InEase

import numpy as np

per_level_scale = 1.381912879967776
log2_per_level_scale = np.log2(per_level_scale)
base_resolution = 16
log2_hashmap_size = 14

HashIn = np.load("./HashIn.npy")
HashOut = np.load("./HashOut.npy")[0:65536]
HashWeight = np.load("./HashWeight.npy")

InEase

# example of train images with masks
ds = HuBMAPDataset(transforms=augmentations)
dl = DataLoader(ds, batch_size=bs, shuffle=False, num_workers=NUM_WORKERS)
imgs, masks = next(iter(dl))

plt.figure(figsize=(16, 16))
for i, (img, mask) in enumerate(zip(imgs, masks)):
    img = ((img.permute(1, 2, 0) * std + mean) * 255.0).numpy().astype(np.uint8)
    plt.subplot(8, 8, i + 1)
    plt.imshow(img, vmin=0, vmax=255)