devforfu

import fire
import matplotlib.pyplot as plt
from matplotlib import rcParams


default_style = {
    'font.family': 'monospace',
    'font.size': 18,
    'figure.figsize': (8, 6)
}

devforfu

"""plotter.py generates simple plots configured from standard input or JSON.

A user is required to provide list of points to be plotted and canvas
parameters to define image rendering style properties.

examples:
    $ python plotter.py stdin -p '1,2;2,3'
    $ python plotter.py stdin -p '1,1;2,2;3,3' -o plot -f svg --show-grid
    $ python plotter.py json -j path/to/the/config.json
"""

devforfu

from argparse import ArgumentParser, ArgumentError, ArgumentTypeError
from functools import partial
import json
from os.path import exists
import matplotlib.pyplot as plt


def main():
    parser = create_parser()
    args = parser.parse_args()

devforfu

from argparse import ArgumentParser, ArgumentTypeError
import matplotlib.pyplot as plt


def main():
    parser = ArgumentParser()
    parser.add_argument(
        '-p', '--points',
        type=list_of_points,
        metavar='PTS', required=True,

devforfu

import argparse
import matplotlib.pyplot as plt


def main():
    parser = argparse.ArgumentParser()
    parser.add_argument(
        '-p', '--points',
        metavar='PTS', required=True,
        help='List of points to plot'

devforfu

class StreamLogger(Callback):

    def __init__(self, streams=None, log_every=1):
        self.streams = streams or [sys.stdout]
        self.log_every = log_every

    def epoch_ended(self, phases, epoch, **kwargs):
        if epoch % self.log_every != 0:
            return
        metrics = merge_dicts([phase.last_metrics for phase in phases])

devforfu

data_path = Path.home()/'data'/'mnist'

mnist_stats = ((0.15,), (0.15,))

epochs = 3

train_ds = MNIST(
    data_path, 
    train=True, 
    download=True,

devforfu

class Net(nn.Module):
    
    def __init__(self):
        super(Net, self).__init__()
        self.conv1 = nn.Conv2d(1, 10, kernel_size=5)
        self.conv2 = nn.Conv2d(10, 20, kernel_size=5)
        self.conv2_drop = nn.Dropout2d()
        self.fc1 = nn.Linear(320, 50)
        self.fc2 = nn.Linear(50, 10)

devforfu

class OneCycleSchedule:

    def __init__(self, t, linear_pct=0.2, eta_max=1.0, eta_min=None,
                 div_factor=100, decay_to_zero=True):

        if eta_min is None:
            eta_min = eta_max / div_factor

        self.t = t
        self.linear_pct = linear_pct

devforfu

class CosineAnnealingSchedule:
    """
    The schedule class that returns a learning rate multiplier from range [0.0, 1.0]
    """
    def __init__(self, eta_min=0.0, eta_max=1.0, t_max=100, t_mult=2):
        self.eta_min = eta_min
        self.eta_max = eta_max
        self.t_max = t_max
        self.t_mult = t_mult
        self.iter = 0