How to use for testing or evaluation purpose?

[arunjung1991]

if not self.training:
a, b = feat.chunk(2)
return a + b
else:
return self.loss(feat, t, partial_index)
how can it be used for testing or validation dataset?

[supertx]

have u handle with this problem.I write some code to test for facial recognization but seem not work.
def compare(f1, f2):
return torch.cosine_similarity(f1, f2)

def main():
net = torch.load("/home/xxx/uniface/Models-LResNet50EIR/LResNet50EIR_28th_epoch.pth")
net.to(torch.device("cpu"))
net.eval()
net.training = False
img1 = openImg("xxx.png")
img2 = openImg("xxx.png")
re1 = net(img1.reshape(1, 3, 112, 112))
re2 = net(img2.reshape(1, 3, 112, 112))
result = compare(re1, re2)
return result