Spaces:
Sleeping
Sleeping
| # ---------------------------------------------------------------------------------------------- | |
| # FastMETRO Official Code | |
| # Copyright (c) POSTECH Algorithmic Machine Intelligence Lab. (P-AMI Lab.) All Rights Reserved | |
| # Licensed under the MIT license. | |
| # ---------------------------------------------------------------------------------------------- | |
| # Modified from DETR (https://github.com/facebookresearch/detr) | |
| # Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved [see https://github.com/facebookresearch/detr/blob/main/LICENSE for details] | |
| # ---------------------------------------------------------------------------------------------- | |
| import math | |
| import torch | |
| from torch import nn | |
| class PositionEmbeddingSine(nn.Module): | |
| """ | |
| This is a more standard version of the position embedding, very similar to the one | |
| used by the Attention is all you need paper, generalized to work on images. | |
| """ | |
| def __init__(self, num_pos_feats=64, temperature=10000, normalize=False, scale=None): | |
| super().__init__() | |
| self.num_pos_feats = num_pos_feats | |
| self.temperature = temperature | |
| self.normalize = normalize | |
| if scale is not None and normalize is False: | |
| raise ValueError("normalize should be True if scale is passed") | |
| if scale is None: | |
| scale = 2 * math.pi | |
| self.scale = scale | |
| def forward(self, bs, h, w, device): | |
| ones = torch.ones((bs, h, w), dtype=torch.bool, device=device) | |
| y_embed = ones.cumsum(1, dtype=torch.float32) | |
| x_embed = ones.cumsum(2, dtype=torch.float32) | |
| if self.normalize: | |
| eps = 1e-6 | |
| y_embed = y_embed / (y_embed[:, -1:, :] + eps) * self.scale | |
| x_embed = x_embed / (x_embed[:, :, -1:] + eps) * self.scale | |
| dim_t = torch.arange(self.num_pos_feats, dtype=torch.float32, device=device) | |
| dim_t = self.temperature ** (2 * torch.div(dim_t, 2, rounding_mode='floor') / self.num_pos_feats) # cancel warning | |
| pos_x = x_embed[:, :, :, None] / dim_t | |
| pos_y = y_embed[:, :, :, None] / dim_t | |
| pos_x = torch.stack((pos_x[:, :, :, 0::2].sin(), pos_x[:, :, :, 1::2].cos()), dim=4).flatten(3) | |
| pos_y = torch.stack((pos_y[:, :, :, 0::2].sin(), pos_y[:, :, :, 1::2].cos()), dim=4).flatten(3) | |
| pos = torch.cat((pos_y, pos_x), dim=3).permute(0, 3, 1, 2) | |
| return pos | |
| def build_position_encoding(pos_type, hidden_dim): | |
| N_steps = hidden_dim // 2 | |
| if pos_type == 'sine': | |
| position_embedding = PositionEmbeddingSine(N_steps, normalize=True) | |
| else: | |
| raise ValueError("not supported {pos_type}") | |
| return position_embedding |