-
Notifications
You must be signed in to change notification settings - Fork 0
/
Figures.py
336 lines (245 loc) · 11.4 KB
/
Figures.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
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
import matplotlib.pyplot as plt
import _pickle as pickle# to serialize objects
import gzip # to decompress
import os, sys
import numpy as np
from termcolor import colored, cprint
import torch
import torch.backends.cudnn as cudnn
sys.path.append('../S3Pool/')
from models import *
from Pruning import PruneNetwork, CountZeroWeights
from Sharing import SharingWeightsNetwork, CountAllWeights
# Load checkpoint.
def Load_model(namemodel):
global net_acc
print('==> Resuming from checkpoint..')
assert os.path.isdir('./checkpoint/'), 'Error: no checkpoint directory found!'
checkpoint = torch.load('./'+namemodel+'.t7')
net = checkpoint['net']
net_acc = checkpoint['acc']
if use_cuda:
net.cuda()
net = torch.nn.DataParallel(net, device_ids=range(torch.cuda.device_count()))
cudnn.benchmark = True
return net
original_namemodel='checkpoint/Pruned_woL2/VGG19_BN_10'
namemodel = original_namemodel+'_Pruned24'
train_iter_data = pickle.load(gzip.open('./'+namemodel+'.p.gz', 'rb'))
train_loss, train_acc, test_loss, test_acc, description, accuracies, all_parameters, total_weights, zero_weights = train_iter_data
def compute_mask(layer, weights):
thresh = weights.std()
m1 = weights.abs() < thresh
mask = torch.ones(weights.size()).cuda()
mask = mask-m1.float()
print('Layer:\t',layer, '\tThreshold:\t', thresh,'\t\tPrunned weights (%):\t', colored((1-mask.mean())*100,'red') )
return mask
def CountZeroWeights(net):
parameters = net.state_dict()
layers = list(filter(lambda x:'features' in x or 'classifier' in x, parameters.keys()))
all_parameters, total_weights, zero_weights= 0, 0, 0
for i in range(len(layers)):
# print(layers[i], parameters[layers[i]].numpy().max(),' ' ,(parameters[layers[i]].numpy()==0).mean()*100)
all_parameters += np.size(parameters[layers[i]].cpu().numpy())
# print(layers[i],'\t\t', all_parameters, parameters[layers[i]].cpu().numpy().shape)
if 'weight' in layers[i]:
compute_mask(layers[i],parameters[layers[i]])
total_weights += np.size(parameters[layers[i]].cpu().numpy())
zero_weights += (parameters[layers[i]].cpu().numpy()==0).sum()
print('total parameters: {},\nTotal weights: {},\nZero weights: {},\nZero weights rate: {},\nZero weights rate in model: {}'.format(
all_parameters, total_weights, zero_weights, colored(zero_weights/total_weights*100, 'blue'), colored(zero_weights/all_parameters*100, 'green')))
return all_parameters, total_weights, zero_weights
use_cuda = torch.cuda.is_available()
original_namemodel='checkpoint/Pruned_woL2/VGG19_BN_10'
net = Load_model(original_namemodel+'_Pruned24')
CountZeroWeights(net)
CountAllWeights(net)
xxx
iterations = len(train_loss)
all_parameters = np.array(all_parameters)
zero_weights = np.array(zero_weights)
remain_weights = all_parameters-zero_weights
font = {'family' : 'normal', 'weight' : 'bold', 'size' : 15}
fig, ax1 = plt.subplots(figsize=(15,10))
plt.title('Compression/Accuracy Curve')
# ax1.set_ylim(0.12, 0.17)
ax1.plot(range(iterations+1), all_parameters/remain_weights, '-or', label='Compression Rate')
ax1.set_xlabel('Iterations', color='k')
ax1.set_ylabel('Compression', color='r')
ax2 = ax1.twinx()
ax2.set_ylabel('Accuracy', color='b')
ax2.plot(range(iterations+1), accuracies, '-ob', label='Accuracy')
handles, labels = ax1.get_legend_handles_labels()
handles1, labels1 = ax2.get_legend_handles_labels()
handles.extend(handles1)
labels.extend(labels1)
ax2.set_ylim(91.5, 93)
ax1.legend(handles, labels,loc="upper left")
plt.show()
plt.rc('font', **font)
fig, ax1 = plt.subplots(figsize=(15,10))
plt.title('Compression/Accuracy Curve')
ax1.set_ylim(0.1, 1.)
ax1.plot(range(iterations+1), remain_weights/all_parameters, '-or', label='Weights rate')
ax1.set_xlabel('Iterations', color='k')
ax1.set_ylabel('Non-zero Weights rate (%)', color='r')
ax2 = ax1.twinx()
ax2.set_ylabel('Accuracy', color='b')
ax2.plot(range(iterations+1), accuracies, '-ob', label='Accuracy')
handles, labels = ax1.get_legend_handles_labels()
handles1, labels1 = ax2.get_legend_handles_labels()
handles.extend(handles1)
labels.extend(labels1)
ax2.set_ylim(91.5, 93)
ax1.legend(handles, labels)
plt.show()
plt.rc('font', **font)
fig, ax1 = plt.subplots(figsize=(15,10))
plt.title('Compression/Accuracy Curve')
ax1.set_ylim(0.11, 0.17)
ax1.plot(range(iterations+1), remain_weights/all_parameters, '-or', label='Weights rate')
ax1.set_xlabel('Iterations', color='k')
ax1.set_ylabel('Non-zero Weights rate (%)', color='r')
ax2 = ax1.twinx()
ax2.set_ylabel('Accuracy', color='b')
ax2.plot(range(iterations+1), accuracies, '-ob', label='Accuracy')
handles, labels = ax1.get_legend_handles_labels()
handles1, labels1 = ax2.get_legend_handles_labels()
handles.extend(handles1)
labels.extend(labels1)
ax2.set_ylim(91.5, 93)
ax1.legend(handles, labels)
plt.show()
###################################################################################
###################################################################################
###################################################################################
###################################################################################
###################################################################################
def Read_training(cluster_name):
original_namemodel='checkpoint/'+cluster_name+'/VGG19_BN_10'
train_iter_data = pickle.load(gzip.open('./'+original_namemodel+'_Shared'+'.p.gz', 'rb'))
train_loss, train_acc, test_loss, test_acc, description, accuracies, all_parameters, total_weights, unique_weights = train_iter_data
iterations = len(train_loss)
all_parameters = np.array(all_parameters)
unique_weights = np.array(unique_weights)
remain_weights = unique_weights
return train_loss, train_acc, test_loss, test_acc, description, accuracies, all_parameters, total_weights, unique_weights
# test(0,'no')
original_namemodel='checkpoint/Pruned_woL2/VGG19_BN_10'
net = Load_model(original_namemodel+'_Pruned24')
a,b,c = CountZeroWeights(net)
CountAllWeights(net)
clusters_list = [0, 4,5,7,9, 11,13,15, 18,20,25,35]
original_namemodel='checkpoint/Kmeans/VGG19_BN_10'
net = Load_model(original_namemodel+'_Shared35')
CountZeroWeights(net)
CountAllWeights(net)
train_loss, train_acc, test_loss, test_acc, description, accuracies, all_parameters, total_weights, unique_weights = Read_training('Kmeans')
font = {'family' : 'normal', 'weight' : 'bold', 'size' : 15}
plt.rc('font', **font)
fig, ax1 = plt.subplots(figsize=(15,10))
ax2 = ax1.twinx()
plt.title('Compression/Accuracy Curve')
# ax1.set_ylim(0.12, 0.17)
ax1.plot(clusters_list, all_parameters/unique_weights, '--or', label='Kmeans Compression Rate')
ax1.set_xlabel('Number of Clusters per layer', color='k')
ax1.set_ylabel('Compression', color='r')
ax1.set_xticks(clusters_list)
ax1.set_xlim(0, 35.2)
ax2.set_ylabel('Accuracy', color='b')
ax2.plot(clusters_list, accuracies, '--ob', label='Accuracy Kmeans')
train_loss, train_acc, test_loss, test_acc, description, accuracies, all_parameters, total_weights, unique_weights = Read_training('GMM')
ax1.plot(clusters_list, all_parameters/unique_weights, '-or', label='GMM Compression Rate')
ax1.set_xlabel('Number of Clusters per layer', color='r')
ax1.set_ylabel('Compression', color='r')
ax1.set_xticks(clusters_list)
ax1.set_xlim(0, 35.2)
ax2.set_ylabel('Accuracy', color='b')
ax2.plot(clusters_list, accuracies, '-ob', label='Accuracy GMM')
handles, labels = ax1.get_legend_handles_labels()
handles1, labels1 = ax2.get_legend_handles_labels()
handles.extend(handles1)
labels.extend(labels1)
ax2.set_ylim(88, 93.5)
ax2.set_xlim(0, 35.2)
ax1.legend(handles, labels, loc="best")
plt.show()
train_loss, train_acc, test_loss, test_acc, description, accuracies1, all_parameters, total_weights, unique_weights1 = Read_training('Non_parametric1')
train_loss, train_acc, test_loss, test_acc, description, accuracies2, all_parameters, total_weights, unique_weights2 = Read_training('Non_parametric2')
train_loss, train_acc, test_loss, test_acc, description, accuracies3, all_parameters, total_weights, unique_weights3 = Read_training('Non_parametric3')
accuracies = [accuracies1[0], accuracies1[1], accuracies2[-1], accuracies3[-1]]
unique_weights = [unique_weights1[0], unique_weights1[1], unique_weights2[-1], unique_weights3[-1]]
plt.rc('font', **font)
fig, ax1 = plt.subplots(figsize=(15,10))
plt.title('Compression/Accuracy Curve')
ax1.set_ylim(0, 1.)
ax1.plot(range(len(accuracies)), unique_weights/all_parameters, '-or', label='Weights rate')
ax1.set_xlabel('Covariance Initialization', color='k')
ax1.set_ylabel('Unique non-zero Weights rate (%)', color='r')
ax1.set_xticks([0,1,2,3])
ax2 = ax1.twinx()
ax2.set_ylabel('Accuracy', color='b')
ax2.plot(range(len(accuracies)), accuracies, '-ob', label='Accuracy')
handles, labels = ax1.get_legend_handles_labels()
handles1, labels1 = ax2.get_legend_handles_labels()
handles.extend(handles1)
labels.extend(labels1)
ax2.set_ylim(91., 93.5)
ax1.legend(handles, labels)
plt.show()
plt.rc('font', **font)
fig, ax1 = plt.subplots(figsize=(15,10))
plt.title('Compression/Accuracy Curve')
ax1.set_ylim(0., 0.000015)
ax1.plot(range(len(accuracies)), unique_weights/all_parameters, '-or', label='Weights rate')
ax1.set_xlabel('Covariance Initialization', color='k')
ax1.set_ylabel('Non-zero Weights rate (%)', color='r')
ax1.set_xticks([0,1,2,3])
ax2 = ax1.twinx()
ax2.set_ylabel('Accuracy', color='b')
ax2.plot(range(len(accuracies)), accuracies, '-ob', label='Accuracy')
handles, labels = ax1.get_legend_handles_labels()
handles1, labels1 = ax2.get_legend_handles_labels()
handles.extend(handles1)
labels.extend(labels1)
ax2.set_ylim(91., 93.5)
ax1.legend(handles, labels)
plt.show()
fig, ax1 = plt.subplots(figsize=(15,10))
plt.title('Error Model Curve')
# ax1.set_ylim(0.12, 0.17)
ax1.plot(range(len(test_acc)), 100-np.array(train_acc), '--b', label='Training')
ax1.plot(range(len(test_acc)), 100-np.array(test_acc), '-r', label='Test')
ax1.set_xlabel('Number of epochs', color='k')
ax1.set_ylabel('Error', color='r')
handles, labels = ax1.get_legend_handles_labels()
ax1.legend(handles, labels)
plt.show()
train_loss, train_acc, test_loss, test_acc, description, accuracies, all_parameters, total_weights, unique_weights = Read_training('Kmeans')
fig, ax1 = plt.subplots(figsize=(15,10))
plt.title('Error Model Curve')
# ax1.set_ylim(0.12, 0.17)
ax1.plot(range(test_acc.shape[1]), 100-np.array(train_acc).T, '--b', label='Training')
ax1.plot(range(test_acc.shape[1]), 100-np.array(test_acc).T, '-r', label='Test')
ax1.set_xlabel('Number of epochs', color='k')
ax1.set_ylabel('Error', color='r')
handles, labels = ax1.get_legend_handles_labels()
ax1.legend(handles[::int(len(test_acc))], labels[::int(len(test_acc))])
plt.show()
# # Maybe S3Pool
# fig, ax1 = plt.subplots(figsize=(15,10))
# plt.title('Error Model Curve')
# # ax1.set_ylim(0.12, 0.17)
# ax1.plot(range(len(test_acc)), 100-np.array(test_acc), '--b', label='Training')
# ax1.plot(range(len(test_acc)), 100-np.array(train_acc), '-r', label='Test')
# ax1.set_xlabel('Number of epochs', color='k')
# ax1.set_ylabel('Accuracy', color='r')
# ax2 = ax1.twinx()
# ax2.set_ylabel('Weights L1 values', color='k')
# ax2.plot(range(len(test_acc)), np.array(total_weights[:-1]), '-k', label='Weights sum')
# handles, labels = ax1.get_legend_handles_labels()
# handles1, labels1 = ax2.get_legend_handles_labels()
# handles.extend(handles1)
# labels.extend(labels1)
# ax1.legend(handles, labels)
# plt.show()