Can you provide the code combined with SimCLR? about un-mix HOT 4 CLOSED

      r = np.random.rand(1)
      # generate mixed sample
      cfg.beta = 1.0
      lam = np.random.beta(cfg.beta, cfg.beta)
      images_reverse = torch.flip(samples[0], (0,))
      if r < cfg.prob:
          mixed_images = lam * samples[0] + (1 - lam) * images_reverse
          mixed_images_flip = torch.flip(mixed_images, (0,))
      else:
          mixed_images = samples[0].clone()
          bbx1, bby1, bbx2, bby2 =rand_bbox(samples[0].size(), lam)
          mixed_images[:, :, bbx1:bbx2, bby1:bby2] = images_reverse[:, :, bbx1:bbx2, bby1:bby2]
          mixed_images_flip = torch.flip(mixed_images, (0,))
          # adjust lambda to exactly match pixel ratio
          lam = 1 - ((bbx2 - bbx1) * (bby2 - bby1) / (samples[0].size()[-1] * samples[0].size()[-2]))

      optimizer.zero_grad()
      loss_o = model(samples)
      loss_m1 = model([samples[1], mixed_images])
      loss_m2 = model([samples[1], mixed_images_flip])
      loss = loss_o + lam*loss_m1 + (1-lam)*loss_m2
      loss.backward()

      # function of rand_bbox for r
      def rand_bbox(size, lam):
          W = size[2]
          H = size[3]
          cut_rat = np.sqrt(1. - lam)
          cut_w = np.int(W * cut_rat)
          cut_h = np.int(H * cut_rat)
      
          # uniform
          cx = np.random.randint(W)
          cy = np.random.randint(H)
      
          bbx1 = np.clip(cx - cut_w // 2, 0, W)
          bby1 = np.clip(cy - cut_h // 2, 0, H)
          bbx2 = np.clip(cx + cut_w // 2, 0, W)
          bby2 = np.clip(cy + cut_h // 2, 0, H)
      
          return bbx1, bby1, bbx2, bby2