import chainer
from chainer import cuda
import chainer.functions as F
import chainer.links as L
from chainer import Variable, optimizers
import numpy as np
import os
import logging
import datetime

log_name='predict_log.txt'
logging.basicConfig(filename=log_name,level=logging.INFO)
logging.info('***************')
logging.info(datetime.datetime.today())
logging.info('Predict Log')

# 中間層の数
n_units   = 1000

# 使用するモデル・オプティマイザのepoch数
epoch = 17

mnist = chainer.datasets.get_mnist()[0]

x_data = []
y_data = []

for data in mnist:
    x_data.append(data[0])
    y_data.append(data[1])
x_data = np.array(x_data).astype(np.float32)
y_data = np.array(y_data).astype(np.int32)

N = len(x_data)
N_test = 1000
print('N is {}'.format(N))

model = chainer.FunctionSet(l1=F.Linear(784, n_units),
                    l2=F.Linear(n_units, n_units),
                    l3=F.Linear(n_units, 10))

chainer.serializers.load_npz('model_tegaki_{}'.format(epoch),model)
optimizer = chainer.optimizers.Adam()
optimizer.setup(model)

chainer.serializers.load_npz('optimizer_tegaki_{}'.format(epoch),optimizer)

# 認識してみる
print('Predict MNIST.')
print('\tPredict\tLabel')
correct = 0
for idx in np.random.permutation(N)[:N_test]:
    #x = Variable(x_data[idx].reshape(1,1,28,28))
    x = Variable(x_data[idx])

    h1 = F.dropout(F.relu(model.l1(x)))
    h2 = F.dropout(F.relu(model.l2(h1)))
    y = model.l3(h2)

    if(int(y_data[idx]) == np.argmax(y.data)):
        correct = correct + 1
    else:
        print('Missed.',np.argmax(y.data),'\t',y_data[idx])
        logging.debug('missed:{}!={}'.format(np.argmax(y.data),y_data[idx]))

accuracy = correct / N_test
print('accuracy:',accuracy,'({} / {})'.format(correct,N_test))
logging.info('accuracy: {}'.format(accuracy))

logging.info('***************')