Spaces:
Runtime error
Runtime error
| import math | |
| import warnings | |
| import torch | |
| from dist import cached_grid | |
| from utils import log as log_utils | |
| LOGGER = log_utils.getLogger(__name__) | |
| def mask_selector(masks_softmaxed, top=2, size_norm=False): | |
| """Select <top> centre most masks and sumthem """ | |
| b, k, *other_dims, h, w = masks_softmaxed.shape | |
| masks_softmaxed = masks_softmaxed.view(b, k, 1, h, w) | |
| g = cached_grid(h, w, device=masks_softmaxed.device, dtype=masks_softmaxed.dtype) | |
| x = g[0, 0] / (w - 1) - .5 | |
| y = g[0, 1] / (h - 1) - .5 | |
| v = (x ** 2 + y ** 2) * 2 | |
| assert len(v.shape) == 2 | |
| v = v.view(*[1] * (len(masks_softmaxed) - 2), h, w) | |
| scores = (masks_softmaxed * (1 - v)).sum([-1, -2]).view(b, k) | |
| scores = scores / (masks_softmaxed.flatten(-3).sum(-1) + 1e-6) | |
| LOGGER.debug_once(f"Selector -- masks in {masks_softmaxed.shape}; scores {scores.shape}") | |
| best_idxs = scores.topk(top, dim=-1).indices[..., None, None, None].expand(-1, -1, -1, h, w) | |
| wrst_idxs = (-scores).topk(k - top, dim=-1).indices[..., None, None, None].expand(-1, -1, -1, h, w) | |
| LOGGER.debug_once(f"Selector -- inds {best_idxs.shape} {wrst_idxs.shape}") | |
| masks_out = torch.empty(b, 2, 1, h, w, device=masks_softmaxed.device, dtype=masks_softmaxed.dtype) | |
| centre_most_masks = torch.gather(masks_softmaxed, 1, best_idxs).sum(1, keepdim=True) | |
| other_masks = torch.gather(masks_softmaxed, 1, wrst_idxs).sum(1, keepdim=True) | |
| LOGGER.debug_once(f"Selector -- best {centre_most_masks.shape} others {other_masks.shape}") | |
| masks_out[:, 1:] = centre_most_masks | |
| masks_out[:, :1] = other_masks | |
| return masks_out.view(b, 2, *other_dims, h, w) | |
| def _no_grad_trunc_normal_(tensor, mean, std, a, b): | |
| # Cut & paste from PyTorch official master until it's in a few official releases - RW | |
| # Method based on https://people.sc.fsu.edu/~jburkardt/presentations/truncated_normal.pdf | |
| def norm_cdf(x): | |
| # Computes standard normal cumulative distribution function | |
| return (1. + math.erf(x / math.sqrt(2.))) / 2. | |
| if (mean < a - 2 * std) or (mean > b + 2 * std): | |
| warnings.warn("mean is more than 2 std from [a, b] in nn.init.trunc_normal_. " | |
| "The distribution of values may be incorrect.", | |
| stacklevel=2) | |
| with torch.no_grad(): | |
| # Values are generated by using a truncated uniform distribution and | |
| # then using the inverse CDF for the normal distribution. | |
| # Get upper and lower cdf values | |
| l = norm_cdf((a - mean) / std) | |
| u = norm_cdf((b - mean) / std) | |
| # Uniformly fill tensor with values from [l, u], then translate to | |
| # [2l-1, 2u-1]. | |
| tensor.uniform_(2 * l - 1, 2 * u - 1) | |
| # Use inverse cdf transform for normal distribution to get truncated | |
| # standard normal | |
| tensor.erfinv_() | |
| # Transform to proper mean, std | |
| tensor.mul_(std * math.sqrt(2.)) | |
| tensor.add_(mean) | |
| # Clamp to ensure it's in the proper range | |
| tensor.clamp_(min=a, max=b) | |
| return tensor | |
| def trunc_normal_(tensor, mean=0., std=1., a=-2., b=2.): | |
| # type: (Tensor, float, float, float, float) -> Tensor | |
| return _no_grad_trunc_normal_(tensor, mean, std, a, b) | |