工业蒸汽量预测

时间：2025-07-23 | 作者： | 阅读：0

本项目使用人工神经网络完成蒸汽量回归预测，包括数据处理、异常值处理、相关性分析、模型构建、模型训练、模型预测等步骤。

工业蒸汽量预测

1. 引言

1.1 项目介绍

火力发电的基本原理是：燃料在燃烧时加热水生成蒸汽，蒸汽压力推动汽轮机旋转，然后汽轮机带动发电机旋转，产生电能。在这一系列的能量转化中，影响发电效率的核心是锅炉的燃烧效率，即燃料燃烧加热水产生高温高压蒸汽。锅炉的燃烧效率的影响因素很多，包括锅炉的可调参数，如燃烧给量，一二次风，引风，返料风，给水水量；以及锅炉的工况，比如锅炉床温、床压，炉膛温度、压力，过热器的温度等。
本项目使用人工神经网络完成蒸汽量回归预测，包括数据处理、异常值处理、相关性分析、模型构建、模型训练、模型预测等步骤。

1.2 数据集介绍

经脱敏后的锅炉传感器采集的数据（采集频率是分钟级别），根据锅炉的工况，预测产生的蒸汽量。

数据集各个字段以及数据类型如下所示：

数据集特征较多，需要筛选有用特征
数据集最后一列为需要预测的目标值
除最后一列外的数据列都是特征

V0V1V2V3V4V5V6V7V8V9...V36V37targetfloatfloatfloatfloatfloatfloatfloatfloatfloatfloatfloatfloatfloatfloat

2. 环境准备

2.1 安装环境

In [?]

pip install missingno -q登录后复制 ? ? ? ?

Note: you may need to restart the kernel to use updated packages.登录后复制 ? ? ? ?

2.2 导入所需模块

In [1]

import numpy as np import pandas as pd import matplotlib.pyplot as pltimport seaborn as snsimport missingno as msnofrom sklearn.model_selection import train_test_splitfrom sklearn.preprocessing import StandardScalerfrom sklearn.pipeline import Pipelinefrom sklearn.linear_model import LogisticRegressionfrom sklearn.tree import DecisionTreeClassifierfrom sklearn.ensemble import GradientBoostingClassifierfrom sklearn.ensemble import RandomForestClassifierfrom sklearn.neighbors import KNeighborsClassifier from sklearn.model_selection import GridSearchCVfrom sklearn.model_selection import cross_val_scorefrom sklearn.metrics import confusion_matrix, classification_report, accuracy_scorefrom sklearn import metricsfrom sklearn.metrics import roc_curve, auc, roc_auc_score登录后复制 ? ?

3. 数据处理

3.1 读取数据

In [2]

IndustrialSteam_train=pd.read_csv(r'data/data178496/zhengqi_train.txt',sep='t')IndustrialSteam_test=pd.read_csv(r'data/data178496/zhengqi_test.txt',sep='t')登录后复制 ? ?In [?]

IndustrialSteam_train.head()登录后复制 ? ? ? ?

V0 V1 V2 V3 V4 V5 V6 V7 V8 V9 ... 0.566 0.016 -0.143 0.407 0.452 -0.901 -1.812 -2.360 -0.436 -2.114 ... 1 0.968 0.437 0.066 0.566 0.194 -0.893 -1.566 -2.360 0.332 -2.114 ... 2 1.013 0.568 0.235 0.370 0.112 -0.797 -1.367 -2.360 0.396 -2.114 ... 3 0.733 0.368 0.283 0.165 0.599 -0.679 -1.200 -2.086 0.403 -2.114 ... 4 0.684 0.638 0.260 0.209 0.337 -0.454 -1.073 -2.086 0.314 -2.114 ... V29 V30 V31 V32 V33 V34 V35 V36 V37 target 0 0.136 0.109 -0.615 0.327 -4.627 -4.789 -5.101 -2.608 -3.508 0.175 1 -0.128 0.124 0.032 0.600 -0.843 0.160 0.364 -0.335 -0.730 0.676 2 -0.009 0.361 0.277 -0.116 -0.843 0.160 0.364 0.765 -0.589 0.633 3 0.015 0.417 0.279 0.603 -0.843 -0.065 0.364 0.333 -0.112 0.206 4 0.183 1.078 0.328 0.418 -0.843 -0.215 0.364 -0.280 -0.028 0.384 [5 rows x 39 columns]登录后复制 ? ? ? ? ? ? ? ?

3.2 数据分析及可视化

3.2.1 缺失值检测

所有属性中有2888个非空值，因此没有丢失值。
所有功能的数据类型都是float64，可以传入模型，无需进行数据类型转换。

In [?]

IndustrialSteam_train.info()登录后复制 ? ?

3.2.2 缺失值可视化

In [?]

missing_values = msno.bar(IndustrialSteam_train, figsize = (16,5),color = ”#483D8B“)登录后复制 ? ? ? ?

<Figure size 1600x500 with 3 Axes>登录后复制 ? ? ? ? ? ? ? ?

3.2.3 描述统计

In [?]

IndustrialSteam_train.describe().T登录后复制 ? ? ? ?

count mean std min 25% 50% 75% maxV0 2888.0 0.123048 0.928031 -4.335 -0.29700 0.3590 0.72600 2.121V1 2888.0 0.056068 0.941515 -5.122 -0.22625 0.2725 0.59900 1.918V2 2888.0 0.289720 0.911236 -3.420 -0.31300 0.3860 0.91825 2.828V3 2888.0 -0.067790 0.970298 -3.956 -0.65225 -0.0445 0.62400 2.457V4 2888.0 0.012921 0.888377 -4.742 -0.38500 0.1100 0.55025 2.689V5 2888.0 -0.558565 0.517957 -2.182 -0.85300 -0.4660 -0.15400 0.489V6 2888.0 0.182892 0.918054 -4.576 -0.31000 0.3880 0.83125 1.895V7 2888.0 0.116155 0.955116 -5.048 -0.29500 0.3440 0.78225 1.918V8 2888.0 0.177856 0.895444 -4.692 -0.15900 0.3620 0.72600 2.245V9 2888.0 -0.169452 0.953813 -12.891 -0.39000 0.0420 0.04200 1.335V10 2888.0 0.034319 0.968272 -2.584 -0.42050 0.1570 0.61925 4.830V11 2888.0 -0.364465 0.858504 -3.160 -0.80325 -0.1120 0.24700 1.455V12 2888.0 0.023177 0.894092 -5.165 -0.41900 0.1230 0.61600 2.657V13 2888.0 0.195738 0.922757 -3.675 -0.39800 0.2895 0.86425 2.475V14 2888.0 0.016081 1.015585 -2.455 -0.66800 -0.1610 0.82975 2.558V15 2888.0 0.096146 1.033048 -2.903 -0.66225 -0.0005 0.73000 4.314V16 2888.0 0.113505 0.983128 -5.981 -0.30000 0.3060 0.77425 2.861V17 2888.0 -0.043458 0.655857 -2.224 -0.36600 0.1650 0.43000 2.023V18 2888.0 0.055034 0.953466 -3.582 -0.36750 0.0820 0.51325 4.441V19 2888.0 -0.114884 1.108859 -3.704 -0.98750 -0.0005 0.73725 3.431V20 2888.0 -0.186226 0.788511 -3.402 -0.67550 -0.1565 0.30400 3.525V21 2888.0 -0.056556 0.781471 -2.643 -0.51700 -0.0565 0.43150 2.259V22 2888.0 0.302893 0.639186 -1.375 -0.06300 0.2165 0.87200 2.018V23 2888.0 0.155978 0.978757 -5.542 0.09725 0.3380 0.36825 1.906V24 2888.0 -0.021813 1.033403 -1.344 -1.19100 0.0950 0.93125 2.423V25 2888.0 -0.051679 0.915957 -3.808 -0.55725 -0.0760 0.35600 7.284V26 2888.0 0.072092 0.889771 -5.131 -0.45200 0.0750 0.64425 2.980V27 2888.0 0.272407 0.270374 -1.164 0.15775 0.3250 0.44200 0.925V28 2888.0 0.137712 0.929899 -2.435 -0.45500 -0.4470 0.73000 4.671V29 2888.0 0.097648 1.061200 -2.912 -0.66400 -0.0230 0.74525 4.580V30 2888.0 0.055477 0.901934 -4.507 -0.28300 0.0535 0.48800 2.689V31 2888.0 0.127791 0.873028 -5.859 -0.17025 0.2995 0.63500 2.013V32 2888.0 0.020806 0.902584 -4.053 -0.40725 0.0390 0.55700 2.395V33 2888.0 0.007801 1.006995 -4.627 -0.49900 -0.0400 0.46200 5.465V34 2888.0 0.006715 1.003291 -4.789 -0.29000 0.1600 0.27300 5.110V35 2888.0 0.197764 0.985675 -5.695 -0.20250 0.3640 0.60200 2.324V36 2888.0 0.030658 0.970812 -2.608 -0.41300 0.1370 0.64425 5.238V37 2888.0 -0.130330 1.017196 -3.630 -0.79825 -0.1855 0.49525 3.000target 2888.0 0.126353 0.983966 -3.044 -0.35025 0.3130 0.79325 2.538登录后复制 ? ? ? ? ? ? ? ?

3.2.4 样本特征分布可视化

In [?]

hist_plot = IndustrialSteam_train.hist(figsize = (20,20), color = ”#483D8B“)登录后复制 ? ? ? ?

<Figure size 2000x2000 with 42 Axes>登录后复制 ? ? ? ? ? ? ? ?

3.2.5 相关性分析

用于查看特征与特征之间的相关性
查看目标值与特征之间的相关性
筛选出与目标值有较强相关性的特征

In [?]

from pylab import mplfrom matplotlib.font_manager import FontPropertiesmyfont=FontProperties(fname=r'/usr/share/fonts/fangzheng/FZSYJW.TTF',size=12)sns.set(font=myfont.get_name())corr = IndustrialSteam_train.corr()# 调用热力图绘制相关性关系plt.figure(figsize=(25,25),dpi=150)sns.heatmap(corr, square=True, linewidths=0.1, annot=True)登录后复制 ? ? ? ?

<matplotlib.axes._subplots.AxesSubplot at 0x7fe003667410>登录后复制 ? ? ? ? ? ? ? ?

<Figure size 3750x3750 with 2 Axes>登录后复制 ? ? ? ? ? ? ? ?

3.2.6 筛选特征

以0.5为界限，同时在训练集和测试集中去除相关系数绝对值低于0.5的特征，确保被输入模型进行训练的特征与预测目标值有较强的相关性。

In [3]

df_train = IndustrialSteam_train[['V0','V1','V3','V4','V8','V12','V16','V31','target']]df_test = IndustrialSteam_test[['V0','V1','V3','V4','V8','V12','V16','V31']]登录后复制 ? ?

3.3 异常值处理

3.3.1 异常值检测

使用箱型图查看离群点（Outlier）
对于离群点考虑使用异常值处理方法

In [?]

plt.figure(figsize=(20,10))sns.boxenplot(data = df_train)plt.xticks(rotation=60)plt.show()登录后复制登录后复制 ? ? ? ?

<Figure size 2000x1000 with 1 Axes>登录后复制登录后复制 ? ? ? ? ? ? ? ?

3.3.2 异常值插补

与缺失值的插补一样，我们也可以插补异常值。
我们可以在这种方法中使用均值、中值和零值来插补。
由于我们进行了填补，而不是直接删除异常值。因此不会丢失数据。
这里的中值是合适的，因为它不受异常值的影响。

In [?]

#median imputationimport pandas as pdimport numpy as nptrain = df_trainsns.boxplot(train['V0'])plt.title(”Box Plot before median imputation“)plt.show()q1 = train['V0'].quantile(0.25)q3 = train['V0'].quantile(0.75)iqr = q3-q1Lower_tail = q1 - 1.5 * iqrUpper_tail = q3 + 1.5 * iqr# V0med = np.median(train['V0'])for i in train['V0']: if i > Upper_tail or i < Lower_tail: train['V0'] = train['V0'].replace(i, med)sns.boxplot(train['V0'])plt.title(”Box Plot after median imputation“)plt.show() # V1med = np.median(train['V1'])for i in train['V1']: if i > Upper_tail or i < Lower_tail: train['V1'] = train['V1'].replace(i, med)# V3med = np.median(train['V3'])for i in train['V3']: if i > Upper_tail or i < Lower_tail: train['V3'] = train['V3'].replace(i, med)# V4med = np.median(train['V4'])for i in train['V4']: if i > Upper_tail or i < Lower_tail: train['V4'] = train['V4'].replace(i, med)# V8med = np.median(train['V8'])for i in train['V8']: if i > Upper_tail or i < Lower_tail: train['V8'] = train['V8'].replace(i, med)# V12med = np.median(train['V12'])for i in train['V12']: if i > Upper_tail or i < Lower_tail: train['V12'] = train['V12'].replace(i, med)# V16med = np.median(train['V16'])for i in train['V16']: if i > Upper_tail or i < Lower_tail: train['V16'] = train['V16'].replace(i, med)# V31med = np.median(train['V31'])for i in train['V31']: if i > Upper_tail or i < Lower_tail: train['V31'] = train['V31'].replace(i, med)登录后复制 ? ? ? ?

<Figure size 640x480 with 1 Axes>登录后复制登录后复制登录后复制 ? ? ? ? ? ? ? ?

/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/ipykernel_launcher.py:18: SettingWithCopyWarning: A value is trying to be set on a copy of a slice from a DataFrame.Try using .loc[row_indexer,col_indexer] = value insteadSee the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy登录后复制 ? ? ? ?

<Figure size 640x480 with 1 Axes>登录后复制登录后复制登录后复制 ? ? ? ? ? ? ? ?

/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/ipykernel_launcher.py:27: SettingWithCopyWarning: A value is trying to be set on a copy of a slice from a DataFrame.Try using .loc[row_indexer,col_indexer] = value insteadSee the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/ipykernel_launcher.py:33: SettingWithCopyWarning: A value is trying to be set on a copy of a slice from a DataFrame.Try using .loc[row_indexer,col_indexer] = value insteadSee the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/ipykernel_launcher.py:39: SettingWithCopyWarning: A value is trying to be set on a copy of a slice from a DataFrame.Try using .loc[row_indexer,col_indexer] = value insteadSee the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/ipykernel_launcher.py:45: SettingWithCopyWarning: A value is trying to be set on a copy of a slice from a DataFrame.Try using .loc[row_indexer,col_indexer] = value insteadSee the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/ipykernel_launcher.py:51: SettingWithCopyWarning: A value is trying to be set on a copy of a slice from a DataFrame.Try using .loc[row_indexer,col_indexer] = value insteadSee the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/ipykernel_launcher.py:57: SettingWithCopyWarning: A value is trying to be set on a copy of a slice from a DataFrame.Try using .loc[row_indexer,col_indexer] = value insteadSee the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/ipykernel_launcher.py:63: SettingWithCopyWarning: A value is trying to be set on a copy of a slice from a DataFrame.Try using .loc[row_indexer,col_indexer] = value insteadSee the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy登录后复制 ? ? ? ?

查看异常值插补后的数据分布

In [?]

plt.figure(figsize=(20,10))sns.boxenplot(data = df_train)plt.xticks(rotation=60)plt.show()登录后复制登录后复制 ? ? ? ?

<Figure size 2000x1000 with 1 Axes>登录后复制登录后复制 ? ? ? ? ? ? ? ?

3.4 数据预处理

3.4.1 拆分特征数据与目标值数据

选择'V0','V1','V3','V4','V8','V12','V16','V31'字段作为特征
选择'target'字段作为目标值

In [4]

data_features = df_train.loc[df_train.index[:], ['V0','V1','V3','V4','V8','V12','V16','V31']]data_label = df_train['target']登录后复制 ? ?

3.4.2 划分数据集

按照0.8：0.2的比例划分训练集与测试集

In [5]

from sklearn.model_selection import train_test_split# 数据集划分x_train, x_test, y_train, y_test = train_test_split(data_features, data_label, test_size=0.2, random_state=6)print(”训练集的特征值：n“, x_train.shape)print(”测试集的标签值：n“, y_test.shape)print(”The length of original data X is:“, data_features.shape[0])print(”The length of train Data is:“, x_train.shape[0])print(”The length of test Data is:“, x_test.shape[0])登录后复制 ? ? ? ?

训练集的特征值： (2310, 8)测试集的标签值： (578,)The length of original data X is: 2888The length of train Data is: 2310The length of test Data is: 578登录后复制 ? ? ? ?

3.4.3 重置索引

对于数据集索引进行重置，确保所有数据都从第0条开始排序

In [6]

x_train=x_train.reset_index(drop=True)x_test=x_test.reset_index(drop=True)y_train=y_train.reset_index(drop=True)y_test=y_test.reset_index(drop=True)登录后复制 ? ?

3.4.4 转换数组

对重置完成的各个数据集，将数据转换成矩阵形式供后续使用。

In [7]

x_train=np.array(x_train)x_test=np.array(x_test)y_train=np.array(y_train)y_test=np.array(y_test)登录后复制 ? ?In [?]

y_train登录后复制 ? ? ? ?

array([-0.987, 1.142, -2.555, ..., -0.42 , 1.024, 1.046])登录后复制 ? ? ? ? ? ? ? ?In [8]

y_train = np.array(y_train)y_train = y_train.reshape(-1,1)y_test = np.array(y_test)y_test = y_test.reshape(-1,1)登录后复制 ? ?

3. 5 归一化（标准化）

对训练集与测试集进行归一化或者标准化处理
本处我们选择标准化

In [?]

from sklearn.preprocessing import MinMaxScalerfrom sklearn.preprocessing import StandardScaler# 1. 实例化一个转换器类transfer = StandardScaler()# 2. 标准化x_train = transfer.fit_transform(x_train)x_test = transfer.fit_transform(x_test)y_train = transfer.fit_transform(y_train)y_test = transfer.fit_transform(y_test)# df_test_x = transfer.fit_transform(df_test)登录后复制 ? ?In [?]

x_train[0]登录后复制 ? ? ? ?

array([-0.187, -0.33 , -1.523, 0.6 , -0.437, 0.389, -0.626, -0.057])登录后复制 ? ? ? ? ? ? ? ?

3.6 设置随机数种子

In [?]

import randomimport paddleseed = 666# 设置随机种子固定结果def set_seed(seed): np.random.seed(seed) random.seed(seed) paddle.seed(seed)set_seed(seed)登录后复制 ? ?

4. 模型构建

搭建全连接神经网络

8节点输入
1输出节点

In [9]

import paddleimport paddle.nn as nn# 定义动态图class Classification(paddle.nn.Layer): def __init__(self): super(Classification, self).__init__() self.fc1 = paddle.nn.Linear(8, 1) # 网络的前向计算函数 def forward(self, inputs): pred = self.fc1(inputs) return pred登录后复制 ? ?

5. 可视化损失函数

In [11]

train_nums = []train_costs = []def draw_train_process(iters,train_costs): title=”training cost“ plt.title(title, fontsize=24) plt.xlabel(”iter“, fontsize=14) plt.ylabel(”cost“, fontsize=14) plt.plot(iters, train_costs,color='red',label='training cost') plt.grid() plt.show()登录后复制 ? ?

6. 自定义损失函数

In [20]

import paddleimport paddle.nn.functional as Fclass kl_loss(paddle.nn.Layer): def __init__(self): super(kl_loss, self).__init__() def forward(self, p, q, label): ce_loss = 0.5 * (F.mse_loss(p, label=label)) + F.mse_loss(q, label=label) kl_loss = self.compute_kl_loss(p, q) # carefully choose hyper-parameters loss = ce_loss + 0.3 * kl_loss return loss def compute_kl_loss(self, p, q): p_loss = F.kl_div(F.log_softmax(p, axis=-1), F.softmax(q, axis=-1), reduction='none') q_loss = F.kl_div(F.log_softmax(q, axis=-1), F.softmax(p, axis=-1), reduction='none') # You can choose whether to use function ”sum“ and ”mean“ depending on your task p_loss = p_loss.sum() q_loss = q_loss.sum() loss = (p_loss + q_loss) / 2 return loss登录后复制 ? ?

7. 模型训练

7.1 开启训练

超参数设定如下：

BATCH_SIZE = 16
EPOCH_NUM = 20
learning_rate=0.0002 损失函数：
kl_loss

In [33]

import paddle.nn.functional as Fimport paddley_preds = []labels_list = []BATCH_SIZE = 16train_data = x_traintrain_data_y = y_traintest_data = x_testtest_data_y = y_testcompute_kl_loss = kl_loss()CET_loss = paddle.nn.CrossEntropyLoss()def train(model): print('start training ... ') # 开启模型训练模式 model.train() EPOCH_NUM = 20 train_num = 0 scheduler = paddle.optimizer.lr.CosineAnnealingDecay(learning_rate=0.0002, T_max=int(train_data.shape[0]/BATCH_SIZE*EPOCH_NUM), verbose=False) optimizer = paddle.optimizer.Adam(learning_rate=scheduler, parameters=model.parameters()) for epoch_id in range(EPOCH_NUM): # 在每轮迭代开始之前，将训练数据的顺序随机的打乱 np.random.shuffle(train_data) # 将训练数据进行拆分，每个batch包含8条数据 mini_batches = [np.append(train_data[k: k+BATCH_SIZE], train_data_y[k: k+BATCH_SIZE], axis = 1) for k in range(0, len(train_data), BATCH_SIZE)] for batch_id, data in enumerate(mini_batches): features_np = np.array(data[:, :8], np.float32) labels_np = np.array(data[:, -1:], np.float32) features = paddle.to_tensor(features_np) labels = paddle.to_tensor(labels_np) #前向计算 #y_pred = model(features) y_pred1 = model(features) y_pred2 = model(features) cost = compute_kl_loss(y_pred1, y_pred2, label=labels) # cost = CET_loss(y_pred, labels) #cost = F.mse_loss(y_pred, label=labels) train_cost = cost.numpy()[0] #反向传播 cost.backward() #最小化loss，更新参数 optimizer.step() # 清除梯度 optimizer.clear_grad() if batch_id % 500 == 0 and epoch_id % 1 == 0: print(”Pass:%d,Cost:%0.5f“%(epoch_id, train_cost)) train_num = train_num + BATCH_SIZE train_nums.append(train_num) train_costs.append(train_cost)model = Classification()train(model)登录后复制 ? ? ? ?

start training ... Pass:0,Cost:2.43391Pass:1,Cost:3.92359Pass:2,Cost:2.98257Pass:3,Cost:2.93184Pass:4,Cost:2.18770Pass:5,Cost:3.19956Pass:6,Cost:4.07202Pass:7,Cost:2.55369Pass:8,Cost:3.19636Pass:9,Cost:3.43391Pass:10,Cost:2.27505Pass:11,Cost:1.95374Pass:12,Cost:2.40070Pass:13,Cost:3.80006Pass:14,Cost:2.00660Pass:15,Cost:3.59392Pass:16,Cost:2.63512Pass:17,Cost:2.65104Pass:18,Cost:2.91626Pass:19,Cost:2.96661登录后复制 ? ? ? ?

7.2 训练过程可视化

In [41]

import matplotlibimport matplotlib.pyplot as pltimport warningswarnings.filterwarnings('ignore')%matplotlib inlinedraw_train_process(train_nums, train_costs)登录后复制 ? ? ? ?

<Figure size 640x480 with 1 Axes>登录后复制登录后复制登录后复制 ? ? ? ? ? ? ? ?

8. 模型预测

使用测试集数据进行预测
对预测后的数据进行形状转换方便后续可视化

In [36]

train_data = x_traintrain_data_y = y_traintest_data = x_testtest_data_y = y_testdef predict(model): print('start evaluating ... ') model.eval() outputs = [] mini_batches = [np.append(test_data[k: k+BATCH_SIZE], test_data_y[k: k+BATCH_SIZE], axis = 1) for k in range(0, len(test_data), BATCH_SIZE)] for data in mini_batches: features_np = np.array(data[:, :8], np.float32) features = paddle.to_tensor(features_np) pred = model(features) out = paddle.argmax(pred, axis=1) outputs.extend(out.numpy()) return outputsoutputs = predict(model)登录后复制 ? ? ? ?

start evaluating ...登录后复制 ? ? ? ?In [38]

predict_result = []for infer_feature in test_data: infer_feature = infer_feature.reshape(1, 8) infer_feature = paddle.to_tensor(np.array(infer_feature, dtype='float32')) result = model(infer_feature) predict_result.append(result)print(predict_result)登录后复制 ? ? ? ?

[Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.06025434]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.26668236]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.04699296]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.15396497]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.11334734]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.22626641]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.02553241]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.00950530]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.56165761]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.33237797]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.16205853]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.44974920]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.11122769]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.78721941]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.29552042]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.05363976]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.34395823]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.02476013]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.24630979]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.41399071]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.09647968]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.25168726]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.12463144]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.17930359]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.22265507]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.19350600]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.16149586]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.97867000]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.14237657]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.19865277]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.06837564]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.33876964]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.08810827]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-1.31689727]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.25119871]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.33098105]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.15266529]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.05445822]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.00509700]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.18745518]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.07550180]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.26571921]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.34217680]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.08926390]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.30792972]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.16032460]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.57145494]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.54297256]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.25611365]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.63418227]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.00557299]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.18981141]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.12302241]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.03510806]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.04525859]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.00092174]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.13034135]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.18077031]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.16647439]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.68720388]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.44611922]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.04261202]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.09249763]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.58799160]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.79527473]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.96323133]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.35711923]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.18494114]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.07933167]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.13546264]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.02684295]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.75451505]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.11692813]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.14113143]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.33698213]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.38676432]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.03340812]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.22545892]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.21347922]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.14782789]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.00954092]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[1.06012034]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.22003201]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.70404851]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.02809775]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.16613525]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.06436903]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.05216470]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.50551772]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.35137683]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.15199861]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.27733982]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.05591117]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.43492565]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.09615488]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.32641587]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.04105541]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.16224015]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-1.01362646]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.23819336]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.46461323]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.37720412]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.03798632]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.65613806]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.23755836]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.69738317]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.25705069]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-1.40520060]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.21922657]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.02883075]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.48285359]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.23328757]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-1.09615302]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.21412531]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.00535169]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.79399991]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.52258801]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.29668379]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.30907032]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.55264854]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.66549766]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.56879592]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.47440132]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.02180216]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.09484225]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.03586899]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.47702056]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.11216781]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.10515133]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.60479486]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.32108724]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.07980973]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.04501529]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.38887209]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.26931447]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.10054627]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.22147605]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.06280908]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.01246339]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.14042327]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.38733354]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.03984775]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.40873846]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.37586206]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.20004481]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.50934064]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.18784773]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.14203307]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.07040142]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.54865539]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.55391824]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.38516435]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.02930367]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.02538344]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.39730838]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.01709569]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.03152084]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.12164506]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.26876831]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.67705190]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.13091573]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.02252182]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.28376839]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.23220628]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.19528371]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.00430268]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.85965836]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.09281531]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.71069968]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.17499715]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.19014531]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.73743546]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.19340679]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.31273481]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.09856954]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.03048946]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.24947995]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.11751899]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.24697509]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.17461219]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.41916654]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.37340474]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.25891006]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.37807691]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[1.02975428]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.10468236]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.53602326]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.55852354]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.21298340]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.09785891]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.21151096]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.04951549]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.80943894]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.04852328]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.15818748]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.34501284]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.21866712]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.17124262]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.12341249]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.34145629]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.23096427]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.11948290]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.44626537]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.13348910]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.33801559]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.15752676]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.57183367]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.12696415]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.73151672]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.04441139]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.33734927]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.12257078]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.42197654]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.07283434]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.12016518]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.22773159]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.00632009]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.58992362]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.33788848]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.22251603]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.36759052]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.01202503]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.19001615]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.16210946]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.07992619]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.00859043]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.93312132]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.34916848]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.43634501]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.67739213]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.02566296]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.25677067]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.20933753]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.06095613]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.49417597]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.04814164]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.47177973]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.34180573]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.21158339]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.06199323]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.17558643]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.15741932]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.19417530]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.46489933]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.53986490]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.52108425]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.32266229]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.23374096]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.16298553]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.22927698]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.16080782]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.20695420]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.59630513]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[1.07300067]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.79350090]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.41532162]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.26769426]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.36136031]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.45680836]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.38632005]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.13972732]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.34614977]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.13971359]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.16951732]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.03548551]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.04101467]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.09060603]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.11193165]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.11157656]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.16849741]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.24559656]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.24626759]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.14612700]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.40763599]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[1.54426730]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.12621829]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.97957349]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.00884005]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.28282693]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.44169435]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.40584773]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.16409793]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.55420506]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.30608740]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.49449006]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.02781016]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.04644860]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.79522288]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.11904773]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.47921237]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.21942130]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.25088710]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.31529802]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.11372295]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.15712427]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.58457565]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.40284061]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.34274596]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.78810346]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.22440287]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.52228999]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.38588661]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.15205957]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.09741980]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.13445786]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.20310366]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.05031108]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.46788487]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.24416491]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.52382004]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.06532061]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.53144491]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.33535960]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.73592830]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.33153993]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.06309694]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.20155904]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.44570580]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.08384399]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.35838705]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.46934620]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.49268493]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.04641798]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.15203755]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.42403418]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.60045838]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.24354461]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.07905102]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.54188943]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.09376356]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.07086556]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.12033543]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.00973678]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.26597875]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.10573357]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.52661669]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.11851332]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.85310972]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.96144092]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.80334806]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.11293778]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.12638660]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.65866530]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.88422561]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.14159554]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.32514289]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.28985247]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.21863467]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.16769901]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.28093368]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.24580508]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[1.33940685]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.07320566]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.19715047]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.44033682]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.13477851]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.37601706]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.61441523]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.38144094]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.55627668]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.00794823]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.14110795]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.04848848]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.36352396]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.27837858]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.07947430]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.10192870]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.24401200]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.83408523]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.44856891]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.06447601]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.17212763]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.18707243]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.11827999]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.37973288]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.56238842]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.46713963]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.47919673]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.07199048]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.47425637]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.12414902]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.13834774]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.31951672]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.27122203]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.06265430]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.51172924]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.50536525]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.51465583]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.44701704]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.39858559]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.31886303]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.29493463]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.00735997]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.42210022]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.06257382]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.76618671]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.59000134]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.81734991]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.48811778]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.50484604]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.17658579]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.15134256]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.22699168]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.11169553]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.96924913]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.01308975]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.80368865]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.40497336]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.16192749]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.14131603]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[1.22107267]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.39507094]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.03073066]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.27309778]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.00708114]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.57821071]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.33658504]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.11668704]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.66753232]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.97967565]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.16306505]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.05496188]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.51948881]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.67708051]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.05163373]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.00239483]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.14013566]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.34191847]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.25270984]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.11080459]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.48523274]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.19357768]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.17594978]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.07142785]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.20822111]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.08473505]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.15897757]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.13883314]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.34538117]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.07930353]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.35454777]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.58794606]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.33299047]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.25655201]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.10445303]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.04779747]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.10455710]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.16646618]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.31524831]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.04964443]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.58060354]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.02770084]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.83897114]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.02168062]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.35580465]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.39009991]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.13909817]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.08075938]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.18590415]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.56150711]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.18525685]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.01324606]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.07201144]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.28585029]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.90698087]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.20296726]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.39485294]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.48382780]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.18443805]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.38270891]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.79779136]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.19123754]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[1.21670949]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.01447338]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.12670536]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.01824123]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.14000210]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.16719995]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.18032512]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.18801895]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.20721790]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.75916696]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.21391419]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.22957066]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.35765213]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.57816088]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.53739858]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.08023025]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.23322487]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.27490732]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.11255787]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.90063202]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.21748936]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.46920335]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-1.03563023]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.07641967]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.28591272]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.60917199]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.27543157]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.30377823]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.01602978]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.36365208]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.63090140]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.06347218]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.05933932]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.70915890]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.72446680]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.96640921]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.29156765]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.35821834]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.42687842]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.43299255]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.41873679]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.24981216]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.69104838]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.04902317]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.06218195]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.31550786]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.45310614]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.17188393]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.12393215]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.57967651]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.20886803]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.14294177]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.49980709]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.28482857]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.00266491]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.56916320]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.31185475]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.71436346]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.66652203]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.32023388]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.45304009]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.32588464]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.16148154]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.42767087]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.60954678]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.85734975]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.55337751]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.25894463]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.32260633]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.03098416]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.13747558]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.09666431]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.45293269]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.19382098]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.40163389]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.42662674]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.05361977]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.38799116]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.02736823]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.02932636]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.10339427]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.46115249]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.36047184]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.33328989]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.11792254]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.06596407]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.10048011]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.21324658]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.11020529]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.08897623]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.17561601]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.11536156]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.93352878]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.35112262]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.22222342]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.04151958]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.39091966]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.44056484]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.50789940]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-1.01449597]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.64409053]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.03456168]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.40700445]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.16290851]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[-0.12790632]]), Tensor(shape=[1, 1], dtype=float32, place=Place(gpu:0), stop_gradient=False, [[0.41777393]])]登录后复制 ? ? ? ?In [39]

predict_result = np.array(predict_result)predict_result_new = predict_result.reshape(-1,1)test_data_y_new = test_data_y登录后复制 ? ?

9. 可视化预测值与真实值

In [40]

# 绘制预测与真值结果plt.figure(figsize=(12,5), dpi=80)plt.plot(test_data_y_new[:100], label=”True value“)plt.plot(predict_result_new[:100], label=”Pred value“)plt.xlabel(”Sample“,fontproperties = 'Times New Roman', size = 18)plt.ylabel(”Value“,fontproperties = 'Times New Roman', size = 18)plt.legend(loc='best')plt.yticks(fontproperties = 'Times New Roman', size = 18)plt.xticks(fontproperties = 'Times New Roman', size = 18)plt.title(”True VS Pred“,fontproperties = 'Times New Roman', size = 18)plt.legend(loc=”best“)plt.show()登录后复制 ? ? ? ?

<Figure size 960x400 with 1 Axes>登录后复制 ? ? ? ? ? ? ? ?