I run the sample.py

1. model.py line: 189 , through the decoder, the size of the dec_out should be n * n ,but the code final can't get the right size. so, it can't be reshape to n * n . w_edge has the same problem.
2. model.py, sample_graph function can't work properly, feed_dict.update gets error(ValueError: setting an array element with a sequence). I changed it to: feed_dict.update({self.weight_bin: weight_bins[0]}) feed_dict.update({self.edges: edges[0]}). and it works. but the decoder still has problem.
3. hope get your answer. thanks for your reading.