forked from cyankaet/orderml
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathgraph_train.py
216 lines (190 loc) · 6.01 KB
/
graph_train.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
203
204
205
206
207
208
209
210
211
212
213
214
215
216
import numpy as np
import matplotlib.pyplot as plt
numTrials = 1786
numFirst = 0
endParams = [[] for i in range(numTrials + numFirst)]
data = [[] for i in range(numTrials + numFirst)]
chis = [[] for i in range(numTrials + numFirst)]
nfs = [[] for i in range(numTrials + numFirst)]
jts = [[] for i in range(numTrials + numFirst)]
bks = [[] for i in range(numTrials + numFirst)]
endtns = []
endnfs = []
endjts = []
endbks = []
# these might not do anything? unclear - march 2022
tn = 160
nf = 216
jt = .2
convergs_tn = []
convergs_nf = []
convergs_jt = []
convergs_bk = []
imagestore = "/wrk/kmm11/orderout/thirdpaperrungraphs/"
# firststore = ""
logstore = "/wrk/kmm11/orderout/thirdpaperrun/"
# for i in range(1,numFirst+1):
# data[i-1] = np.loadtxt(firststore + "temps/tnLog-" + str(i) + ".npy")
# chis[i-1] = np.loadtxt(firststore + "chis/chiLog-" + str(i) + ".npy")
# nfs[i-1] = np.loadtxt(firststore + "nf/nfLog-" + str(i) + ".npy")
# jts[i-1] = np.loadtxt(firststore + "jt/jtLog-" + str(i) + ".npy")
# bks[i-1] = np.loadtxt(firststore + "bk/bkLog-" + str(i) + ".npy")
for i in range(numFirst+1,numTrials+1):
data[i-1] = np.loadtxt(logstore + "temps/tnLog-" + str(i) + ".npy")
chis[i-1] = np.loadtxt(logstore + "chis/chiLog-" + str(i) + ".npy")
nfs[i-1] = np.loadtxt(logstore + "nf/nfLog-" + str(i) + ".npy")
jts[i-1] = np.loadtxt(logstore + "jt/jtLog-" + str(i) + ".npy")
bks[i-1] = np.loadtxt(logstore + "bk/bkLog-" + str(i) + ".npy")
#plot each episode tn track, process ending values for plotting below
for i in range (0, len(data)):
try:
value = data[i][-1]
except:
value = data[i]
try:
nf = nfs[i][-1]
except:
nf = nfs[i]
try:
jt = jts[i][-1]
except:
jt = jts[i]
try:
bk = bks[i][-1]
except:
bk = bks[i]
#print(value)
endtns.append(value)
endnfs.append(nf)
endjts.append(jt)
endbks.append(bk)
#print(endtns)
plt.xlabel("measurement steps")
plt.ylabel("tn value")
#plt.show()
plt.savefig(imagestore + "tns.png")
plt.close()
# if numFirst > 0:
# convergs_tn = np.concatenate((np.loadtxt(firststore + "convergsTn.npy"),np.loadtxt(logstore + "convergsTn.npy")))
# convergs_nf = np.concatenate((np.loadtxt(firststore + "convergsNf.npy"),np.loadtxt(logstore + "convergsNf.npy")))
# convergs_jt = np.concatenate((np.loadtxt(firststore + "convergsJt.npy"),np.loadtxt(logstore + "convergsJt.npy")))
# convergs_bk = np.concatenate((np.loadtxt(firststore + "convergsBk.npy"),np.loadtxt(logstore + "convergsBk.npy")))
# else:
convergs_tn = np.loadtxt(logstore + "convergsTn.npy")
convergs_nf = np.loadtxt(logstore + "convergsNf.npy")
convergs_jt = np.loadtxt(logstore + "convergsJt.npy")
convergs_bk = np.loadtxt(logstore + "convergsBk.npy")
print("convergence for tn")
plt.plot(convergs_tn, 'ro', markersize=3)
plt.xlabel("episodes")
plt.ylabel("step at which convergence is reached")
#plt.show()
plt.savefig(imagestore + 'convergs_tn.png')
plt.close()
print("convergence for nf")
plt.plot(convergs_nf, 'ro')
plt.xlabel("episodes")
plt.ylabel("step at which convergence is reached")
#plt.show()
plt.savefig(imagestore + 'convergs_nf.png')
plt.close()
print("convergence for jt")
plt.plot(convergs_jt, 'ro')
plt.xlabel("episodes")
plt.ylabel("step at which convergence is reached")
plt.savefig(imagestore + 'convergs_jt.png')
plt.close()
print("convergence for bk")
plt.plot(convergs_bk, 'ro')
plt.xlabel("episodes")
plt.ylabel("step at which convergence is reached")
plt.savefig(imagestore + "convergs_bk.png")
#plt.show()
plt.close()
#actually plot final tn values for each episode
# plt.plot(endtns, 'bo')
# plt.xlabel("episodes")
# plt.ylabel("tn value")
# #plt.show()
# plt.savefig(imagestore + "endtns.png")
# plt.close()
print("endtns shape:", np.shape(endtns))
print("convergs_tn shape:", np.shape(convergs_tn))
plt.scatter(convergs_tn, endtns)
plt.xlabel("episodes")
plt.ylabel("tn value")
#plt.show()
plt.savefig(imagestore + "steptns.png")
plt.close()
# print("final nfs")
# plt.plot(endnfs, 'go')
# plt.xlabel("episodes")
# plt.ylabel("nf value")
# #plt.show()
# plt.savefig(imagestore + "endnfs.png")
# plt.close()
# print("final jts")
# plt.plot(endjts, 'mo')
# plt.xlabel("episodes")
# plt.ylabel("jt value")
# #plt.show()
# plt.savefig(imagestore + "endjts.png")
# plt.close()
#track chisq values through the episode
for i in range (0, len(chis)):
plt.plot(chis[i], label = str(i))
plt.xlabel("steps")
plt.ylabel("chis")
#plt.show()
plt.savefig(imagestore + "chis.png")
plt.close()
# if numFirst > 0:
# rewards = np.concatenate((np.loadtxt(firststore + 'runrewards.npy'),np.loadtxt(logstore + 'runrewards.npy')))
# else:
rewards = np.loadtxt(logstore + 'runrewards.npy')
plt.plot(rewards, 'ro', markersize=3)
plt.xlabel("episode")
plt.ylabel("reward")
plt.savefig(imagestore + "runrewards.png")
#plt.show()
plt.close()
actions = []
for i in range(1,numTrials+1):
arr = np.loadtxt(logstore + "acts/actLog-" + str(i) + ".npy")
if arr.ndim == 0:
arr = np.array([arr])
actions.append(arr)
beg_actions = actions[:100]
end_actions = actions[-100:]
beg_tns = endtns[:100]
end_tns = endtns[-100:]
temps = [1, 5, 10, 15, 20, 25, 30, 35, 40, 50]
beg_np = np.zeros(len(temps))
avg_beg = []
for actions in beg_actions:
avg_beg.append(np.average(actions))
for action in actions:
beg_np[temps.index(action)] += 1
end_np = np.zeros(len(temps))
avg_end = []
for actions in end_actions:
avg_end.append(np.average(actions))
for action in actions:
end_np[temps.index(action)] += 1
fig, axs = plt.subplots(1, 2, sharey=True, figsize=(10,5))
axs[0].bar(temps, beg_np)
axs[0].title.set_text("First 100 Eps.")
axs[1].bar(temps,end_np)
axs[1].title.set_text("Last 100 Eps.")
axs[0].set_xlabel("Selected Action")
axs[1].set_xlabel("Selected Action")
axs[0].set_ylabel("Number of Picks")
plt.savefig(imagestore + "actions.png")
plt.close()
fig, axs = plt.subplots(1, 2, sharey=True, figsize=(15,5))
axs[0].plot(beg_tns, avg_beg, 'o')
axs[0].title.set_text("First 100 Eps.")
axs[1].plot(end_tns, avg_end, 'o')
axs[1].title.set_text("Last 100 Eps.")
plt.savefig(imagestore + "avg_temp.png")
plt.close()