Spaces:
Paused
Paused
File size: 7,905 Bytes
2aac0e2 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 |
# Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved
import math
import numpy as np
from scipy import interpolate
import torch
import torch.nn as nn
import torch.nn.functional as F
__all__ = [
"window_partition",
"window_unpartition",
"add_decomposed_rel_pos",
"get_abs_pos",
"PatchEmbed",
]
def window_partition(x, window_size):
"""
Partition into non-overlapping windows with padding if needed.
Args:
x (tensor): input tokens with [B, H, W, C].
window_size (int): window size.
Returns:
windows: windows after partition with [B * num_windows, window_size, window_size, C].
(Hp, Wp): padded height and width before partition
"""
B, H, W, C = x.shape
pad_h = (window_size - H % window_size) % window_size
pad_w = (window_size - W % window_size) % window_size
if pad_h > 0 or pad_w > 0:
x = F.pad(x, (0, 0, 0, pad_w, 0, pad_h))
Hp, Wp = H + pad_h, W + pad_w
x = x.view(B, Hp // window_size, window_size, Wp // window_size, window_size, C)
windows = x.permute(0, 1, 3, 2, 4, 5).contiguous().view(-1, window_size, window_size, C)
return windows, (Hp, Wp)
def window_unpartition(windows, window_size, pad_hw, hw):
"""
Window unpartition into original sequences and removing padding.
Args:
x (tensor): input tokens with [B * num_windows, window_size, window_size, C].
window_size (int): window size.
pad_hw (Tuple): padded height and width (Hp, Wp).
hw (Tuple): original height and width (H, W) before padding.
Returns:
x: unpartitioned sequences with [B, H, W, C].
"""
Hp, Wp = pad_hw
H, W = hw
B = windows.shape[0] // (Hp * Wp // window_size // window_size)
x = windows.view(B, Hp // window_size, Wp // window_size, window_size, window_size, -1)
x = x.permute(0, 1, 3, 2, 4, 5).contiguous().view(B, Hp, Wp, -1)
if Hp > H or Wp > W:
x = x[:, :H, :W, :].contiguous()
return x
def get_rel_pos(q_size, k_size, rel_pos, interp_type):
"""
Get relative positional embeddings according to the relative positions of
query and key sizes.
Args:
q_size (int): size of query q.
k_size (int): size of key k.
rel_pos (Tensor): relative position embeddings (L, C).
Returns:
Extracted positional embeddings according to relative positions.
"""
max_rel_dist = int(2 * max(q_size, k_size) - 1)
# Interpolate rel pos if needed.
if rel_pos.shape[0] != max_rel_dist:
if interp_type == "vitdet":
# the vitdet impl:
# https://github.com/facebookresearch/detectron2/blob/96c752ce821a3340e27edd51c28a00665dd32a30/detectron2/modeling/backbone/utils.py#L77.
rel_pos_resized = F.interpolate(
rel_pos.reshape(1, rel_pos.shape[0], -1).permute(0, 2, 1),
size=max_rel_dist,
mode="linear",
)
rel_pos_resized = rel_pos_resized.reshape(-1, max_rel_dist).permute(1, 0)
elif interp_type == "beit":
# steal from beit https://github.com/microsoft/unilm/tree/master/beit
# modified by Yuxin Fang
src_size = rel_pos.shape[0]
dst_size = max_rel_dist
q = 1.0903078
dis = []
cur = 1
for i in range(src_size // 2):
dis.append(cur)
cur += q ** (i + 1)
r_ids = [-_ for _ in reversed(dis)]
x = r_ids + [0] + dis
t = dst_size // 2.0
dx = np.arange(-t, t + 0.1, 1.0)
all_rel_pos_bias = []
for i in range(rel_pos.shape[1]):
# a hack from https://github.com/baaivision/EVA/issues/8,
# could also be used in fine-tuning but the performance haven't been tested.
z = rel_pos[:, i].view(src_size).cpu().float().detach().numpy()
f = interpolate.interp1d(x, z, kind='cubic', fill_value="extrapolate")
all_rel_pos_bias.append(
torch.Tensor(f(dx)).contiguous().view(-1, 1).to(rel_pos.device))
rel_pos_resized = torch.cat(all_rel_pos_bias, dim=-1)
else:
raise NotImplementedError()
else:
rel_pos_resized = rel_pos
# Scale the coords with short length if shapes for q and k are different.
q_coords = torch.arange(q_size)[:, None] * max(k_size / q_size, 1.0)
k_coords = torch.arange(k_size)[None, :] * max(q_size / k_size, 1.0)
relative_coords = (q_coords - k_coords) + (k_size - 1) * max(q_size / k_size, 1.0)
return rel_pos_resized[relative_coords.long()]
def add_decomposed_rel_pos(attn, q, rel_pos_h, rel_pos_w, q_size, k_size, interp_type):
"""
Calculate decomposed Relative Positional Embeddings from :paper:`mvitv2`.
https://github.com/facebookresearch/mvit/blob/19786631e330df9f3622e5402b4a419a263a2c80/mvit/models/attention.py # noqa B950
Args:
attn (Tensor): attention map.
q (Tensor): query q in the attention layer with shape (B, q_h * q_w, C).
rel_pos_h (Tensor): relative position embeddings (Lh, C) for height axis.
rel_pos_w (Tensor): relative position embeddings (Lw, C) for width axis.
q_size (Tuple): spatial sequence size of query q with (q_h, q_w).
k_size (Tuple): spatial sequence size of key k with (k_h, k_w).
Returns:
attn (Tensor): attention map with added relative positional embeddings.
"""
q_h, q_w = q_size
k_h, k_w = k_size
Rh = get_rel_pos(q_h, k_h, rel_pos_h, interp_type)
Rw = get_rel_pos(q_w, k_w, rel_pos_w, interp_type)
B, _, dim = q.shape
r_q = q.reshape(B, q_h, q_w, dim)
rel_h = torch.einsum("bhwc,hkc->bhwk", r_q, Rh)
rel_w = torch.einsum("bhwc,wkc->bhwk", r_q, Rw)
attn = (
attn.view(B, q_h, q_w, k_h, k_w) + rel_h[:, :, :, :, None] + rel_w[:, :, :, None, :]
).view(B, q_h * q_w, k_h * k_w)
return attn
def get_abs_pos(abs_pos, has_cls_token, hw):
"""
Calculate absolute positional embeddings. If needed, resize embeddings and remove cls_token
dimension for the original embeddings.
Args:
abs_pos (Tensor): absolute positional embeddings with (1, num_position, C).
has_cls_token (bool): If true, has 1 embedding in abs_pos for cls token.
hw (Tuple): size of input image tokens.
Returns:
Absolute positional embeddings after processing with shape (1, H, W, C)
"""
h, w = hw
if has_cls_token:
abs_pos = abs_pos[:, 1:]
xy_num = abs_pos.shape[1]
size = int(math.sqrt(xy_num))
assert size * size == xy_num
if size != h or size != w:
new_abs_pos = F.interpolate(
abs_pos.reshape(1, size, size, -1).permute(0, 3, 1, 2),
size=(h, w),
mode="bicubic",
align_corners=False,
)
return new_abs_pos.permute(0, 2, 3, 1)
else:
return abs_pos.reshape(1, h, w, -1)
class PatchEmbed(nn.Module):
"""
Image to Patch Embedding.
"""
def __init__(
self, kernel_size=(16, 16), stride=(16, 16), padding=(0, 0), in_chans=3, embed_dim=768
):
"""
Args:
kernel_size (Tuple): kernel size of the projection layer.
stride (Tuple): stride of the projection layer.
padding (Tuple): padding size of the projection layer.
in_chans (int): Number of input image channels.
embed_dim (int): embed_dim (int): Patch embedding dimension.
"""
super().__init__()
self.proj = nn.Conv2d(
in_chans, embed_dim, kernel_size=kernel_size, stride=stride, padding=padding
)
def forward(self, x):
x = self.proj(x)
# B C H W -> B H W C
x = x.permute(0, 2, 3, 1)
return x |