-
Notifications
You must be signed in to change notification settings - Fork 0
/
LTH_fig1_mini.py
129 lines (95 loc) · 5.83 KB
/
LTH_fig1_mini.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
import torch
from torch import nn
import torchvision
import matplotlib.pyplot as plt
from d2l import torch as d2l
from dl_assignment_7_common import * # Your functions should go here if you want to use them from scripts
import os
import numpy as np
import torch.nn.utils.prune as prune
print('start')
run_name = 'mini'
cp_path = f'./checkpoints/{run_name}'
if not os.path.exists(cp_path):
os.makedirs(cp_path)
result_path = f'./results/{run_name}'
if not os.path.exists(result_path):
os.makedirs(result_path)
device = d2l.try_gpu()
dataset_used = get_dataset('mnist', dir = './data', batch_size = 60, shuffle = True, download = False)
L1_size, random_size = 5, 10
remaining_percentages = np.array([ 2, 1, 0.7, 0.5, 0.2])
prune_fractions = (100-remaining_percentages)/100
epochs_random = [ 50, 50, 50, 50, 50]
epochs_L1 = [ 50, 50, 50, 50, 50]
early_stop_iterations_lenet_L1 = np.zeros([len(prune_fractions),L1_size])
early_stop_iterations_lenet_L1 = np.insert(early_stop_iterations_lenet_L1, 0, prune_fractions, axis=1)
early_stop_iterations_lenet_random = np.zeros([len(prune_fractions),random_size])
early_stop_iterations_lenet_random = np.insert(early_stop_iterations_lenet_random, 0, prune_fractions, axis=1)
early_stop_testacc_lenet_L1 = np.zeros([len(prune_fractions),L1_size])
early_stop_testacc_lenet_L1 = np.insert(early_stop_testacc_lenet_L1, 0, prune_fractions, axis=1)
early_stop_testacc_lenet_random = np.zeros([len(prune_fractions),random_size])
early_stop_testacc_lenet_random = np.insert(early_stop_testacc_lenet_random, 0, prune_fractions, axis=1)
net, optimizer = create_network(arch = 'LeNet', input = 784, output = 10)
_, early_stop_values = train(net, optimizer, dataset_used, epochs = 50, file_specifier = f'LTH_fig1_base', val_interval = 2,cp_path=cp_path, plot = False)
for j in range(L1_size):
for i, fraction in enumerate(prune_fractions):
print(f'run {j} for fraction {fraction} L1 pruned')
trained_net = torch.load(f"{cp_path}/model_LeNet-after-LTH_fig1_base.pth")
net, optimizer = create_network(arch = 'LeNet', input = 784, output = 10)
net.load_state_dict(trained_net.state_dict())
mask = L1_prune(net, fraction)
init_net = torch.load(f"{cp_path}/model_LeNet-before-LTH_fig1_base.pth")
net, optimizer = create_network(arch = 'LeNet', input = 784, output = 10)
net.load_state_dict(init_net.state_dict())
init_net_pruned = prune_using_mask(net, mask)
optimizer = torch.optim.Adam(init_net_pruned.parameters(), lr=0.0012)
_, early_stop_values = train(init_net_pruned, optimizer, dataset_used, epochs = epochs_L1[i], file_specifier = f'LTH_L1_pruned{fraction}', val_interval = 2, cp_path=cp_path , plot = False)
early_stop_iterations_lenet_L1[i,j+1] = early_stop_values['iteration']
early_stop_testacc_lenet_L1[i,j+1] = early_stop_values['test_acc']
np.savetxt(f'{result_path}/early_stop_iterations_lenet_L1_{run_name}.txt',early_stop_iterations_lenet_L1)
np.savetxt(f'{result_path}/early_stop_testacc_lenet_L1_{run_name}.txt',early_stop_testacc_lenet_L1)
print('start random')
for j in range(random_size):
for i, fraction in enumerate(prune_fractions):
print(f'run {j} for fraction {fraction} random pruned')
net, optimizer = create_network(arch = 'LeNet', input = 784, output = 10)
_, early_stop_values = train(net, optimizer, dataset_used, epochs = 10, file_specifier = f'LTH_fig1_base', val_interval = 2, cp_path=cp_path , plot = False)
trained_net = torch.load(f"{cp_path}/model_LeNet-after-LTH_fig1_base.pth")
net, optimizer = create_network(arch = 'LeNet', input = 784, output = 10)
net.state_dict(trained_net.state_dict())
mask = random_prune(net, fraction)
init_net = torch.load(f"{cp_path}/model_LeNet-before-LTH_fig1_base.pth")
net, optimizer = create_network(arch = 'LeNet', input = 784, output = 10)
net.state_dict(init_net.state_dict())
init_net_pruned = prune_using_mask(net, mask)
optimizer = torch.optim.Adam(init_net_pruned.parameters(), lr=0.0012)
_, early_stop_values = train(init_net_pruned, optimizer, dataset_used, epochs = epochs_L1[i], file_specifier = f'LTH_random_pruned{fraction}', val_interval = 2, cp_path=cp_path , plot = False)
early_stop_iterations_lenet_random[i,j+1] = early_stop_values['iteration']
early_stop_testacc_lenet_random[i,j+1] = early_stop_values['test_acc']
np.savetxt(f'{result_path}/early_stop_iterations_lenet_random_{run_name}.txt',early_stop_iterations_lenet_random)
np.savetxt(f'{result_path}/early_stop_testacc_lenet_random_{run_name}.txt',early_stop_testacc_lenet_random)
# # Equally spaced x values for plotting
# x_spaced = np.arange(len(prune_fractions))
# plt.figure(figsize=(10, 6))
# plt.errorbar(x_spaced,early_stop_iterations_lenet_random_avg,yerr=np.std(early_stop_iterations_lenet_random[:,1:],axis=1),label='random')
# plt.errorbar(x_spaced,early_stop_iterations_lenet_L1_avg,yerr=np.std(early_stop_iterations_lenet_L1[:,1:],axis=1),label='L1')
# plt.xticks(x_spaced, prune_fractions)
# plt.xlabel('Percent of Weights Remaining')
# plt.ylabel('EarlyStop Iteration (Val)')
# plt.legend()
# plt.grid(True)
# plt.tight_layout()
# plt.savefig('./results/fig1.png')
# plt.show()
# plt.figure(figsize=(10, 6))
# plt.errorbar(x_spaced,early_stop_testacc_lenet_random_avg,yerr=np.std(early_stop_testacc_lenet_random[:,1:],axis=1),label='random')
# plt.errorbar(x_spaced,early_stop_testacc_lenet_L1_avg,yerr=np.std(early_stop_testacc_lenet_L1[:,1:],axis=1),label='L1')
# plt.xticks(x_spaced, prune_fractions)
# plt.xlabel('Percent of Weights Remaining')
# plt.ylabel('Accuracy at Early-Stop (Test)')
# plt.legend()
# plt.grid(True)
# plt.tight_layout()
# plt.savefig('./results/fig2.png')
# plt.show()