+def create_sequences_pairs(train = False):
+ nb, length = 10000, 1024
+ noise_level = 1e-2
+
+ nb_classes = 4
+ ha = torch.randint(nb_classes, (nb, ), device = device) + 1
+ # hb = torch.randint(nb_classes, (nb, ), device = device)
+ hb = ha
+
+ pos = torch.empty(nb, device = device).uniform_(0.0, 0.9)
+ a = torch.linspace(0, 1, length, device = device).view(1, -1).expand(nb, -1)
+ a = a - pos.view(nb, 1)
+ a = (a >= 0).float() * torch.exp(-a * math.log(2) / 0.1)
+ a = a * ha.float().view(-1, 1).expand_as(a) / (1 + nb_classes)
+ noise = a.new(a.size()).normal_(0, noise_level)
+ a = a + noise
+
+ pos = torch.empty(nb, device = device).uniform_(0.5)
+ b1 = torch.linspace(0, 1, length, device = device).view(1, -1).expand(nb, -1)
+ b1 = b1 - pos.view(nb, 1)
+ b1 = (b1 >= 0).float() * torch.exp(-b1 * math.log(2) / 0.1)
+ pos = pos + hb.float() / (nb_classes + 1) * 0.5
+ b2 = torch.linspace(0, 1, length, device = device).view(1, -1).expand(nb, -1)
+ b2 = b2 - pos.view(nb, 1)
+ b2 = (b2 >= 0).float() * torch.exp(-b2 * math.log(2) / 0.1)
+
+ b = b1 + b2
+ noise = b.new(b.size()).normal_(0, noise_level)
+ b = b + noise
+
+ ######################################################################
+ # for k in range(10):
+ # file = open(f'/tmp/dat{k:02d}', 'w')
+ # for i in range(a.size(1)):
+ # file.write(f'{a[k, i]:f} {b[k,i]:f}\n')
+ # file.close()
+ # exit(0)
+ ######################################################################
+
+ a = (a - a.mean()) / a.std()
+ b = (b - b.mean()) / b.std()
+
+ return a, b, ha
+
+######################################################################
+