skip to navigation
skip to content

Not Logged In

MLizard 0.1.0

Machine Learning workflow automatization

Latest Version: 0.1.2

Python machine learning infrastructure project. The idea of MLizard is to make it easy to run lots of different experiments on lots of different options, constantly changing or exchanging parts of the process, without loosing track of what you did, when you did it, how you did it, what came out of it, which files are connected to it and so on. So this is how it looks like:

"""
# for this demo we use the docstring as config
alpha = 0.7
beta = 7
gamma = "Foo"
"""
# we import the experiment factory
from mlizard.experiment import createExperiment

# at the beginning of the file we create an experiment
ex = createExperiment("Demo", config_string=__doc__)

@ex.stage
def part0(rnd):
   return rnd.randint(10)

@ex.stage
def part1(X, alpha, beta, logger):
   X -= alpha
   X *= beta
   logger.info("multiplied by %f and added %f", alpha, beta)
   return X

@main
def mainFunction():
   # this is the main method, here we put everything together
   X = part0() # note that we do not need to pass rnd
   X = part1(X) # and no alpha, beta, and logger

So we have to create an experiment and decorate all of our functions. But what do we get for this? - automatic option passing (alpha, beta) - a logger - a random number generator that is seeded by the experiment - automatic caching of intermediate results

More to come (Roadmap): - easy option sweeps - report file - online results view - database of runs/options/results - git integration (track version of code for every result)

 
File Type Py Version Uploaded on Size
MLizard-0.1.0.tar.gz (md5) Source 2012-10-23 8KB
  • Downloads (All Versions):
  • 25 downloads in the last day
  • 61 downloads in the last week
  • 305 downloads in the last month