-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathlstm_pendulum_main.py
202 lines (154 loc) · 6.14 KB
/
lstm_pendulum_main.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
# -*- coding: utf-8 -*-
"""
Created on Wed Nov 4 11:21:58 2020
@author: Lenovo
"""
#%% Import Packages
import torch
import numpy as np
import torch.optim as optim
import nn_pend_func as pf
# import pend_phy as pph
import matplotlib.pyplot as plt
import os
import torch.nn as nn
import torch.nn.functional as F
import torchvision.models as models
from sklearn.preprocessing import MinMaxScaler
from torch.autograd import Variable
import matplotlib.animation as animation
#%% User Params
Animate = True
batch_size = 1
num_of_epochs = 1000
input_window = 50
output_len = 450
train_size = 8000
#%% Import data
file_name = 'dataset_d_pend_single_long.npy'
trainset = pf.PendDataSetSingle(file_name, input_len = input_window, outputlen = output_len, train = True)
trainloader = torch.utils.data.DataLoader(trainset, batch_size, shuffle=True)
trainSetSize = len(trainset)
dataiter = iter(trainloader)
#%% Initialize model and optimizer
input_size = 4
hidden_size = 100
num_layers = 1
num_classes = output_len
model = pf.LSTM_p(num_classes, input_size, hidden_size, num_layers, input_window, batch_size = batch_size)
criterion = nn.MSELoss()
optimizer = torch.optim.Adam(model.parameters(), lr=0.001)
model.cuda()
model.train()
#%% Train Model
for epoch in range(num_of_epochs):
dataiter = iter(trainloader)
for i in range(train_size//(batch_size * (input_window + output_len))):
data_item, data_label = dataiter.next()
data_x_torch = torch.zeros((batch_size, input_window, input_size))
data_y_torch = torch.zeros((batch_size, output_len, input_size))
for j in range(input_size):
data_xj = data_item[:,j,:]
data_yj = data_label[:,j,:]
data_x_torch[:,:,j] = (data_xj).view(batch_size, input_window).float()
data_y_torch[:,:,j] = (data_yj).view(batch_size, output_len).float()
outputs = model(data_x_torch.cuda())
# myoutput = pf.recursive_predict(model, data_x1_torch.cuda(), 20)
optimizer.zero_grad()
# obtain the loss function
loss = criterion(outputs.cuda().view(batch_size, output_len, -1), data_y_torch.cuda())
loss.backward()
optimizer.step()
if epoch % 10 == 0:
print("Epoch: %d, loss: %1.5f" % (epoch, loss.item()))
#%% predict
ind2plot = 0
file_name = 'dataset_d_pend_single_long.npy'
testset = pf.PendDataSetSingle(file_name, input_len = input_window, outputlen = output_len, train = False)
testloader = torch.utils.data.DataLoader(testset, batch_size, shuffle=False)
dataiter = iter(testloader)
# sc = MinMaxScaler()
data_item, data_label = dataiter.next()
data_x_torch = torch.zeros((batch_size, input_window, input_size))
data_y_torch = torch.zeros((batch_size, output_len, input_size))
for j in range(input_size):
data_xj = data_item[:,j,:]
data_yj = data_label[:,j,:]
# data_xj = sc.fit_transform(data_xj)
# data_yj = sc.fit_transform(data_yj)
data_x_torch[:,:,j] = (data_xj).view(batch_size, input_window).float()
data_y_torch[:,:,j] = (data_yj).view(batch_size, output_len).float()
model.eval()
model_prediction = model(data_x_torch.cuda())
predicted_x_full = np.array([],dtype = object)
true_x_full = np.array([],dtype = object)
for j in range(input_size):
data_xj = data_item[:,j,:]
data_yj = data_label[:,j,:]
dataX_plot = data_xj.data.numpy()
dataY_plot = data_yj.data.numpy()
My_train_predict = model_prediction[:,:,j]
# My_train_predict = pf.recursive_predict(model, dataX_scaled.cuda(), train_size-input_window)
my_data_predict = My_train_predict.cpu().data.numpy()
out_x_true, out_y_true = pf.debatch_data(dataX_plot, dataY_plot)
out_x_model, out_y_model = pf.debatch_data(dataX_plot, my_data_predict.reshape(batch_size,output_len))
predicted_x_full = np.append(predicted_x_full, {'{}'.format(j): out_y_model})
true_x_full = np.append(true_x_full, {'{}'.format(j): out_y_true})
plt.figure()
plt.plot(out_x_true, out_y_true, 'o')
plt.plot(out_x_model, out_y_model, '.')
plt.suptitle('Time-Series Prediction')
plt.show()
#%% Animate
dt = 0.05
x1_m = predicted_x_full[0]['0']
y1_m = predicted_x_full[1]['1']
x2_m = predicted_x_full[2]['2']
y2_m = predicted_x_full[3]['3']
x1_t = true_x_full[0]['0']
y1_t = true_x_full[1]['1']
x2_t = true_x_full[2]['2']
y2_t = true_x_full[3]['3']
if Animate:
fig = plt.figure()
ax = fig.add_subplot(121, autoscale_on=False, xlim=(-6, 6), ylim=(-10, 6))
ax.set_aspect('equal')
ax.grid()
plt.title('Model Prediction')
line, = ax.plot([], [], 'o-r', lw=2)
time_template = 'time = %.1fs'
time_text = ax.text(0.05, 0.9, '', transform=ax.transAxes)
input_temp = 'Initial few seconds \nare used As input!'
input_text_t = ax.text(0.05, 0.8, '', transform=ax.transAxes)
ax = fig.add_subplot(122, autoscale_on=False, xlim=(-6, 6), ylim=(-10, 6))
ax.set_aspect('equal')
ax.grid()
plt.title('Ground truth')
line_t, = ax.plot([], [], 'o-b', lw=2)
time_template = 'time = %.1fs'
time_text_t = ax.text(0.05, 0.9, '', transform=ax.transAxes)
def init_m():
line.set_data([], [])
time_text.set_text('')
line_t.set_data([], [])
time_text_t.set_text('')
input_text_t.set_text('')
return line, time_text, line_t, time_text_t, input_text_t
def animate_m(i):
thisx = [0, x1_m[i], x2_m[i]]
thisy = [0, y1_m[i], y2_m[i]]
line.set_data(thisx, thisy)
time_text.set_text(time_template % (i*dt))
thisx = [0, x1_t[i], x2_t[i]]
thisy = [0, y1_t[i], y2_t[i]]
line_t.set_data(thisx, thisy)
time_text_t.set_text(time_template % (i*dt))
if i < input_window:
input_text_t.set_text(input_temp)
else:
input_text_t.set_text('model output')
return line, time_text, line_t, time_text_t, input_text_t
ani = animation.FuncAnimation(fig, animate_m, range(1, len(y1_m)),
interval=dt*1000, blit=True, init_func=init_m)
plt.show()
ani.save('anim.gif', dpi=80, writer='imagemagick')