On the May 25 commit, many part of the code have been re-written in a map-reduce style. I tried to parallellize the classifications with multiprocessing.Pool. The speedup is around 200%, but the result is inconsistent witht he sequential version. I hope someone could make another attempt. Refer to multiprocessing or Jobjib (built upon multiprocessing): https://pythonhosted.org/joblib/index.html
On the May 25 commit, many part of the code have been re-written in a map-reduce style. I tried to parallellize the classifications with
multiprocessing.Pool. The speedup is around 200%, but the result is inconsistent witht he sequential version. I hope someone could make another attempt. Refer tomultiprocessingorJobjib(built uponmultiprocessing): https://pythonhosted.org/joblib/index.html