教电脑学会加法运算
教电脑学会加法运算---RNN的应用例子
刚看到这个题目，我心里就想，计算机进行加法运算还不简单吗？ print x+y 不就行啦？ so easy！
如果我们认真探究一下，我们人类如何做加法运算的话，让计算机模仿人类来学习加法运算，那就没那么简单了。
我们想想下面的情景：
- 我们小的时候对数字开始认识，是从分糖块开始认识数字的，最好的计算工具就是双手。
- 随着掌握的增多，我记得10以内的加法，3+5=8 会随口而出，这个貌似不用数手指头了。
- 现在我们每个人都可以随口而出的10以内2数的加法运算，没有一个人会想想，也没有人数手指头，都是随口而出的。
以上情景告诉我们，我们的大脑不是计算器，简单的运算是靠记忆算出来的，如果让计算机也模仿人脑，而不采用计算器，那么这个话题就是今天我们要讨论的。
这是一个循环神经网络（RNN）的应用，通过大量数据集（1+1=2， 2+3=5， ...)教会计算机学会加法运算，像我们人类大脑一样会脱口而出说出答案，而且也像我们人类一样会偶尔犯错误弄错答案。
当然我们也不会记忆所有的数据，有些可以靠推断进行运算。
下面的例子是教会电脑三位数的加法（XYZ+ABC），数据集模拟生成50000个记录，实际上全部可能应该是 1000X1000即100万，就是我们的数据集只是实际的5%，这个也符合我们人类的认知习惯，学以一小部分就可以掌握全部知识。
采用seq2seq (顺序到顺序)的神经网络的加法运算的实现
输入: "535+61"
输出: "596"
字串使用多个空格填充采用序列到序列的神经网络进行机器学习 "Sequence to Sequence Learning with Neural Networks"
本代码来自 keras的官方github上的examples中的 addition_rnn.py, 描述使用RNN进行加法运算的例子，本人做了一点改动，主要是便于理解，内容详见：
[keras examples]（https://github.com/fchollet/keras/tree/master/examples）
%matplotlib inline
from __future__ import print_function
from keras.models import Sequential
from keras import layers
import numpy as np
from six.moves import range
from IPython.display import SVG
from keras.utils.vis_utils import model_to_dot
import pandas as pd
import matplotlib.pyplot as plt
Using TensorFlow backend.
采用 XYZ+ABC的加法训练集，需要我们自动生成
class CharacterTable(object):
"""
生成字符集：
编码为热独数
解码热独数为字符输出
解码字符的向量的概率
"""
"""Given a set of characters:
+ Encode them to a one hot integer representation
+ Decode the one hot integer representation to their character output
+ Decode a vector of probabilities to their character output
"""
def __init__(self, chars):
"""
初始化字符表
参数
字符为可能输入的字母
"""
"""Initialize character table.
# Arguments
chars: Characters that can appear in the input.
"""
self.chars = sorted(set(chars))
self.char_indices = dict((c, i) for i, c in enumerate(self.chars))
self.indices_char = dict((i, c) for i, c in enumerate(self.chars))
def encode(self, C, num_rows):
"""One hot encode given string C.
# Arguments
num_rows: Number of rows in the returned one hot encoding. This is
used to keep the # of rows for each data the same.
"""
x = np.zeros((num_rows, len(self.chars)))
for i, c in enumerate(C):
x[i, self.char_indices[c]] = 1
return x
def decode(self, x, calc_argmax=True):
if calc_argmax:
x = x.argmax(axis=-1)
return ''.join(self.indices_char[x] for x in x)
class colors:
ok = '\033[92m'
fail = '\033[91m'
close = '\033[0m'
# 模型和数据集的参数 Parameters for the model and dataset.
# TRAINING_SIZE训练集的大小，数据集模拟生成50000个记录作为训练集，占实际可能的总体数据的5%
# DIGITS 输入数字的长度
# MAXLEN 输入数据的最大长度
# 例如 '345+678' DIGITS =3 MAXLEN = 3+1+3
# INVERT是数字是否反转 我个人感觉有点怪怪的感觉，就做了 False，略过了。
TRAINING_SIZE = 50000
DIGITS = 3
INVERT = False
# Maximum length of input is 'int + int' (e.g., '345+678'). Maximum length of
# int is DIGITS.
MAXLEN = DIGITS + 1 + DIGITS
# 模拟生成50000个记录（questions和expected）作为训练集
# 所有的数字，加号和空格
# All the numbers, plus sign and space for padding.
chars = '0123456789+ '
ctable = CharacterTable(chars)
questions = []
expected = []
seen = set()
print('Generating data...')
while len(questions) < TRAINING_SIZE:
f = lambda: int(''.join(np.random.choice(list('0123456789'))
for i in range(np.random.randint(1, DIGITS + 1))))
a, b = f(), f()
# Skip any addition questions we've already seen
# Also skip any such that x+Y == Y+x (hence the sorting).
key = tuple(sorted((a, b)))
if key in seen:
continue
seen.add(key)
# Pad the data with spaces such that it is always MAXLEN.
q = '{}+{}'.format(a, b)
query = q + ' ' * (MAXLEN - len(q))
ans = str(a + b)
# Answers can be of maximum size DIGITS + 1.
ans += ' ' * (DIGITS + 1 - len(ans))
if INVERT:
# Reverse the query, e.g., '12+345 ' becomes ' 543+21'. (Note the
# space used for padding.)
query = query[::-1]
questions.append(query)
expected.append(ans)
print('Total addition questions:', len(questions))
# 打印数据集的前100个问题和答案
print(questions[1:100])
print('Total addition expected:', len(expected))
print(expected[1:100])
Generating data...
Total addition questions: 50000
['51+9 ', '346+7 ', '2+11 ', '750+8 ', '539+84 ', '439+3 ', '778+7 ', '0+438 ', '47+4 ', '7+22 ', '8+7 ', '0+95 ', '900+285', '84+1 ', '975+6 ', '7+6 ', '828+830', '183+8 ', '63+254 ', '781+31 ', '4+2 ', '27+9 ', '453+550', '6+38 ', '927+47 ', '500+507', '593+536', '55+11 ', '479+2 ', '19+56 ', '377+946', '776+718', '85+920 ', '28+327 ', '2+92 ', '70+4 ', '0+9 ', '8+3 ', '32+200 ', '37+8 ', '6+64 ', '204+949', '96+94 ', '36+21 ', '63+101 ', '442+77 ', '463+988', '608+24 ', '2+6 ', '453+11 ', '413+7 ', '61+590 ', '1+556 ', '76+140 ', '9+6 ', '7+0 ', '9+260 ', '1+73 ', '2+5 ', '4+5 ', '84+8 ', '1+1 ', '5+0 ', '544+0 ', '906+4 ', '72+11 ', '213+954', '110+67 ', '245+28 ', '224+4 ', '975+412', '96+58 ', '26+335 ', '84+43 ', '9+8 ', '5+7 ', '1+0 ', '5+1 ', '6+95 ', '453+69 ', '61+230 ', '3+179 ', '1+4 ', '474+12 ', '3+81 ', '6+46 ', '52+4 ', '55+8 ', '337+22 ', '35+0 ', '815+29 ', '202+98 ', '796+81 ', '89+47 ', '68+827 ', '0+2 ', '74+191 ', '7+357 ', '99+52 ']
Total addition expected: 50000
['60 ', '353 ', '13 ', '758 ', '623 ', '442 ', '785 ', '438 ', '51 ', '29 ', '15 ', '95 ', '1185', '85 ', '981 ', '13 ', '1658', '191 ', '317 ', '812 ', '6 ', '36 ', '1003', '44 ', '974 ', '1007', '1129', '66 ', '481 ', '75 ', '1323', '1494', '1005', '355 ', '94 ', '74 ', '9 ', '11 ', '232 ', '45 ', '70 ', '1153', '190 ', '57 ', '164 ', '519 ', '1451', '632 ', '8 ', '464 ', '420 ', '651 ', '557 ', '216 ', '15 ', '7 ', '269 ', '74 ', '7 ', '9 ', '92 ', '2 ', '5 ', '544 ', '910 ', '83 ', '1167', '177 ', '273 ', '228 ', '1387', '154 ', '361 ', '127 ', '17 ', '12 ', '1 ', '6 ', '101 ', '522 ', '291 ', '182 ', '5 ', '486 ', '84 ', '52 ', '56 ', '63 ', '359 ', '35 ', '844 ', '300 ', '877 ', '136 ', '895 ', '2 ', '265 ', '364 ', '151 ']
# 数据集做预处理，向量化
# 数据打乱后，45000个做训练数据集，5000个做验证数据集
print('Vectorization...')
x = np.zeros((len(questions), MAXLEN, len(chars)), dtype=np.bool)
y = np.zeros((len(questions), DIGITS + 1, len(chars)), dtype=np.bool)
for i, sentence in enumerate(questions):
x[i] = ctable.encode(sentence, MAXLEN)
for i, sentence in enumerate(expected):
y[i] = ctable.encode(sentence, DIGITS + 1)
# Shuffle (x, y) in unison as the later parts of x will almost all be larger
# digits.
indices = np.arange(len(y))
np.random.shuffle(indices)
x = x[indices]
y = y[indices]
# Explicitly set apart 10% for validation data that we never train over.
split_at = len(x) - len(x) // 10
(x_train, x_val) = x[:split_at], x[split_at:]
(y_train, y_val) = y[:split_at], y[split_at:]
print('Training Data:')
print(x_train.shape)
print(y_train.shape)
print('Validation Data:')
print(x_val.shape)
print(y_val.shape)
Vectorization...
Training Data:
(45000, 7, 12)
(45000, 4, 12)
Validation Data:
(5000, 7, 12)
(5000, 4, 12)
# 下面的代码是构建模型的部分，具体可以看model.summary的内容和生成的模型的SVG图形
# 构建自己的一个RNN模型而不是直接采用 GRU 或 SimpleRNN.
# Try replacing GRU, or SimpleRNN.
RNN = layers.LSTM
HIDDEN_SIZE = 128
BATCH_SIZE = 128
LAYERS = 1
print('Build model...')
model = Sequential()
# 编码输入序列使用RNN，输出产生隐藏层。
# 值得一提的是： 你的输入序列是个可变长度，使用input_shape=(None, num_feature)，其中num_feature=len(chars)。
# "Encode" the input sequence using an RNN, producing an output of HIDDEN_SIZE.
# Note: In a situation where your input sequences have a variable length,
# use input_shape=(None, num_feature).
model.add(RNN(HIDDEN_SIZE, input_shape=(MAXLEN, len(chars))))
# 作为解码RNN的输入，重复提供RNN循环一次每次的最新状态。
# 重复 'DIGITS + 1'次，作为输出的最大长度，例如：当DIGITS=3,最大输出为 999+999=1998.
# As the decoder RNN's input, repeatedly provide with the last hidden state of
# RNN for each time step. Repeat 'DIGITS + 1' times as that's the maximum
# length of output, e.g., when DIGITS=3, max output is 999+999=1998.
model.add(layers.RepeatVector(DIGITS + 1))
# 解码 RNN可以是多个层，也可以是单个层。
# The decoder RNN could be multiple layers stacked or a single layer.
for _ in range(LAYERS):
# 通过设置 return_sequence=True，不仅返回最后一次的输出，也包括所有的输出，即便在是(num_samples, timesteps, output_dim).
# 这个是 keras中LSTM，采用TimeDistributed包装层的问题，这个演示中语焉不详，大家可是看看LSTM网络的不同方式和TimeDistributed层的作用
# By setting return_sequences to True, return not only the last output but
# all the outputs so far in the form of (num_samples, timesteps,
# output_dim). This is necessary as TimeDistributed in the below expects
# the first dimension to be the timesteps.
model.add(RNN(HIDDEN_SIZE, return_sequences=True))
# 为每个输入的时间片段提供一个连接层。
# 对每一步的输出序列，决定选择哪个字符。
# Apply a dense layer to the every temporal slice of an input. For each of step
# of the output sequence, decide which character should be chosen.
model.add(layers.TimeDistributed(layers.Dense(len(chars))))
model.add(layers.Activation('softmax'))
model.compile(loss='categorical_crossentropy',
optimizer='adam',
metrics=['accuracy'])
model.summary()
SVG(model_to_dot(model,show_shapes=True).create(prog='dot', format='svg'))
Build model...
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
lstm_1 (LSTM) (None, 128) 72192
_________________________________________________________________
repeat_vector_1 (RepeatVecto (None, 4, 128) 0
_________________________________________________________________
lstm_2 (LSTM) (None, 4, 128) 131584
_________________________________________________________________
time_distributed_1 (TimeDist (None, 4, 12) 1548
_________________________________________________________________
activation_1 (Activation) (None, 4, 12) 0
=================================================================
Total params: 205,324
Trainable params: 205,324
Non-trainable params: 0
_________________________________________________________________
RNN模型
# 训练模型，每次训练都做一回训练结果的验证。
# Train the model each generation and show predictions against the validation
# dataset.
loss = []
acc = []
val_loss = []
val_acc = []
for iteration in range(1, 200):
print()
print('-' * 50)
print('Iteration', iteration)
hist = model.fit(x_train, y_train,
batch_size=BATCH_SIZE,
epochs=1,
validation_data=(x_val, y_val))
# Select 10 samples from the validation set at random so we can visualize
# errors.
print(hist.history)
loss.append(hist.history['loss'][0])
acc.append(hist.history['acc'][0])
val_loss.append(hist.history['val_loss'][0])
val_acc.append(hist.history['val_acc'][0])
for i in range(10):
ind = np.random.randint(0, len(x_val))
rowx, rowy = x_val[np.array([ind])], y_val[np.array([ind])]
preds = model.predict_classes(rowx, verbose=0)
q = ctable.decode(rowx[0])
correct = ctable.decode(rowy[0])
guess = ctable.decode(preds[0], calc_argmax=False)
print('Q', q[::-1] if INVERT else q)
print('T', correct)
if correct == guess:
print(colors.ok + '☑' + colors.close, end=" ")
else:
print(colors.fail + '☒' + colors.close, end=" ")
print(guess)
print('---')
训练过程
--------------------------------------------------
Iteration 1
Train on 45000 samples, validate on 5000 samples
Epoch 1/1
45000/45000 [==============================] - 38s - loss: 1.8916 - acc: 0.3204 - val_loss: 1.7901 - val_acc: 0.3464
{'val_loss': [1.7900752948760987], 'val_acc': [0.34644999999999998], 'loss': [1.8915647541681926], 'acc': [0.32042222222222222]}
Q 18+532
T 550
�[91m☒�[0m 129
---
Q 326+29
T 355
�[91m☒�[0m 122
---
Q 78+753
T 831
�[91m☒�[0m 100
---
Q 738+62
T 800
�[91m☒�[0m 102
---
Q 28+159
T 187
�[91m☒�[0m 109
---
Q 230+401
T 631
�[91m☒�[0m 122
---
Q 427+724
T 1151
�[91m☒�[0m 102
---
Q 57+777
T 834
�[91m☒�[0m 109
---
Q 298+55
T 353
�[91m☒�[0m 109
---
Q 416+29
T 445
�[91m☒�[0m 129
---
--------------------------------------------------
Iteration 2
Train on 45000 samples, validate on 5000 samples
Epoch 1/1
45000/45000 [==============================] - 37s - loss: 1.7554 - acc: 0.3522 - val_loss: 1.6911 - val_acc: 0.3728
{'val_loss': [1.6910695877075195], 'val_acc': [0.37275000000000003], 'loss': [1.7553744444105359], 'acc': [0.35220555555025734]}
Q 192+12
T 204
�[91m☒�[0m 221
---
Q 353+99
T 452
�[91m☒�[0m 103
---
Q 381+886
T 1267
�[91m☒�[0m 1222
---
Q 35+50
T 85
�[91m☒�[0m 55
---
Q 150+62
T 212
�[91m☒�[0m 567
---
Q 285+4
T 289
�[91m☒�[0m 33
---
Q 698+699
T 1397
�[91m☒�[0m 1692
---
Q 447+3
T 450
�[91m☒�[0m 35
---
Q 91+215
T 306
�[91m☒�[0m 121
---
Q 695+7
T 702
�[91m☒�[0m 103
---
--------------------------------------------------
Iteration 3
Train on 45000 samples, validate on 5000 samples
Epoch 1/1
45000/45000 [==============================] - 37s - loss: 1.6050 - acc: 0.3986 - val_loss: 1.5398 - val_acc: 0.4189
{'val_loss': [1.5398105825424195], 'val_acc': [0.41894999999999999], 'loss': [1.6049728581110636], 'acc': [0.39855000001589458]}
Q 9+123
T 132
�[91m☒�[0m 123
---
Q 676+76
T 752
�[91m☒�[0m 762
---
Q 956+59
T 1015
�[91m☒�[0m 1021
---
Q 215+249
T 464
�[91m☒�[0m 355
---
Q 793+975
T 1768
�[91m☒�[0m 1687
---
Q 615+1
T 616
�[91m☒�[0m 621
---
Q 60+947
T 1007
�[91m☒�[0m 1021
---
Q 39+820
T 859
�[91m☒�[0m 891
---
Q 321+1
T 322
�[91m☒�[0m 22
---
Q 339+7
T 346
�[91m☒�[0m 337
---
--------------------------------------------------
Iteration 4
Train on 45000 samples, validate on 5000 samples
Epoch 1/1
45000/45000 [==============================] - 37s - loss: 1.4618 - acc: 0.4521 - val_loss: 1.3862 - val_acc: 0.4834
{'val_loss': [1.3862218780517579], 'val_acc': [0.4834], 'loss': [1.4617731448067559], 'acc': [0.45206111112170749]}
Q 58+326
T 384
�[91m☒�[0m 353
---
Q 323+9
T 332
�[92m☑�[0m 332
---
Q 845+808
T 1653
�[91m☒�[0m 1554
---
Q 773+6
T 779
�[91m☒�[0m 744
---
Q 496+259
T 755
�[91m☒�[0m 667
---
Q 34+83
T 117
�[91m☒�[0m 13
---
Q 26+431
T 457
�[91m☒�[0m 466
---
Q 91+25
T 116
�[91m☒�[0m 12
---
Q 23+25
T 48
�[91m☒�[0m 36
---
Q 165+5
T 170
�[91m☒�[0m 166
---
--------------------------------------------------
Iteration 5
Train on 45000 samples, validate on 5000 samples
Epoch 1/1
45000/45000 [==============================] - 37s - loss: 1.3244 - acc: 0.5041 - val_loss: 1.2633 - val_acc: 0.5250
{'val_loss': [1.2633318698883056], 'val_acc': [0.52500000000000002], 'loss': [1.3243814388910928], 'acc': [0.50412777777777773]}
Q 49+51
T 100
�[91m☒�[0m 10
---
Q 885+176
T 1061
�[91m☒�[0m 100
---
Q 318+68
T 386
�[91m☒�[0m 421
---
Q 98+523
T 621
�[91m☒�[0m 641
---
Q 73+469
T 542
�[91m☒�[0m 521
---
Q 49+75
T 124
�[91m☒�[0m 111
---
Q 76+365
T 441
�[91m☒�[0m 421
---
Q 279+617
T 896
�[91m☒�[0m 804
---
Q 402+322
T 724
�[91m☒�[0m 767
---
Q 835+435
T 1270
�[91m☒�[0m 1277
---
--------------------------------------------------
Iteration 11
Train on 45000 samples, validate on 5000 samples
Epoch 1/1
45000/45000 [==============================] - 36s - loss: 0.6989 - acc: 0.7403 - val_loss: 0.6459 - val_acc: 0.7578
{'val_loss': [0.64586523256301875], 'val_acc': [0.75775000000000003], 'loss': [0.69888072902891374], 'acc': [0.7403388889100817]}
Q 256+30
T 286
�[91m☒�[0m 277
---
Q 8+721
T 729
�[92m☑�[0m 729
---
Q 341+36
T 377
�[92m☑�[0m 377
---
Q 263+112
T 375
�[91m☒�[0m 474
---
Q 1+705
T 706
�[91m☒�[0m 707
---
Q 713+95
T 808
�[92m☑�[0m 808
---
Q 951+737
T 1688
�[91m☒�[0m 1679
---
Q 47+767
T 814
�[91m☒�[0m 824
---
Q 415+625
T 1040
�[91m☒�[0m 967
---
Q 300+4
T 304
�[91m☒�[0m 305
---
--------------------------------------------------
Iteration 21
Train on 45000 samples, validate on 5000 samples
Epoch 1/1
45000/45000 [==============================] - 38s - loss: 0.0879 - acc: 0.9817 - val_loss: 0.1036 - val_acc: 0.9716
{'val_loss': [0.10357598056793213], 'val_acc': [0.97165000000000001], 'loss': [0.087886847303973309], 'acc': [0.98168333332273694]}
Q 571+58
T 629
�[92m☑�[0m 629
---
Q 396+781
T 1177
�[92m☑�[0m 1177
---
Q 234+980
T 1214
�[92m☑�[0m 1214
---
Q 145+627
T 772
�[91m☒�[0m 762
---
Q 91+137
T 228
�[92m☑�[0m 228
---
Q 7+419
T 426
�[92m☑�[0m 426
---
Q 772+75
T 847
�[92m☑�[0m 847
---
Q 705+128
T 833
�[92m☑�[0m 833
---
Q 520+6
T 526
�[92m☑�[0m 526
---
Q 732+74
T 806
�[92m☑�[0m 806
---
--------------------------------------------------
Iteration 22
Train on 45000 samples, validate on 5000 samples
Epoch 1/1
45000/45000 [==============================] - 38s - loss: 0.0881 - acc: 0.9788 - val_loss: 0.0858 - val_acc: 0.9790
{'val_loss': [0.085751581120491027], 'val_acc': [0.97899999999999998], 'loss': [0.088109891496764292], 'acc': [0.97884444448682995]}
Q 63+61
T 124
�[92m☑�[0m 124
---
Q 807+60
T 867
�[92m☑�[0m 867
---
Q 768+63
T 831
�[92m☑�[0m 831
---
Q 438+52
T 490
�[92m☑�[0m 490
---
Q 51+763
T 814
�[92m☑�[0m 814
---
Q 775+84
T 859
�[92m☑�[0m 859
---
Q 642+69
T 711
�[92m☑�[0m 711
---
Q 66+622
T 688
�[92m☑�[0m 688
---
Q 166+500
T 666
�[92m☑�[0m 666
---
Q 624+28
T 652
�[92m☑�[0m 652
---
--------------------------------------------------
Iteration 30
Train on 45000 samples, validate on 5000 samples
Epoch 1/1
45000/45000 [==============================] - 38s - loss: 0.0276 - acc: 0.9949 - val_loss: 0.0443 - val_acc: 0.9864
{'val_loss': [0.044335523164272306], 'val_acc': [0.98640000000000005], 'loss': [0.027582680843936072], 'acc': [0.99492777774598862]}
Q 498+7
T 505
�[92m☑�[0m 505
---
Q 444+207
T 651
�[92m☑�[0m 651
---
Q 546+746
T 1292
�[92m☑�[0m 1292
---
Q 673+68
T 741
�[92m☑�[0m 741
---
Q 316+45
T 361
�[92m☑�[0m 361
---
Q 8+109
T 117
�[92m☑�[0m 117
---
Q 538+594
T 1132
�[92m☑�[0m 1132
---
Q 58+876
T 934
�[92m☑�[0m 934
---
Q 99+867
T 966
�[92m☑�[0m 966
---
Q 302+840
T 1142
�[92m☑�[0m 1142
---
--------------------------------------------------
Iteration 40
Train on 45000 samples, validate on 5000 samples
Epoch 1/1
45000/45000 [==============================] - 38s - loss: 0.0084 - acc: 0.9992 - val_loss: 0.0225 - val_acc: 0.9932
{'val_loss': [0.022462421262264252], 'val_acc': [0.99324999999999997], 'loss': [0.0084158510135279758], 'acc': [0.99916666670905219]}
Q 666+930
T 1596
�[92m☑�[0m 1596
---
Q 21+26
T 47
�[92m☑�[0m 47
---
Q 70+245
T 315
�[92m☑�[0m 315
---
Q 927+65
T 992
�[92m☑�[0m 992
---
Q 614+16
T 630
�[92m☑�[0m 630
---
Q 799+40
T 839
�[92m☑�[0m 839
---
Q 11+28
T 39
�[92m☑�[0m 39
---
Q 429+86
T 515
�[92m☑�[0m 515
---
Q 295+94
T 389
�[92m☑�[0m 389
---
Q 728+5
T 733
�[92m☑�[0m 733
---
--------------------------------------------------
Iteration 41
Train on 45000 samples, validate on 5000 samples
Epoch 1/1
45000/45000 [==============================] - 38s - loss: 0.0347 - acc: 0.9901 - val_loss: 0.0987 - val_acc: 0.9671
{'val_loss': [0.098733154940605167], 'val_acc': [0.96714999999999995], 'loss': [0.034694219418366749], 'acc': [0.99013888886769608]}
Q 560+9
T 569
�[92m☑�[0m 569
---
Q 416+390
T 806
�[91m☒�[0m 706
---
Q 991+860
T 1851
�[92m☑�[0m 1851
---
Q 85+532
T 617
�[92m☑�[0m 617
---
Q 54+13
T 67
�[92m☑�[0m 67
---
Q 436+330
T 766
�[92m☑�[0m 766
---
Q 14+809
T 823
�[92m☑�[0m 823
---
Q 55+602
T 657
�[92m☑�[0m 657
---
Q 90+824
T 914
�[92m☑�[0m 914
---
Q 909+44
T 953
�[92m☑�[0m 953
---
--------------------------------------------------
Iteration 50
Train on 45000 samples, validate on 5000 samples
Epoch 1/1
45000/45000 [==============================] - 41s - loss: 0.0336 - acc: 0.9891 - val_loss: 0.0226 - val_acc: 0.9926
{'val_loss': [0.022579708782583474], 'val_acc': [0.99265000000000003], 'loss': [0.033597714668015637], 'acc': [0.98904999999999998]}
Q 82+44
T 126
�[92m☑�[0m 126
---
Q 97+434
T 531
�[92m☑�[0m 531
---
Q 157+85
T 242
�[92m☑�[0m 242
---
Q 951+319
T 1270
�[92m☑�[0m 1270
---
Q 131+0
T 131
�[92m☑�[0m 131
---
Q 80+935
T 1015
�[92m☑�[0m 1015
---
Q 313+41
T 354
�[92m☑�[0m 354
---
Q 393+11
T 404
�[92m☑�[0m 404
---
Q 214+197
T 411
�[92m☑�[0m 411
---
Q 371+1
T 372
�[92m☑�[0m 372
---
--------------------------------------------------
Iteration 80
Train on 45000 samples, validate on 5000 samples
Epoch 1/1
45000/45000 [==============================] - 38s - loss: 0.0309 - acc: 0.9907 - val_loss: 0.0110 - val_acc: 0.9969
{'val_loss': [0.011006494026631116], 'val_acc': [0.99690000000000001], 'loss': [0.030902873884472583], 'acc': [0.99073333335452607]}
Q 929+75
T 1004
�[92m☑�[0m 1004
---
Q 211+71
T 282
�[92m☑�[0m 282
---
Q 205+6
T 211
�[92m☑�[0m 211
---
Q 0+148
T 148
�[92m☑�[0m 148
---
Q 51+518
T 569
�[92m☑�[0m 569
---
Q 813+618
T 1431
�[92m☑�[0m 1431
---
Q 649+86
T 735
�[92m☑�[0m 735
---
Q 59+514
T 573
�[92m☑�[0m 573
---
Q 790+31
T 821
�[92m☑�[0m 821
---
Q 707+932
T 1639
�[92m☑�[0m 1639
---
--------------------------------------------------
Iteration 81
Train on 45000 samples, validate on 5000 samples
Epoch 1/1
45000/45000 [==============================] - 38s - loss: 0.0020 - acc: 0.9999 - val_loss: 0.0081 - val_acc: 0.9979
{'val_loss': [0.0081367643300443888], 'val_acc': [0.99785000000000001], 'loss': [0.0020035561124069822], 'acc': [0.99992777777777775]}
Q 245+34
T 279
�[92m☑�[0m 279
---
Q 2+78
T 80
�[92m☑�[0m 80
---
Q 63+391
T 454
�[92m☑�[0m 454
---
Q 45+888
T 933
�[92m☑�[0m 933
---
Q 28+653
T 681
�[92m☑�[0m 681
---
Q 45+826
T 871
�[92m☑�[0m 871
---
Q 33+814
T 847
�[92m☑�[0m 847
---
Q 552+978
T 1530
�[92m☑�[0m 1530
---
Q 802+2
T 804
�[92m☑�[0m 804
---
Q 22+538
T 560
�[92m☑�[0m 560
---
--------------------------------------------------
Iteration 82
Train on 45000 samples, validate on 5000 samples
Epoch 1/1
45000/45000 [==============================] - 38s - loss: 0.0012 - acc: 1.0000 - val_loss: 0.0069 - val_acc: 0.9979
{'val_loss': [0.0068942253075540069], 'val_acc': [0.99790000000000001], 'loss': [0.0012494151224899622], 'acc': [0.99998888888888893]}
Q 7+881
T 888
�[92m☑�[0m 888
---
Q 166+500
T 666
�[92m☑�[0m 666
---
Q 333+3
T 336
�[92m☑�[0m 336
---
Q 720+55
T 775
�[92m☑�[0m 775
---
Q 752+189
T 941
�[92m☑�[0m 941
---
Q 935+479
T 1414
�[92m☑�[0m 1414
---
Q 30+882
T 912
�[92m☑�[0m 912
---
Q 66+37
T 103
�[92m☑�[0m 103
---
Q 289+717
T 1006
�[91m☒�[0m 9006
---
Q 174+72
T 246
�[92m☑�[0m 246
---
--------------------------------------------------
Iteration 83
Train on 45000 samples, validate on 5000 samples
Epoch 1/1
45000/45000 [==============================] - 38s - loss: 0.0010 - acc: 1.0000 - val_loss: 0.0062 - val_acc: 0.9981
{'val_loss': [0.0062321711093187336], 'val_acc': [0.99814999999999998], 'loss': [0.0010194639238839348], 'acc': [0.99999444444444441]}
Q 604+393
T 997
�[92m☑�[0m 997
---
Q 206+9
T 215
�[92m☑�[0m 215
---
Q 708+38
T 746
�[92m☑�[0m 746
---
Q 31+434
T 465
�[92m☑�[0m 465
---
Q 240+48
T 288
�[92m☑�[0m 288
---
Q 772+75
T 847
�[92m☑�[0m 847
---
Q 979+445
T 1424
�[92m☑�[0m 1424
---
Q 591+168
T 759
�[92m☑�[0m 759
---
Q 55+7
T 62
�[92m☑�[0m 62
---
Q 60+877
T 937
�[92m☑�[0m 937
---
--------------------------------------------------
Iteration 84
Train on 45000 samples, validate on 5000 samples
Epoch 1/1
45000/45000 [==============================] - 39s - loss: 9.5497e-04 - acc: 1.0000 - val_loss: 0.0072 - val_acc: 0.9979
{'val_loss': [0.0072223034381866452], 'val_acc': [0.99785000000000001], 'loss': [0.00095497155957337883], 'acc': [0.99998333333333334]}
Q 91+25
T 116
�[92m☑�[0m 116
---
Q 429+86
T 515
�[92m☑�[0m 515
---
Q 5+218
T 223
�[92m☑�[0m 223
---
Q 5+583
T 588
�[92m☑�[0m 588
---
Q 84+362
T 446
�[92m☑�[0m 446
---
Q 428+4
T 432
�[92m☑�[0m 432
---
Q 35+768
T 803
�[92m☑�[0m 803
---
Q 430+21
T 451
�[92m☑�[0m 451
---
Q 95+486
T 581
�[92m☑�[0m 581
---
Q 0+642
T 642
�[92m☑�[0m 642
---
--------------------------------------------------
Iteration 85
Train on 45000 samples, validate on 5000 samples
Epoch 1/1
45000/45000 [==============================] - 39s - loss: 9.4473e-04 - acc: 1.0000 - val_loss: 0.0069 - val_acc: 0.9980
{'val_loss': [0.0069201477531343697], 'val_acc': [0.998], 'loss': [0.00094472687745259865], 'acc': [0.99998333333333334]}
Q 772+75
T 847
�[92m☑�[0m 847
---
Q 94+634
T 728
�[92m☑�[0m 728
---
Q 836+335
T 1171
�[92m☑�[0m 1171
---
Q 64+8
T 72
�[92m☑�[0m 72
---
Q 8+984
T 992
�[92m☑�[0m 992
---
Q 806+61
T 867
�[92m☑�[0m 867
---
Q 1+576
T 577
�[92m☑�[0m 577
---
Q 57+256
T 313
�[92m☑�[0m 313
---
Q 393+37
T 430
�[92m☑�[0m 430
---
Q 69+968
T 1037
�[92m☑�[0m 1037
---
--------------------------------------------------
Iteration 86
Train on 45000 samples, validate on 5000 samples
Epoch 1/1
45000/45000 [==============================] - 40s - loss: 0.0420 - acc: 0.9873 - val_loss: 0.0184 - val_acc: 0.9942
{'val_loss': [0.018363300597667696], 'val_acc': [0.99424999999999997], 'loss': [0.041988874252306088], 'acc': [0.98727777777777781]}
Q 636+4
T 640
�[92m☑�[0m 640
---
Q 63+439
T 502
�[92m☑�[0m 502
---
Q 245+34
T 279
�[92m☑�[0m 279
---
Q 574+387
T 961
�[92m☑�[0m 961
---
Q 802+70
T 872
�[92m☑�[0m 872
---
Q 659+234
T 893
�[92m☑�[0m 893
---
Q 29+505
T 534
�[92m☑�[0m 534
---
Q 87+217
T 304
�[92m☑�[0m 304
---
Q 849+988
T 1837
�[92m☑�[0m 1837
---
Q 9+874
T 883
�[92m☑�[0m 883
---
--------------------------------------------------
Iteration 87
Train on 45000 samples, validate on 5000 samples
Epoch 1/1
45000/45000 [==============================] - 40s - loss: 0.0031 - acc: 0.9997 - val_loss: 0.0095 - val_acc: 0.9975
{'val_loss': [0.0095259528324007983], 'val_acc': [0.99750000000000005], 'loss': [0.003114491027055515], 'acc': [0.99965000000000004]}
Q 659+447
T 1106
�[92m☑�[0m 1106
---
Q 39+614
T 653
�[92m☑�[0m 653
---
Q 550+776
T 1326
�[92m☑�[0m 1326
---
Q 825+66
T 891
�[92m☑�[0m 891
---
Q 848+344
T 1192
�[92m☑�[0m 1192
---
Q 823+648
T 1471
�[92m☑�[0m 1471
---
Q 241+364
T 605
�[92m☑�[0m 605
---
Q 3+283
T 286
�[92m☑�[0m 286
---
Q 177+87
T 264
�[92m☑�[0m 264
---
Q 770+995
T 1765
�[92m☑�[0m 1765
---
--------------------------------------------------
Iteration 88
Train on 45000 samples, validate on 5000 samples
Epoch 1/1
45000/45000 [==============================] - 40s - loss: 0.0014 - acc: 0.9999 - val_loss: 0.0088 - val_acc: 0.9973
{'val_loss': [0.0088253471940755845], 'val_acc': [0.99729999999999996], 'loss': [0.0014127093569686014], 'acc': [0.99994444444444441]}
Q 93+749
T 842
�[92m☑�[0m 842
---
Q 76+537
T 613
�[92m☑�[0m 613
---
Q 52+256
T 308
�[92m☑�[0m 308
---
Q 43+554
T 597
�[92m☑�[0m 597
---
Q 470+32
T 502
�[92m☑�[0m 502
---
Q 729+66
T 795
�[92m☑�[0m 795
---
Q 58+869
T 927
�[92m☑�[0m 927
---
Q 112+655
T 767
�[92m☑�[0m 767
---
Q 503+9
T 512
�[92m☑�[0m 512
---
Q 3+199
T 202
�[92m☑�[0m 202
---
--------------------------------------------------
Iteration 89
Train on 45000 samples, validate on 5000 samples
Epoch 1/1
45000/45000 [==============================] - 38s - loss: 0.0010 - acc: 1.0000 - val_loss: 0.0072 - val_acc: 0.9980
{'val_loss': [0.0071900495752692225], 'val_acc': [0.99804999999999999], 'loss': [0.001022424675565627], 'acc': [0.99998888888888893]}
Q 23+54
T 77
�[92m☑�[0m 77
---
Q 7+281
T 288
�[92m☑�[0m 288
---
Q 566+330
T 896
�[92m☑�[0m 896
---
Q 700+97
T 797
�[92m☑�[0m 797
---
Q 6+826
T 832
�[92m☑�[0m 832
---
Q 58+446
T 504
�[92m☑�[0m 504
---
Q 461+601
T 1062
�[92m☑�[0m 1062
---
Q 541+78
T 619
�[92m☑�[0m 619
---
Q 188+359
T 547
�[92m☑�[0m 547
---
Q 283+83
T 366
�[92m☑�[0m 366
---
--------------------------------------------------
Iteration 90
Train on 45000 samples, validate on 5000 samples
Epoch 1/1
45000/45000 [==============================] - 38s - loss: 8.3694e-04 - acc: 1.0000 - val_loss: 0.0066 - val_acc: 0.9981
{'val_loss': [0.0066044864896684886], 'val_acc': [0.99809999999999999], 'loss': [0.00083693527401321459], 'acc': [0.99999444444444441]}
Q 915+99
T 1014
�[92m☑�[0m 1014
---
Q 874+530
T 1404
�[92m☑�[0m 1404
---
Q 40+244
T 284
�[92m☑�[0m 284
---
Q 321+503
T 824
�[92m☑�[0m 824
---
Q 146+9
T 155
�[92m☑�[0m 155
---
Q 400+528
T 928
�[92m☑�[0m 928
---
Q 620+818
T 1438
�[92m☑�[0m 1438
---
Q 86+24
T 110
�[92m☑�[0m 110
---
Q 659+67
T 726
�[92m☑�[0m 726
---
Q 584+737
T 1321
�[92m☑�[0m 1321
---
--------------------------------------------------
Iteration 91
Train on 45000 samples, validate on 5000 samples
Epoch 1/1
45000/45000 [==============================] - 38s - loss: 7.5902e-04 - acc: 1.0000 - val_loss: 0.0068 - val_acc: 0.9981
{'val_loss': [0.0068030334025621416], 'val_acc': [0.99809999999999999], 'loss': [0.00075902411151263448], 'acc': [0.99998888888888893]}
Q 96+511
T 607
�[92m☑�[0m 607
---
Q 97+558
T 655
�[92m☑�[0m 655
---
Q 16+78
T 94
�[92m☑�[0m 94
---
Q 97+87
T 184
�[92m☑�[0m 184
---
Q 34+480
T 514
�[92m☑�[0m 514
---
Q 518+50
T 568
�[92m☑�[0m 568
---
Q 786+370
T 1156
�[92m☑�[0m 1156
---
Q 55+23
T 78
�[92m☑�[0m 78
---
Q 96+725
T 821
�[92m☑�[0m 821
---
Q 638+9
T 647
�[92m☑�[0m 647
---
--------------------------------------------------
Iteration 92
Train on 45000 samples, validate on 5000 samples
Epoch 1/1
45000/45000 [==============================] - 38s - loss: 0.0319 - acc: 0.9898 - val_loss: 0.0276 - val_acc: 0.9911
{'val_loss': [0.027552293083071708], 'val_acc': [0.99109999999999998], 'loss': [0.03191252367479934], 'acc': [0.98983333332273693]}
Q 58+497
T 555
�[92m☑�[0m 555
---
Q 71+40
T 111
�[92m☑�[0m 111
---
Q 0+175
T 175
�[92m☑�[0m 175
---
Q 600+97
T 697
�[92m☑�[0m 697
---
Q 940+115
T 1055
�[92m☑�[0m 1055
---
Q 25+863
T 888
�[92m☑�[0m 888
---
Q 480+82
T 562
�[92m☑�[0m 562
---
Q 463+2
T 465
�[92m☑�[0m 465
---
Q 177+47
T 224
�[92m☑�[0m 224
---
Q 455+83
T 538
�[92m☑�[0m 538
---
--------------------------------------------------
Iteration 93
Train on 45000 samples, validate on 5000 samples
Epoch 1/1
45000/45000 [==============================] - 39s - loss: 0.0047 - acc: 0.9990 - val_loss: 0.0096 - val_acc: 0.9969
{'val_loss': [0.009624259917996824], 'val_acc': [0.99690000000000001], 'loss': [0.0046642686388972737], 'acc': [0.99896666666666667]}
Q 4+459
T 463
�[92m☑�[0m 463
---
Q 846+408
T 1254
�[92m☑�[0m 1254
---
Q 678+15
T 693
�[92m☑�[0m 693
---
Q 85+474
T 559
�[92m☑�[0m 559
---
Q 573+12
T 585
�[92m☑�[0m 585
---
Q 437+243
T 680
�[92m☑�[0m 680
---
Q 224+34
T 258
�[92m☑�[0m 258
---
Q 96+588
T 684
�[92m☑�[0m 684
---
Q 410+41
T 451
�[92m☑�[0m 451
---
Q 18+210
T 228
�[92m☑�[0m 228
---
--------------------------------------------------
Iteration 94
Train on 45000 samples, validate on 5000 samples
Epoch 1/1
45000/45000 [==============================] - 39s - loss: 0.0012 - acc: 1.0000 - val_loss: 0.0066 - val_acc: 0.9982
{'val_loss': [0.0066092373952269558], 'val_acc': [0.99819999999999998], 'loss': [0.0011797052747259537], 'acc': [0.99998333333333334]}
Q 647+774
T 1421
�[92m☑�[0m 1421
---
Q 976+86
T 1062
�[92m☑�[0m 1062
---
Q 275+598
T 873
�[92m☑�[0m 873
---
Q 342+634
T 976
�[92m☑�[0m 976
---
Q 830+42
T 872
�[92m☑�[0m 872
---
Q 85+85
T 170
�[92m☑�[0m 170
---
Q 677+521
T 1198
�[92m☑�[0m 1198
---
Q 940+949
T 1889
�[92m☑�[0m 1889
---
Q 41+750
T 791
�[92m☑�[0m 791
---
Q 12+6
T 18
�[92m☑�[0m 18
---
--------------------------------------------------
Iteration 95
Train on 45000 samples, validate on 5000 samples
Epoch 1/1
45000/45000 [==============================] - 39s - loss: 8.5304e-04 - acc: 1.0000 - val_loss: 0.0059 - val_acc: 0.9982
{'val_loss': [0.0059119948847219349], 'val_acc': [0.99819999999999998], 'loss': [0.00085304205637011261], 'acc': [1.0]}
Q 13+57
T 70
�[92m☑�[0m 70
---
Q 534+71
T 605
�[92m☑�[0m 605
---
Q 762+703
T 1465
�[92m☑�[0m 1465
---
Q 136+8
T 144
�[92m☑�[0m 144
---
Q 707+932
T 1639
�[92m☑�[0m 1639
---
Q 31+51
T 82
�[92m☑�[0m 82
---
Q 2+804
T 806
�[92m☑�[0m 806
---
Q 823+648
T 1471
�[92m☑�[0m 1471
---
Q 655+86
T 741
�[92m☑�[0m 741
---
Q 612+799
T 1411
�[92m☑�[0m 1411
---
--------------------------------------------------
Iteration 96
Train on 45000 samples, validate on 5000 samples
Epoch 1/1
45000/45000 [==============================] - 40s - loss: 0.0021 - acc: 0.9996 - val_loss: 0.0922 - val_acc: 0.9789
{'val_loss': [0.092193102195858953], 'val_acc': [0.97894999999999999], 'loss': [0.0021130382461680306], 'acc': [0.99959444446563717]}
Q 97+558
T 655
�[92m☑�[0m 655
---
Q 183+583
T 766
�[92m☑�[0m 766
---
Q 20+32
T 52
�[92m☑�[0m 52
---
Q 26+58
T 84
�[92m☑�[0m 84
---
Q 166+66
T 232
�[92m☑�[0m 232
---
Q 98+523
T 621
�[92m☑�[0m 621
---
Q 92+673
T 765
�[92m☑�[0m 765
---
Q 12+565
T 577
�[92m☑�[0m 577
---
Q 460+97
T 557
�[92m☑�[0m 557
---
Q 703+17
T 720
�[92m☑�[0m 720
---
--------------------------------------------------
Iteration 97
Train on 45000 samples, validate on 5000 samples
Epoch 1/1
45000/45000 [==============================] - 40s - loss: 0.0276 - acc: 0.9914 - val_loss: 0.0154 - val_acc: 0.9949
{'val_loss': [0.015444085666537285], 'val_acc': [0.99495], 'loss': [0.027583559429811107], 'acc': [0.99141666666666661]}
Q 535+32
T 567
�[92m☑�[0m 567
---
Q 863+966
T 1829
�[92m☑�[0m 1829
---
Q 699+374
T 1073
�[92m☑�[0m 1073
---
Q 29+190
T 219
�[92m☑�[0m 219
---
Q 1+994
T 995
�[92m☑�[0m 995
---
Q 422+84
T 506
�[92m☑�[0m 506
---
Q 633+525
T 1158
�[92m☑�[0m 1158
---
Q 604+376
T 980
�[92m☑�[0m 980
---
Q 863+966
T 1829
�[92m☑�[0m 1829
---
Q 874+530
T 1404
�[92m☑�[0m 1404
---
--------------------------------------------------
Iteration 98
Train on 45000 samples, validate on 5000 samples
Epoch 1/1
45000/45000 [==============================] - 40s - loss: 0.0018 - acc: 0.9999 - val_loss: 0.0067 - val_acc: 0.9979
{'val_loss': [0.0066717502040788534], 'val_acc': [0.99790000000000001], 'loss': [0.001755557962672578], 'acc': [0.99988888888888894]}
Q 561+65
T 626
�[92m☑�[0m 626
---
Q 979+81
T 1060
�[92m☑�[0m 1060
---
Q 361+78
T 439
�[92m☑�[0m 439
---
Q 3+104
T 107
�[92m☑�[0m 107
---
Q 104+974
T 1078
�[92m☑�[0m 1078
---
Q 660+545
T 1205
�[92m☑�[0m 1205
---
Q 72+243
T 315
�[92m☑�[0m 315
---
Q 99+617
T 716
�[92m☑�[0m 716
---
Q 567+55
T 622
�[92m☑�[0m 622
---
Q 290+286
T 576
�[92m☑�[0m 576
---
--------------------------------------------------
Iteration 99
Train on 45000 samples, validate on 5000 samples
Epoch 1/1
45000/45000 [==============================] - 39s - loss: 8.7092e-04 - acc: 1.0000 - val_loss: 0.0058 - val_acc: 0.9984
{'val_loss': [0.0057973034713417288], 'val_acc': [0.99839999999999995], 'loss': [0.00087091847145929937], 'acc': [0.99998888888888893]}
Q 95+638
T 733
�[92m☑�[0m 733
---
Q 407+26
T 433
�[92m☑�[0m 433
---
Q 565+786
T 1351
�[92m☑�[0m 1351
---
Q 505+227
T 732
�[92m☑�[0m 732
---
Q 778+309
T 1087
�[92m☑�[0m 1087
---
Q 929+138
T 1067
�[92m☑�[0m 1067
---
Q 436+330
T 766
�[92m☑�[0m 766
---
Q 356+55
T 411
�[92m☑�[0m 411
---
Q 276+195
T 471
�[92m☑�[0m 471
---
Q 149+962
T 1111
�[92m☑�[0m 1111
---
--------------------------------------------------
Iteration 100
Train on 45000 samples, validate on 5000 samples
Epoch 1/1
45000/45000 [==============================] - 37s - loss: 7.1227e-04 - acc: 1.0000 - val_loss: 0.0054 - val_acc: 0.9983
{'val_loss': [0.0053955351205542687], 'val_acc': [0.99834999999999996], 'loss': [0.00071227497477084395], 'acc': [0.99999444444444441]}
Q 31+437
T 468
�[92m☑�[0m 468
---
Q 426+38
T 464
�[92m☑�[0m 464
---
Q 941+842
T 1783
�[92m☑�[0m 1783
---
Q 323+463
T 786
�[92m☑�[0m 786
---
Q 646+72
T 718
�[92m☑�[0m 718
---
Q 834+296
T 1130
�[92m☑�[0m 1130
---
Q 688+214
T 902
�[91m☒�[0m 802
---
Q 349+669
T 1018
�[92m☑�[0m 1018
---
Q 232+67
T 299
�[92m☑�[0m 299
---
Q 613+309
T 922
�[92m☑�[0m 922
---
--------------------------------------------------
Iteration 101
Train on 45000 samples, validate on 5000 samples
Epoch 1/1
45000/45000 [==============================] - 37s - loss: 6.9362e-04 - acc: 1.0000 - val_loss: 0.0060 - val_acc: 0.9980
{'val_loss': [0.0059655166847631339], 'val_acc': [0.99804999999999999], 'loss': [0.00069362363646634749], 'acc': [0.99998333333333334]}
Q 472+332
T 804
�[92m☑�[0m 804
---
Q 8+636
T 644
�[92m☑�[0m 644
---
Q 258+138
T 396
�[91m☒�[0m 496
---
Q 873+788
T 1661
�[91m☒�[0m 1671
---
Q 238+452
T 690
�[92m☑�[0m 690
---
Q 932+11
T 943
�[92m☑�[0m 943
---
Q 2+74
T 76
�[92m☑�[0m 76
---
Q 33+633
T 666
�[92m☑�[0m 666
---
Q 971+81
T 1052
�[92m☑�[0m 1052
---
Q 5+218
T 223
�[92m☑�[0m 223
---
--------------------------------------------------
Iteration 102
Train on 45000 samples, validate on 5000 samples
Epoch 1/1
45000/45000 [==============================] - 37s - loss: 0.0038 - acc: 0.9990 - val_loss: 0.0370 - val_acc: 0.9884
{'val_loss': [0.036963613131642342], 'val_acc': [0.98839999999999995], 'loss': [0.0037796633920735784], 'acc': [0.99903888891008163]}
Q 769+8
T 777
�[92m☑�[0m 777
---
Q 982+72
T 1054
�[92m☑�[0m 1054
---
Q 803+148
T 951
�[92m☑�[0m 951
---
Q 768+63
T 831
�[92m☑�[0m 831
---
Q 711+580
T 1291
�[92m☑�[0m 1291
---
Q 901+516
T 1417
�[92m☑�[0m 1417
---
Q 383+56
T 439
�[92m☑�[0m 439
---
Q 9+951
T 960
�[92m☑�[0m 960
---
Q 91+594
T 685
�[92m☑�[0m 685
---
Q 708+425
T 1133
�[92m☑�[0m 1133
---
--------------------------------------------------
Iteration 103
Train on 45000 samples, validate on 5000 samples
Epoch 1/1
45000/45000 [==============================] - 38s - loss: 0.0352 - acc: 0.9895 - val_loss: 0.0113 - val_acc: 0.9969
{'val_loss': [0.011299159627594054], 'val_acc': [0.99690000000000001], 'loss': [0.035152347180123132], 'acc': [0.9894722222222222]}
Q 537+648
T 1185
�[92m☑�[0m 1185
---
Q 244+8
T 252
�[92m☑�[0m 252
---
Q 277+351
T 628
�[92m☑�[0m 628
---
Q 967+82
T 1049
�[92m☑�[0m 1049
---
Q 496+259
T 755
�[92m☑�[0m 755
---
Q 165+696
T 861
�[92m☑�[0m 861
---
Q 212+8
T 220
�[92m☑�[0m 220
---
Q 215+71
T 286
�[92m☑�[0m 286
---
Q 75+586
T 661
�[92m☑�[0m 661
---
Q 51+518
T 569
�[92m☑�[0m 569
---
--------------------------------------------------
Iteration 104
Train on 45000 samples, validate on 5000 samples
Epoch 1/1
45000/45000 [==============================] - 38s - loss: 0.0015 - acc: 0.9999 - val_loss: 0.0069 - val_acc: 0.9980
{'val_loss': [0.0068515809554606675], 'val_acc': [0.99804999999999999], 'loss': [0.0014726223781704902], 'acc': [0.99992222222222227]}
Q 73+53
T 126
�[92m☑�[0m 126
---
Q 9+184
T 193
�[92m☑�[0m 193
---
Q 226+73
T 299
�[92m☑�[0m 299
---
Q 110+83
T 193
�[92m☑�[0m 193
---
Q 235+9
T 244
�[92m☑�[0m 244
---
Q 2+667
T 669
�[92m☑�[0m 669
---
Q 203+4
T 207
�[92m☑�[0m 207
---
Q 45+2
T 47
�[92m☑�[0m 47
---
Q 2+804
T 806
�[92m☑�[0m 806
---
Q 398+628
T 1026
�[92m☑�[0m 1026
---
--------------------------------------------------
Iteration 105
Train on 45000 samples, validate on 5000 samples
Epoch 1/1
45000/45000 [==============================] - 38s - loss: 8.6203e-04 - acc: 1.0000 - val_loss: 0.0056 - val_acc: 0.9985
{'val_loss': [0.005607170724496245], 'val_acc': [0.99850000000000005], 'loss': [0.00086202879945437112], 'acc': [0.99997777777777774]}
Q 751+229
T 980
�[92m☑�[0m 980
---
Q 725+134
T 859
�[92m☑�[0m 859
---
Q 0+674
T 674
�[92m☑�[0m 674
---
Q 29+687
T 716
�[92m☑�[0m 716
---
Q 428+9
T 437
�[92m☑�[0m 437
---
Q 93+847
T 940
�[92m☑�[0m 940
---
Q 29+922
T 951
�[92m☑�[0m 951
---
Q 24+135
T 159
�[92m☑�[0m 159
---
Q 217+818
T 1035
�[92m☑�[0m 1035
---
Q 60+974
T 1034
�[92m☑�[0m 1034
---
--------------------------------------------------
Iteration 106
Train on 45000 samples, validate on 5000 samples
Epoch 1/1
45000/45000 [==============================] - 37s - loss: 7.0560e-04 - acc: 1.0000 - val_loss: 0.0053 - val_acc: 0.9987
{'val_loss': [0.0052594130329787735], 'val_acc': [0.99865000000000004], 'loss': [0.00070560133192274302], 'acc': [0.99998333333333334]}
Q 95+349
T 444
�[92m☑�[0m 444
---
Q 680+50
T 730
�[92m☑�[0m 730
---
Q 8+285
T 293
�[92m☑�[0m 293
---
Q 93+67
T 160
�[92m☑�[0m 160
---
Q 905+13
T 918
�[92m☑�[0m 918
---
Q 153+2
T 155
�[92m☑�[0m 155
---
Q 20+99
T 119
�[92m☑�[0m 119
---
Q 281+3
T 284
�[92m☑�[0m 284
---
Q 55+952
T 1007
�[92m☑�[0m 1007
---
Q 660+12
T 672
�[92m☑�[0m 672
---
--------------------------------------------------
Iteration 107
Train on 45000 samples, validate on 5000 samples
Epoch 1/1
45000/45000 [==============================] - 37s - loss: 5.6956e-04 - acc: 1.0000 - val_loss: 0.0053 - val_acc: 0.9987
{'val_loss': [0.0053374502418562769], 'val_acc': [0.99870000000000003], 'loss': [0.00056955946002983386], 'acc': [1.0]}
Q 87+37
T 124
�[92m☑�[0m 124
---
Q 6+968
T 974
�[92m☑�[0m 974
---
Q 1+733
T 734
�[92m☑�[0m 734
---
Q 2+78
T 80
�[92m☑�[0m 80
---
Q 61+260
T 321
�[92m☑�[0m 321
---
Q 374+94
T 468
�[92m☑�[0m 468
---
Q 25+582
T 607
�[92m☑�[0m 607
---
Q 62+919
T 981
�[92m☑�[0m 981
---
Q 522+0
T 522
�[92m☑�[0m 522
---
Q 963+61
T 1024
�[92m☑�[0m 1024
---
--------------------------------------------------
Iteration 108
Train on 45000 samples, validate on 5000 samples
Epoch 1/1
45000/45000 [==============================] - 37s - loss: 5.5234e-04 - acc: 1.0000 - val_loss: 0.0056 - val_acc: 0.9984
{'val_loss': [0.0056324227192439142], 'val_acc': [0.99844999999999995], 'loss': [0.00055234070724497234], 'acc': [0.99999444444444441]}
Q 560+1
T 561
�[92m☑�[0m 561
---
Q 58+801
T 859
�[92m☑�[0m 859
---
Q 41+592
T 633
�[92m☑�[0m 633
---
Q 486+317
T 803
�[92m☑�[0m 803
---
Q 467+312
T 779
�[92m☑�[0m 779
---
Q 146+0
T 146
�[92m☑�[0m 146
---
Q 4+448
T 452
�[92m☑�[0m 452
---
Q 5+583
T 588
�[92m☑�[0m 588
---
Q 233+91
T 324
�[92m☑�[0m 324
---
Q 20+128
T 148
�[92m☑�[0m 148
---
--------------------------------------------------
Iteration 109
Train on 45000 samples, validate on 5000 samples
Epoch 1/1
45000/45000 [==============================] - 37s - loss: 0.0199 - acc: 0.9938 - val_loss: 0.1282 - val_acc: 0.9617
{'val_loss': [0.12818750436902046], 'val_acc': [0.9617], 'loss': [0.019872152358955807], 'acc': [0.99382777778837417]}
Q 0+642
T 642
�[92m☑�[0m 642
---
Q 115+3
T 118
�[92m☑�[0m 118
---
Q 5+269
T 274
�[92m☑�[0m 274
---
Q 175+23
T 198
�[92m☑�[0m 198
---
Q 85+267
T 352
�[92m☑�[0m 352
---
Q 735+74
T 809
�[92m☑�[0m 809
---
Q 9+145
T 154
�[92m☑�[0m 154
---
Q 56+24
T 80
�[92m☑�[0m 80
---
Q 942+3
T 945
�[92m☑�[0m 945
---
Q 821+18
T 839
�[92m☑�[0m 839
---
--------------------------------------------------
Iteration 110
Train on 45000 samples, validate on 5000 samples
Epoch 1/1
45000/45000 [==============================] - 37s - loss: 0.0168 - acc: 0.9950 - val_loss: 0.0078 - val_acc: 0.9976
{'val_loss': [0.0077659709699451924], 'val_acc': [0.99765000000000004], 'loss': [0.016768622599252395], 'acc': [0.99503333333333333]}
Q 938+374
T 1312
�[92m☑�[0m 1312
---
Q 93+67
T 160
�[92m☑�[0m 160
---
Q 657+810
T 1467
�[92m☑�[0m 1467
---
Q 915+7
T 922
�[92m☑�[0m 922
---
Q 80+86
T 166
�[92m☑�[0m 166
---
Q 528+585
T 1113
�[92m☑�[0m 1113
---
Q 2+395
T 397
�[92m☑�[0m 397
---
Q 667+5
T 672
�[92m☑�[0m 672
---
Q 748+96
T 844
�[92m☑�[0m 844
---
Q 383+53
T 436
�[92m☑�[0m 436
---
--------------------------------------------------
Iteration 120
Train on 45000 samples, validate on 5000 samples
Epoch 1/1
45000/45000 [==============================] - 37s - loss: 5.8026e-04 - acc: 1.0000 - val_loss: 0.0044 - val_acc: 0.9987
{'val_loss': [0.0043738329783082012], 'val_acc': [0.99870000000000003], 'loss': [0.00058025503645961485], 'acc': [1.0]}
Q 442+17
T 459
�[92m☑�[0m 459
---
Q 54+240
T 294
�[92m☑�[0m 294
---
Q 51+93
T 144
�[92m☑�[0m 144
---
Q 835+86
T 921
�[92m☑�[0m 921
---
Q 816+79
T 895
�[92m☑�[0m 895
---
Q 771+166
T 937
�[92m☑�[0m 937
---
Q 167+7
T 174
�[92m☑�[0m 174
---
Q 80+935
T 1015
�[92m☑�[0m 1015
---
Q 2+201
T 203
�[92m☑�[0m 203
---
Q 6+893
T 899
�[92m☑�[0m 899
---
--------------------------------------------------
Iteration 121
Train on 45000 samples, validate on 5000 samples
Epoch 1/1
45000/45000 [==============================] - 37s - loss: 4.9366e-04 - acc: 1.0000 - val_loss: 0.0042 - val_acc: 0.9988
{'val_loss': [0.0042192966386675832], 'val_acc': [0.99880000000000002], 'loss': [0.00049365888878496154], 'acc': [1.0]}
Q 276+579
T 855
�[92m☑�[0m 855
---
Q 86+94
T 180
�[92m☑�[0m 180
---
Q 524+95
T 619
�[92m☑�[0m 619
---
Q 610+635
T 1245
�[92m☑�[0m 1245
---
Q 994+77
T 1071
�[92m☑�[0m 1071
---
Q 49+232
T 281
�[92m☑�[0m 281
---
Q 782+50
T 832
�[92m☑�[0m 832
---
Q 49+59
T 108
�[92m☑�[0m 108
---
Q 98+700
T 798
�[92m☑�[0m 798
---
Q 515+67
T 582
�[92m☑�[0m 582
---
--------------------------------------------------
Iteration 122
Train on 45000 samples, validate on 5000 samples
Epoch 1/1
45000/45000 [==============================] - 37s - loss: 4.3406e-04 - acc: 1.0000 - val_loss: 0.0045 - val_acc: 0.9984
{'val_loss': [0.0045446309591643513], 'val_acc': [0.99844999999999995], 'loss': [0.00043406057129614055], 'acc': [1.0]}
Q 296+31
T 327
�[92m☑�[0m 327
---
Q 95+638
T 733
�[92m☑�[0m 733
---
Q 312+67
T 379
�[92m☑�[0m 379
---
Q 198+894
T 1092
�[91m☒�[0m 1192
---
Q 541+78
T 619
�[92m☑�[0m 619
---
Q 130+95
T 225
�[92m☑�[0m 225
---
Q 14+285
T 299
�[92m☑�[0m 299
---
Q 783+326
T 1109
�[92m☑�[0m 1109
---
Q 2+711
T 713
�[92m☑�[0m 713
---
Q 416+771
T 1187
�[92m☑�[0m 1187
---
--------------------------------------------------
Iteration 123
Train on 45000 samples, validate on 5000 samples
Epoch 1/1
45000/45000 [==============================] - 37s - loss: 4.0240e-04 - acc: 1.0000 - val_loss: 0.0040 - val_acc: 0.9987
{'val_loss': [0.0040482636697590354], 'val_acc': [0.99870000000000003], 'loss': [0.00040240162641016976], 'acc': [1.0]}
Q 1+988
T 989
�[92m☑�[0m 989
---
Q 516+65
T 581
�[92m☑�[0m 581
---
Q 72+208
T 280
�[92m☑�[0m 280
---
Q 65+819
T 884
�[92m☑�[0m 884
---
Q 496+37
T 533
�[92m☑�[0m 533
---
Q 87+87
T 174
�[92m☑�[0m 174
---
Q 713+3
T 716
�[92m☑�[0m 716
---
Q 9+468
T 477
�[92m☑�[0m 477
---
Q 740+5
T 745
�[92m☑�[0m 745
---
Q 404+81
T 485
�[92m☑�[0m 485
---
--------------------------------------------------
Iteration 130
Train on 45000 samples, validate on 5000 samples
Epoch 1/1
45000/45000 [==============================] - 38s - loss: 3.7717e-04 - acc: 1.0000 - val_loss: 0.0042 - val_acc: 0.9986
{'val_loss': [0.0041852503305301074], 'val_acc': [0.99860000000000004], 'loss': [0.00037716550330320993], 'acc': [1.0]}
Q 11+425
T 436
�[92m☑�[0m 436
---
Q 816+72
T 888
�[92m☑�[0m 888
---
Q 411+39
T 450
�[92m☑�[0m 450
---
Q 965+855
T 1820
�[92m☑�[0m 1820
---
Q 0+663
T 663
�[92m☑�[0m 663
---
Q 850+76
T 926
�[92m☑�[0m 926
---
Q 8+399
T 407
�[92m☑�[0m 407
---
Q 157+87
T 244
�[92m☑�[0m 244
---
Q 329+976
T 1305
�[92m☑�[0m 1305
---
Q 1+656
T 657
�[92m☑�[0m 657
---
--------------------------------------------------
Iteration 131
Train on 45000 samples, validate on 5000 samples
Epoch 1/1
45000/45000 [==============================] - 38s - loss: 3.4509e-04 - acc: 1.0000 - val_loss: 0.0043 - val_acc: 0.9986
{'val_loss': [0.00431418531849049], 'val_acc': [0.99855000000000005], 'loss': [0.00034508801567264728], 'acc': [1.0]}
Q 848+95
T 943
�[92m☑�[0m 943
---
Q 98+996
T 1094
�[92m☑�[0m 1094
---
Q 109+3
T 112
�[92m☑�[0m 112
---
Q 165+696
T 861
�[92m☑�[0m 861
---
Q 50+36
T 86
�[92m☑�[0m 86
---
Q 415+530
T 945
�[92m☑�[0m 945
---
Q 25+68
T 93
�[92m☑�[0m 93
---
Q 44+600
T 644
�[92m☑�[0m 644
---
Q 8+256
T 264
�[92m☑�[0m 264
---
Q 238+452
T 690
�[92m☑�[0m 690
---
--------------------------------------------------
Iteration 132
Train on 45000 samples, validate on 5000 samples
Epoch 1/1
45000/45000 [==============================] - 38s - loss: 3.0463e-04 - acc: 1.0000 - val_loss: 0.0043 - val_acc: 0.9989
{'val_loss': [0.0042662992852739992], 'val_acc': [0.99885000000000002], 'loss': [0.00030463080654541652], 'acc': [1.0]}
Q 20+675
T 695
�[92m☑�[0m 695
---
Q 416+688
T 1104
�[92m☑�[0m 1104
---
Q 7+13
T 20
�[92m☑�[0m 20
---
Q 61+260
T 321
�[92m☑�[0m 321
---
Q 6+459
T 465
�[92m☑�[0m 465
---
Q 920+784
T 1704
�[92m☑�[0m 1704
---
Q 40+270
T 310
�[92m☑�[0m 310
---
Q 601+492
T 1093
�[92m☑�[0m 1093
---
Q 4+76
T 80
�[92m☑�[0m 80
---
Q 2+690
T 692
�[92m☑�[0m 692
---
--------------------------------------------------
Iteration 140
Train on 45000 samples, validate on 5000 samples
Epoch 1/1
45000/45000 [==============================] - 38s - loss: 2.9599e-04 - acc: 1.0000 - val_loss: 0.0040 - val_acc: 0.9987
{'val_loss': [0.004036861338280141], 'val_acc': [0.99870000000000003], 'loss': [0.00029598567740370829], 'acc': [1.0]}
Q 64+83
T 147
�[92m☑�[0m 147
---
Q 639+76
T 715
�[92m☑�[0m 715
---
Q 48+206
T 254
�[92m☑�[0m 254
---
Q 54+453
T 507
�[92m☑�[0m 507
---
Q 340+19
T 359
�[92m☑�[0m 359
---
Q 522+958
T 1480
�[92m☑�[0m 1480
---
Q 256+58
T 314
�[92m☑�[0m 314
---
Q 2+458
T 460
�[92m☑�[0m 460
---
Q 750+450
T 1200
�[92m☑�[0m 1200
---
Q 845+549
T 1394
�[92m☑�[0m 1394
---
--------------------------------------------------
Iteration 141
Train on 45000 samples, validate on 5000 samples
Epoch 1/1
45000/45000 [==============================] - 38s - loss: 0.0229 - acc: 0.9937 - val_loss: 0.1356 - val_acc: 0.9604
{'val_loss': [0.13560998322889209], 'val_acc': [0.96040000000000003], 'loss': [0.022925767360006771], 'acc': [0.99366111115349665]}
Q 49+30
T 79
�[92m☑�[0m 79
---
Q 41+720
T 761
�[92m☑�[0m 761
---
Q 854+27
T 881
�[92m☑�[0m 881
---
Q 589+54
T 643
�[92m☑�[0m 643
---
Q 840+2
T 842
�[91m☒�[0m 852
---
Q 440+95
T 535
�[91m☒�[0m 545
---
Q 7+864
T 871
�[92m☑�[0m 871
---
Q 713+95
T 808
�[91m☒�[0m 898
---
Q 587+29
T 616
�[92m☑�[0m 616
---
Q 752+61
T 813
�[92m☑�[0m 813
---
--------------------------------------------------
Iteration 150
Train on 45000 samples, validate on 5000 samples
Epoch 1/1
45000/45000 [==============================] - 38s - loss: 4.3214e-04 - acc: 1.0000 - val_loss: 0.0037 - val_acc: 0.9987
{'val_loss': [0.0037493921197950838], 'val_acc': [0.99875000000000003], 'loss': [0.00043214037894374793], 'acc': [0.99998888888888893]}
Q 704+36
T 740
�[92m☑�[0m 740
---
Q 429+82
T 511
�[92m☑�[0m 511
---
Q 341+630
T 971
�[92m☑�[0m 971
---
Q 270+41
T 311
�[92m☑�[0m 311
---
Q 264+4
T 268
�[92m☑�[0m 268
---
Q 113+46
T 159
�[92m☑�[0m 159
---
Q 210+527
T 737
�[92m☑�[0m 737
---
Q 92+673
T 765
�[92m☑�[0m 765
---
Q 180+895
T 1075
�[92m☑�[0m 1075
---
Q 416+29
T 445
�[92m☑�[0m 445
---
--------------------------------------------------
Iteration 151
Train on 45000 samples, validate on 5000 samples
Epoch 1/1
45000/45000 [==============================] - 38s - loss: 3.4723e-04 - acc: 1.0000 - val_loss: 0.0034 - val_acc: 0.9991
{'val_loss': [0.0034389279445633291], 'val_acc': [0.99909999999999999], 'loss': [0.00034723099155558479], 'acc': [1.0]}
Q 774+32
T 806
�[92m☑�[0m 806
---
Q 107+799
T 906
�[92m☑�[0m 906
---
Q 54+629
T 683
�[92m☑�[0m 683
---
Q 38+678
T 716
�[92m☑�[0m 716
---
Q 863+656
T 1519
�[92m☑�[0m 1519
---
Q 580+829
T 1409
�[92m☑�[0m 1409
---
Q 80+935
T 1015
�[92m☑�[0m 1015
---
Q 3+964
T 967
�[92m☑�[0m 967
---
Q 485+983
T 1468
�[92m☑�[0m 1468
---
Q 69+943
T 1012
�[92m☑�[0m 1012
---
--------------------------------------------------
Iteration 160
Train on 45000 samples, validate on 5000 samples
Epoch 1/1
45000/45000 [==============================] - 37s - loss: 3.6144e-04 - acc: 1.0000 - val_loss: 0.0032 - val_acc: 0.9991
{'val_loss': [0.0031691691763699056], 'val_acc': [0.99909999999999999], 'loss': [0.00036144207823866355], 'acc': [1.0]}
Q 8+285
T 293
�[92m☑�[0m 293
---
Q 98+36
T 134
�[92m☑�[0m 134
---
Q 593+791
T 1384
�[92m☑�[0m 1384
---
Q 987+28
T 1015
�[92m☑�[0m 1015
---
Q 6+845
T 851
�[92m☑�[0m 851
---
Q 239+18
T 257
�[92m☑�[0m 257
---
Q 607+2
T 609
�[92m☑�[0m 609
---
Q 13+194
T 207
�[92m☑�[0m 207
---
Q 929+75
T 1004
�[92m☑�[0m 1004
---
Q 356+2
T 358
�[92m☑�[0m 358
---
--------------------------------------------------
Iteration 161
Train on 45000 samples, validate on 5000 samples
Epoch 1/1
45000/45000 [==============================] - 37s - loss: 3.1419e-04 - acc: 1.0000 - val_loss: 0.0033 - val_acc: 0.9990
{'val_loss': [0.0032845419609919191], 'val_acc': [0.999], 'loss': [0.00031418984433015189], 'acc': [1.0]}
Q 20+316
T 336
�[92m☑�[0m 336
---
Q 567+289
T 856
�[92m☑�[0m 856
---
Q 627+290
T 917
�[92m☑�[0m 917
---
Q 23+182
T 205
�[92m☑�[0m 205
---
Q 40+655
T 695
�[92m☑�[0m 695
---
Q 60+632
T 692
�[92m☑�[0m 692
---
Q 297+45
T 342
�[92m☑�[0m 342
---
Q 37+229
T 266
�[92m☑�[0m 266
---
Q 456+982
T 1438
�[92m☑�[0m 1438
---
Q 734+863
T 1597
�[92m☑�[0m 1597
---
--------------------------------------------------
Iteration 162
Train on 45000 samples, validate on 5000 samples
Epoch 1/1
45000/45000 [==============================] - 37s - loss: 2.7604e-04 - acc: 1.0000 - val_loss: 0.0032 - val_acc: 0.9991
{'val_loss': [0.0031552912279963494], 'val_acc': [0.99909999999999999], 'loss': [0.00027604213720187546], 'acc': [1.0]}
Q 0+852
T 852
�[92m☑�[0m 852
---
Q 90+20
T 110
�[92m☑�[0m 110
---
Q 82+19
T 101
�[92m☑�[0m 101
---
Q 72+104
T 176
�[92m☑�[0m 176
---
Q 71+832
T 903
�[92m☑�[0m 903
---
Q 7+599
T 606
�[92m☑�[0m 606
---
Q 723+453
T 1176
�[92m☑�[0m 1176
---
Q 409+42
T 451
�[92m☑�[0m 451
---
Q 104+948
T 1052
�[92m☑�[0m 1052
---
Q 364+69
T 433
�[92m☑�[0m 433
---
--------------------------------------------------
Iteration 182
Train on 45000 samples, validate on 5000 samples
Epoch 1/1
45000/45000 [==============================] - 38s - loss: 3.4842e-04 - acc: 1.0000 - val_loss: 0.0040 - val_acc: 0.9989
{'val_loss': [0.0039902217447757718], 'val_acc': [0.99890000000000001], 'loss': [0.00034841958908364177], 'acc': [1.0]}
Q 1+887
T 888
�[92m☑�[0m 888
---
Q 51+71
T 122
�[92m☑�[0m 122
---
Q 228+83
T 311
�[92m☑�[0m 311
---
Q 569+32
T 601
�[92m☑�[0m 601
---
Q 8+198
T 206
�[92m☑�[0m 206
---
Q 39+448
T 487
�[92m☑�[0m 487
---
Q 410+471
T 881
�[92m☑�[0m 881
---
Q 22+436
T 458
�[92m☑�[0m 458
---
Q 462+996
T 1458
�[92m☑�[0m 1458
---
Q 74+420
T 494
�[92m☑�[0m 494
---
--------------------------------------------------
Iteration 183
Train on 45000 samples, validate on 5000 samples
Epoch 1/1
45000/45000 [==============================] - 38s - loss: 2.9004e-04 - acc: 1.0000 - val_loss: 0.0040 - val_acc: 0.9988
{'val_loss': [0.0039785686612129213], 'val_acc': [0.99880000000000002], 'loss': [0.00029003697105476424], 'acc': [1.0]}
Q 654+936
T 1590
�[92m☑�[0m 1590
---
Q 6+530
T 536
�[92m☑�[0m 536
---
Q 982+460
T 1442
�[92m☑�[0m 1442
---
Q 286+407
T 693
�[92m☑�[0m 693
---
Q 169+20
T 189
�[92m☑�[0m 189
---
Q 56+848
T 904
�[92m☑�[0m 904
---
Q 912+8
T 920
�[92m☑�[0m 920
---
Q 16+942
T 958
�[92m☑�[0m 958
---
Q 454+480
T 934
�[92m☑�[0m 934
---
Q 8+853
T 861
�[92m☑�[0m 861
---
--------------------------------------------------
Iteration 192
Train on 45000 samples, validate on 5000 samples
Epoch 1/1
45000/45000 [==============================] - 38s - loss: 0.0148 - acc: 0.9955 - val_loss: 0.0072 - val_acc: 0.9978
{'val_loss': [0.0072148059517145157], 'val_acc': [0.99780000000000002], 'loss': [0.014829264489975241], 'acc': [0.99554444444444445]}
Q 3+39
T 42
�[92m☑�[0m 42
---
Q 13+458
T 471
�[92m☑�[0m 471
---
Q 18+210
T 228
�[92m☑�[0m 228
---
Q 71+872
T 943
�[92m☑�[0m 943
---
Q 166+500
T 666
�[92m☑�[0m 666
---
Q 57+606
T 663
�[92m☑�[0m 663
---
Q 705+34
T 739
�[92m☑�[0m 739
---
Q 6+413
T 419
�[92m☑�[0m 419
---
Q 484+648
T 1132
�[92m☑�[0m 1132
---
Q 734+2
T 736
�[92m☑�[0m 736
---
--------------------------------------------------
Iteration 193
Train on 45000 samples, validate on 5000 samples
Epoch 1/1
45000/45000 [==============================] - 38s - loss: 8.2297e-04 - acc: 1.0000 - val_loss: 0.0036 - val_acc: 0.9989
{'val_loss': [0.0035778338901698591], 'val_acc': [0.99890000000000001], 'loss': [0.00082297382206759516], 'acc': [0.9999555555555556]}
Q 314+9
T 323
�[92m☑�[0m 323
---
Q 320+106
T 426
�[92m☑�[0m 426
---
Q 591+168
T 759
�[92m☑�[0m 759
---
Q 100+4
T 104
�[92m☑�[0m 104
---
Q 37+471
T 508
�[92m☑�[0m 508
---
Q 463+382
T 845
�[92m☑�[0m 845
---
Q 0+663
T 663
�[92m☑�[0m 663
---
Q 90+14
T 104
�[92m☑�[0m 104
---
Q 905+478
T 1383
�[92m☑�[0m 1383
---
Q 627+7
T 634
�[92m☑�[0m 634
---
--------------------------------------------------
Iteration 194
Train on 45000 samples, validate on 5000 samples
Epoch 1/1
45000/45000 [==============================] - 38s - loss: 3.8677e-04 - acc: 1.0000 - val_loss: 0.0035 - val_acc: 0.9990
{'val_loss': [0.0035221813589334486], 'val_acc': [0.99895], 'loss': [0.00038677227831859556], 'acc': [1.0]}
Q 52+835
T 887
�[92m☑�[0m 887
---
Q 97+559
T 656
�[92m☑�[0m 656
---
Q 274+751
T 1025
�[92m☑�[0m 1025
---
Q 63+994
T 1057
�[92m☑�[0m 1057
---
Q 564+94
T 658
�[92m☑�[0m 658
---
Q 2+987
T 989
�[92m☑�[0m 989
---
Q 0+796
T 796
�[92m☑�[0m 796
---
Q 85+722
T 807
�[92m☑�[0m 807
---
Q 9+874
T 883
�[92m☑�[0m 883
---
Q 616+879
T 1495
�[92m☑�[0m 1495
---
--------------------------------------------------
Iteration 195
Train on 45000 samples, validate on 5000 samples
Epoch 1/1
45000/45000 [==============================] - 38s - loss: 2.9780e-04 - acc: 1.0000 - val_loss: 0.0033 - val_acc: 0.9990
{'val_loss': [0.0032784495733678342], 'val_acc': [0.999], 'loss': [0.00029779781326651574], 'acc': [1.0]}
Q 183+960
T 1143
�[92m☑�[0m 1143
---
Q 52+583
T 635
�[92m☑�[0m 635
---
Q 64+83
T 147
�[92m☑�[0m 147
---
Q 657+335
T 992
�[92m☑�[0m 992
---
Q 366+32
T 398
�[92m☑�[0m 398
---
Q 716+6
T 722
�[92m☑�[0m 722
---
Q 6+968
T 974
�[92m☑�[0m 974
---
Q 88+559
T 647
�[92m☑�[0m 647
---
Q 876+38
T 914
�[92m☑�[0m 914
---
Q 847+4
T 851
�[92m☑�[0m 851
---
--------------------------------------------------
Iteration 198
Train on 45000 samples, validate on 5000 samples
Epoch 1/1
45000/45000 [==============================] - 39s - loss: 1.9145e-04 - acc: 1.0000 - val_loss: 0.0031 - val_acc: 0.9989
{'val_loss': [0.003103120744228363], 'val_acc': [0.99890000000000001], 'loss': [0.00019145288352285408], 'acc': [1.0]}
Q 1+191
T 192
�[92m☑�[0m 192
---
Q 106+242
T 348
�[92m☑�[0m 348
---
Q 55+8
T 63
�[92m☑�[0m 63
---
Q 909+42
T 951
�[92m☑�[0m 951
---
Q 30+640
T 670
�[92m☑�[0m 670
---
Q 508+2
T 510
�[92m☑�[0m 510
---
Q 645+7
T 652
�[92m☑�[0m 652
---
Q 232+94
T 326
�[92m☑�[0m 326
---
Q 0+906
T 906
�[92m☑�[0m 906
---
Q 36+67
T 103
�[92m☑�[0m 103
---
--------------------------------------------------
Iteration 199
Train on 45000 samples, validate on 5000 samples
Epoch 1/1
45000/45000 [==============================] - 39s - loss: 1.7051e-04 - acc: 1.0000 - val_loss: 0.0032 - val_acc: 0.9990
{'val_loss': [0.0031555439017713072], 'val_acc': [0.999], 'loss': [0.00017050956679094169], 'acc': [1.0]}
Q 14+285
T 299
�[92m☑�[0m 299
---
Q 13+368
T 381
�[92m☑�[0m 381
---
Q 916+69
T 985
�[92m☑�[0m 985
---
Q 292+79
T 371
�[92m☑�[0m 371
---
Q 454+50
T 504
�[92m☑�[0m 504
---
Q 12+6
T 18
�[92m☑�[0m 18
---
Q 18+182
T 200
�[92m☑�[0m 200
---
Q 6+433
T 439
�[92m☑�[0m 439
---
Q 38+69
T 107
�[92m☑�[0m 107
---
Q 21+57
T 78
�[92m☑�[0m 78
---
# 图形化整个训练过程
ax = pd.DataFrame(
{
'val_loss': val_loss,
'val_acc': val_acc,
'loss': loss,
'acc': acc,
}
).rolling(50).mean()[50:].plot(title='Training loss', logy=True)
ax.set_xlabel("Epochs")
ax.set_ylabel("Loss&Acc")
本人的运行环境是 python3 keras2 使用了Jupyter notebook，需要完整代码的可以email：582711548@qq.com索取。
网友评论