syncdiffusion/utils.py

import numpy as np
import torch

def seed_everything(seed):
    torch.manual_seed(seed)
    torch.cuda.manual_seed(seed)
    torch.cuda.manual_seed_all(seed)
    torch.backends.cudnn.deterministic = True
    torch.backends.cudnn.benchmark = False

def get_views(panorama_height, panorama_width, window_size=64, stride=8):
    panorama_height /= 8
    panorama_width /= 8
    num_blocks_height = (panorama_height - window_size) // stride + 1
    num_blocks_width = (panorama_width - window_size) // stride + 1
    total_num_blocks = int(num_blocks_height * num_blocks_width)
    views = []
    for i in range(total_num_blocks):
        h_start = int((i // num_blocks_width) * stride)
        h_end = h_start + window_size
        w_start = int((i % num_blocks_width) * stride)
        w_end = w_start + window_size
        views.append((h_start, h_end, w_start, w_end))
    return views

def exponential_decay_list(init_weight, decay_rate, num_steps):
    weights = [init_weight * (decay_rate ** i) for i in range(num_steps)]
    return torch.tensor(weights)