def forward(self, x):
if self.upsample:
x = F.UpSampling(x, scale=self.upsample, sample_type='nearest')
"""
if self.reflection_padding != 0:
x = self.reflection_pad(x)
"""
out = self.conv2d(x)
return out
def test_upsampling():
print("test upsampling")
tmp_dir = DIR + "upsampling/"
os.makedirs(tmp_dir + "0/", exist_ok=True)
shape = np.random.randint(low=3, high=5, size=(4))
print(shape)
a = np.random.randint(low=-127, high=127, size=shape)
np.save(tmp_dir + "0/in_0.npy", a.astype("int32"))
params = {'method': 'NEAREST_NEIGHBOR', 'layout': 'NCHW', 'scale': 1}
save_dict(params, tmp_dir + "0/attr.txt")
params = {'scale': 1, 'sample_type': 'nearest'}
b = nd.UpSampling(nd.array(a), **params)
np.save(tmp_dir + "0/out_0.npy", b.asnumpy().astype("int32"))
print(b.shape)
print(b)
os.makedirs(tmp_dir + "1/", exist_ok=True)
shape = np.random.randint(low=3, high=4, size=(4))
print(shape)
a = np.random.randint(low=-127, high=127, size=shape)
np.save(tmp_dir + "1/in_0.npy", a.astype("int32"))
params = {'method': 'NEAREST_NEIGHBOR', 'layout': 'NCHW', 'scale': 2}
save_dict(params, tmp_dir + "1/attr.txt")
params = {'scale': 2, 'sample_type': 'nearest'}
b = nd.UpSampling(nd.array(a), **params)
np.save(tmp_dir + "1/out_0.npy", b.asnumpy().astype("int32"))
print(b.shape)
os.makedirs(tmp_dir + "2/", exist_ok=True)
shape = np.random.randint(low=2, high=4, size=(4))
print(shape)
a = np.random.randint(low=-127, high=127, size=shape)
np.save(tmp_dir + "2/in_0.npy", a.astype("int32"))
params = {'method': 'NEAREST_NEIGHBOR', 'layout': 'NCHW', 'scale': 3}
save_dict(params, tmp_dir + "2/attr.txt")
params = {'scale': 3, 'sample_type': 'nearest'}
b = nd.UpSampling(nd.array(a), **params)
np.save(tmp_dir + "2/out_0.npy", b.asnumpy().astype("int32"))
print(b.shape)
print(b)
def forward(self, x):
if self.upsample:
x = F.UpSampling(x, scale=self.upsample, sample_type='nearest')
out = self.conv2d(x)
return out