Automated testing of graph pickling
Since we can pickle graphs, we should do some fuzz testing.
Construct random graphs, apply them a bunch of times, and see that we get the same results if we pcikle and unpickle in the middle of training as if we just keep training.
Construct random graphs, apply them a bunch of times, and see that we get the same results if we pcikle and unpickle in the middle of training as if we just keep training.
Leave a comment