deepwalk
deepwalk copied to clipboard
Option to only dump random walks to disk and skip training
gensim version downgraded to 0.10.1 as 0.10.2 does not install via easy_install due to this bug: https://groups.google.com/forum/#!topic/gensim/NSOXuP4IE9Q
In [1]: import gensim
In [2]: gensim.version Out[2]: '0.10.1'
vvkulkarni@descartes:~/deepwalk$ deepwalk --input ./example_graphs/karate.adjlist --output karate.embeddings Number of nodes: 34 Number of walks: 340 Data size (walks*length): 13600 Walking... Training...
vvkulkarni@descartes:~/deepwalk$ ls -ltr karate.embeddings -rw-rw-r-- 1 vvkulkarni vvkulkarni 20847 Jan 26 15:02 karate.embeddings
This is the solution I used in polyglot https://github.com/aboSamoor/polyglot/blob/master/setup.py#L20-L22
Pushing in changes to only dump walks if needed. Change needed for extended work.
vvkulkarni@curie:~/toolkits/viveks_deepwalk/deepwalk$ deepwalk --input example_graphs/karate.adjlist --output karate.embeddings --max-memory-data-size 0 Number of nodes: 34 Number of walks: 340 Data size (walks_length): 13600 Data size 13600 is larger than limit (max-memory-data-size: 0). Dumping walks to disk. Walking... Counting vertex frequency... Training... vvkulkarni@curie:~/toolkits/viveks_deepwalk/deepwalk$ deepwalk --input example_graphs/karate.adjlist --output karate.embeddings --max-memory-data-size 0 --only-walk Number of nodes: 34 Number of walks: 340 Data size (walks_length): 13600 Data size 13600 is larger than limit (max-memory-data-size: 0). Dumping walks to disk. Walking...