module mlapp.mlpost_rest

Short summary

module lightmlrestapi.mlapp.mlpost_rest

Machine Learning Post request

source on GitHub

Classes

class	truncated documentation
MachineLearningPost	Implements a simple REST API which handles a post request, no authentification is required. The model ingests …

Static Methods

staticmethod	truncated documentation
data2json	numpy.array cannot be converted into json. We change the type into a list.

Methods

method	truncated documentation
__init__
_load
_predict_single
check_single	Checks the sequence load + predict returns something with the given observations.
on_post

Documentation

Machine Learning Post request

source on GitHub

class lightmlrestapi.mlapp.mlpost_rest.MachineLearningPost(load_function, predict_function, secret=None, folder='.', log_features=True, log_prediction=True, load_params=None, ccall='single', version=None)

Bases: BaseLogging

Implements a simple REST API which handles a post request, no authentification is required. The model ingests a vector X and outputs another one or a number Y. An basic example of an application is given by dummy_application.

source on GitHub

Parameters:

• predict_function – predict function

• load_function – load function

• secret – see BaseLogging

• folder – see BaseLogging

• log_features – log the features

• log_prediction – log the prediction

• load_params – given to the loading function

• ccall – see below

• version – API REST version

Some models can only computes predictions for a sequence of observations, not just one. Parameter ccall defines what the prediction function can ingest. * single: only one observation * multi: only multiple ones * both: the function determines what it must do

source on GitHub

__init__(load_function, predict_function, secret=None, folder='.', log_features=True, log_prediction=True, load_params=None, ccall='single', version=None)

Parameters:

• predict_function – predict function

• load_function – load function

• secret – see BaseLogging

• folder – see BaseLogging

• log_features – log the features

• log_prediction – log the prediction

• load_params – given to the loading function

• ccall – see below

• version – API REST version

Some models can only computes predictions for a sequence of observations, not just one. Parameter ccall defines what the prediction function can ingest. * single: only one observation * multi: only multiple ones * both: the function determines what it must do

source on GitHub

_call_convention = {'both': 2, 'multi': 1, 'single': 0}

_load()

_predict_single(obj, features)

check_single(features)

Checks the sequence load + predict returns something with the given observations.

source on GitHub

static data2json(data)

numpy.array cannot be converted into json. We change the type into a list.

source on GitHub

on_post(req, resp)

Parameters:

• req – request

• resp –

  …

source on GitHub