0

After I train a doc2vec model, I want to reuse the document vectors in another module. It seems there are two ways to implement this: save the model and save doc-vectors as a dictionary. I just wonder which one is more memory-efficient and which one reads faster in the target module?

YangGuo
  • 33
  • 5
  • Please post your benchmark results. How many seconds did the first and second method take in your environment? – J_H Dec 21 '17 at 03:05
  • That's going to depend on which library you're using and other choices in your code which you haven't disclosed. As @J_H suggests, try both approaches, and compare the timings. (And also the size-on-disk, and apparent RAM usage, etc.) – gojomo Dec 21 '17 at 04:57
  • Thanks guys. After some experiments, I found there is no significant difference between these two ways(At least for my dataset). – YangGuo Dec 21 '17 at 11:53

1 Answers1

0

Performance measurements matter. For an interesting number of datasets the two ways of saving will take similar time.

J_H
  • 17,926
  • 4
  • 24
  • 44