-
Notifications
You must be signed in to change notification settings - Fork 0
/
cal_flops.py
38 lines (28 loc) · 855 Bytes
/
cal_flops.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
"""
Output model complexity
"""
import time
import torch
from torchstat import stat
from torchsummary import summary
from vision.ssd.mb_tiny_fd import create_mb_tiny_fd
from vision.ssd.mb_tiny_RFB_fd import create_Mb_Tiny_RFB_fd
device = "cpu" # default cpu
width = 320
height = 240
# fd = create_mb_tiny_fd(2)
fd = create_Mb_Tiny_RFB_fd(2)
print(fd)
fd.eval()
fd.to(device)
x = torch.randn(1, 3, width, height).to(device)
summary(fd.to("cuda"), (3, width, height))
from ptflops import get_model_complexity_info
flops, params = get_model_complexity_info(fd.to(device), (3, width, height), print_per_layer_stat=True, as_strings=True)
print("FLOPS:", flops)
print("PARAMS:", params)
for i in range(5):
time_time = time.time()
features = fd(x)
print("inference time :{} s".format(time.time() - time_time))
stat(fd, (3, width, height))