MatPlotLib Meter Logger

PLTMeterLogger.py

import numpy as np
import matplotlib.pyplot as plt
import torch
class ValueMeter:
    def __init__(self):
        self.reset()

    def add(self, value):
        self.values.append(value)

    @property
    def value(self):
        return self.values[-1]

    @property
    def mean(self):
        return np.mean(self.values)

    @property
    def std(self):
        return np.std(self.values)

    def reset(self):
        self.values = []


class PLTMeterLogger(object):
    ''' A class to package and visualize meters.
    '''
    def __init__(self, name="Main"):
        self.meters = {}
        self.name = name
        plt.rc('text', usetex=True)
        plt.rc('font', size=32, family='serif')



    def add(self, meter, value):
        if isinstance(value, torch.Tensor):
            value = value.item()

        if meter in self.meters:
            self.meters[meter].add(value)
        else:
            self.meters[meter] = ValueMeter()
            self.meters[meter].add(value)


    def plot(self, path='./'):
        with plt.style.context(('ggplot')):
            for k, v in self.meters.items():
                plt.figure(figsize=(8, 6))
                plt.title(k)
                plt.plot(v.values)
                plt.savefig('{}/{}.png'.format(path, k))
                plt.clf()

    def save(self, path='./', hyperparams={}):
        np.savez('{}/{}.npz'.format(path, self.name), {k: v.values for k, v in self.meters.items()}, **hyperparams)

    def toAcc(self, output, target):
        acc = output.argmax(dim=1).eq(target).sum().item() * 100.0 / output.size(0)
        return acc

    def reset(self):
        self.meters = {}

    def print(self, start=""):
        for k, v in self.meters.items():
            start += "{}: {:6.3f}\t".format(k.split(" ")[-1], v.mean)
        print(start)