xgfs / node2vec-c Goto Github PK

nv: 1032440, ne: 987454
Need 0.257797 Mb for storing second-order degrees

Generating a corpus for negative samples..

Using vectorized operations
lr 0.020144, Progress 19.42%
Calculations took 12.13 s to run

Thanks for the excellent code.

and I met one question, my data set is too big (which can not be held in one machine's mem), and I should break it to small daily set.
so I should first generate each day's walk result (sequence) and then train by other code(suan as Gensim) as word2vec.

All I want is the random walking result

as for the walking result, should I just return before the part listed as following?
and then save dw_rw to disk for latter training?

I was wondering if there is (algorithmically) any difference between your C++ implementation of Node2Vec and the original reference implementation from Stanford. For instance, I saw something about subsampling frequent nodes in the C++ code.
I'm asking because I tried a few other Node2Vec implementations, including the one in pytorch_geometric, and have had trouble replicating the good performance of your C++ code. So I was wondering if there is anything missing in those.

./node2vec -input /data/ECommAI/round1_b/user_item.bcsr -output embedding.bin -dim 64
I use this command, but the following error occurs:

nv: 985597, ne: 69327988
Segmentation fault (core dumped)

32GB memory