chronometry

Chronometry

ProgressBar

Estimator

Estimator is an object that estimates the running time of a single argument function. You can use it to avoid running a script for too long. For example, if you want to cluster a large dataset and running it might take too long, and cost too much if you use cloud computing, you can create a function with one argument x which takes a sample with x rows and clusters it; then you can use Estimator to estimate how long it takes to run it on the full dataset by providing the actual number of rows to the estimate() method.

Estimator uses a Polynomial Linear Regression model and gives more weight to larger numbers for the training.

Usage

from chronometry import Estimator
from time import sleep

def multiply_with_no_delay(x, y):
    return (x ** 2 + 0.1 * x ** 3 + 1) * 0.00001 + y * 0.001

def multiply(x, y):
    sleep_time = multiply_with_no_delay(x, y)
    if sleep_time > 30:
        raise
    sleep(sleep_time)
    if y == 6:
        sleep(12)
    elif 7 < y < 15:
        raise Exception()
    return sleep_time

estimator = Estimator(function=multiply, polynomial_degree=3, timeout=5)
# the `unit` argument chooses the unit of time to be used. By default unit='s'

estimator.auto_explore()
estimator.predict_time(x=10000, y=10000)

The above code runs for about 53 seconds and then estimates that multiply(10000, 10000) will take 1002371.7 seconds which is only slightly smaller than the correct number: 1001010 seconds.

max_time is the maximum time allowed for the estimate function to run.

If you are using Estimator in Jupyter, you can plot the measurements with the plot() method (no arguments needed) which returns a matplotlib AxesSubplot object and displays it at the same time.

estimator.plot('x')

estimator.plot('y')