mirror of https://github.com/alibaba/EasyCV.git
70 lines
2.2 KiB
Python
70 lines
2.2 KiB
Python
# Copyright (c) Alibaba, Inc. and its affiliates.
|
|
import copy
|
|
import random
|
|
import unittest
|
|
|
|
import numpy as np
|
|
import torch
|
|
|
|
from easycv.models import modelzoo
|
|
from easycv.models.backbones import MNASNet
|
|
|
|
|
|
class MnasnetTest(unittest.TestCase):
|
|
|
|
def setUp(self):
|
|
print(('Testing %s.%s' % (type(self).__name__, self._testMethodName)))
|
|
|
|
def test_mnasnet_withfc(self):
|
|
for alpha in [0.5, 0.75, 1.0, 1.3]:
|
|
with torch.no_grad():
|
|
# input data
|
|
batch_size = random.randint(10, 30)
|
|
a = torch.rand(batch_size, 3, 224, 224).to('cuda')
|
|
|
|
num_classes = random.randint(10, 1000)
|
|
net = MNASNet(alpha=alpha, num_classes=num_classes).to('cuda')
|
|
net.init_weights()
|
|
net.eval()
|
|
|
|
self.assertTrue(len(list(net(a)[-1].shape)) == 2)
|
|
|
|
self.assertTrue(net(a)[-1].size(1) == num_classes)
|
|
|
|
self.assertTrue(net(a)[-1].size(0) == batch_size)
|
|
|
|
def test_mnasnet_withoutfc(self):
|
|
for alpha in [0.5, 0.75, 1.0, 1.3]:
|
|
with torch.no_grad():
|
|
# input data
|
|
batch_size = random.randint(10, 30)
|
|
a = torch.rand(batch_size, 3, 224, 224).to('cuda')
|
|
|
|
net = MNASNet(alpha=alpha, num_classes=0).to('cuda')
|
|
net.init_weights()
|
|
net.eval()
|
|
|
|
self.assertTrue(net(a)[-1].size(1) == 1280)
|
|
|
|
self.assertTrue(net(a)[-1].size(0) == batch_size)
|
|
|
|
def test_mnasnet_load_modelzoo(self):
|
|
for alpha in [0.5, 1.0]:
|
|
with torch.no_grad():
|
|
net = MNASNet(alpha=1.0, num_classes=1000).to('cuda')
|
|
original_weight = net.layers[0].weight
|
|
original_weight = copy.deepcopy(
|
|
original_weight.cpu().data.numpy())
|
|
|
|
net.init_weights(net.pretrained)
|
|
load_weight = net.layers[0].weight.cpu().data.numpy()
|
|
|
|
self.assertFalse(np.allclose(original_weight, load_weight))
|
|
|
|
self.assertTrue(net.pretrained == modelzoo.mnasnet['MNASNet' +
|
|
'1.0'])
|
|
|
|
|
|
if __name__ == '__main__':
|
|
unittest.main()
|