118 lines
4.2 KiB
Python
118 lines
4.2 KiB
Python
|
|
|
|
import torch
|
|
from typing import Tuple, Callable
|
|
import math
|
|
|
|
def do_nothing(x: torch.Tensor, mode:str=None):
|
|
return x
|
|
|
|
|
|
def bipartite_soft_matching_random2d(metric: torch.Tensor,
|
|
w: int, h: int, sx: int, sy: int, r: int,
|
|
no_rand: bool = False) -> Tuple[Callable, Callable]:
|
|
"""
|
|
Partitions the tokens into src and dst and merges r tokens from src to dst.
|
|
Dst tokens are partitioned by choosing one randomy in each (sx, sy) region.
|
|
|
|
Args:
|
|
- metric [B, N, C]: metric to use for similarity
|
|
- w: image width in tokens
|
|
- h: image height in tokens
|
|
- sx: stride in the x dimension for dst, must divide w
|
|
- sy: stride in the y dimension for dst, must divide h
|
|
- r: number of tokens to remove (by merging)
|
|
- no_rand: if true, disable randomness (use top left corner only)
|
|
"""
|
|
B, N, _ = metric.shape
|
|
|
|
if r <= 0:
|
|
return do_nothing, do_nothing
|
|
|
|
with torch.no_grad():
|
|
|
|
hsy, wsx = h // sy, w // sx
|
|
|
|
# For each sy by sx kernel, randomly assign one token to be dst and the rest src
|
|
idx_buffer = torch.zeros(1, hsy, wsx, sy*sx, 1, device=metric.device)
|
|
|
|
if no_rand:
|
|
rand_idx = torch.zeros(1, hsy, wsx, 1, 1, device=metric.device, dtype=torch.int64)
|
|
else:
|
|
rand_idx = torch.randint(sy*sx, size=(1, hsy, wsx, 1, 1), device=metric.device)
|
|
|
|
idx_buffer.scatter_(dim=3, index=rand_idx, src=-torch.ones_like(rand_idx, dtype=idx_buffer.dtype))
|
|
idx_buffer = idx_buffer.view(1, hsy, wsx, sy, sx, 1).transpose(2, 3).reshape(1, N, 1)
|
|
rand_idx = idx_buffer.argsort(dim=1)
|
|
|
|
num_dst = int((1 / (sx*sy)) * N)
|
|
a_idx = rand_idx[:, num_dst:, :] # src
|
|
b_idx = rand_idx[:, :num_dst, :] # dst
|
|
|
|
def split(x):
|
|
C = x.shape[-1]
|
|
src = x.gather(dim=1, index=a_idx.expand(B, N - num_dst, C))
|
|
dst = x.gather(dim=1, index=b_idx.expand(B, num_dst, C))
|
|
return src, dst
|
|
|
|
metric = metric / metric.norm(dim=-1, keepdim=True)
|
|
a, b = split(metric)
|
|
scores = a @ b.transpose(-1, -2)
|
|
|
|
# Can't reduce more than the # tokens in src
|
|
r = min(a.shape[1], r)
|
|
|
|
node_max, node_idx = scores.max(dim=-1)
|
|
edge_idx = node_max.argsort(dim=-1, descending=True)[..., None]
|
|
|
|
unm_idx = edge_idx[..., r:, :] # Unmerged Tokens
|
|
src_idx = edge_idx[..., :r, :] # Merged Tokens
|
|
dst_idx = node_idx[..., None].gather(dim=-2, index=src_idx)
|
|
|
|
def merge(x: torch.Tensor, mode="mean") -> torch.Tensor:
|
|
src, dst = split(x)
|
|
n, t1, c = src.shape
|
|
|
|
unm = src.gather(dim=-2, index=unm_idx.expand(n, t1 - r, c))
|
|
src = src.gather(dim=-2, index=src_idx.expand(n, r, c))
|
|
dst = dst.scatter_reduce(-2, dst_idx.expand(n, r, c), src, reduce=mode)
|
|
|
|
return torch.cat([unm, dst], dim=1)
|
|
|
|
def unmerge(x: torch.Tensor) -> torch.Tensor:
|
|
unm_len = unm_idx.shape[1]
|
|
unm, dst = x[..., :unm_len, :], x[..., unm_len:, :]
|
|
_, _, c = unm.shape
|
|
|
|
src = dst.gather(dim=-2, index=dst_idx.expand(B, r, c))
|
|
|
|
# Combine back to the original shape
|
|
out = torch.zeros(B, N, c, device=x.device, dtype=x.dtype)
|
|
out.scatter_(dim=-2, index=b_idx.expand(B, num_dst, c), src=dst)
|
|
out.scatter_(dim=-2, index=a_idx.expand(B, a_idx.shape[1], 1).gather(dim=1, index=unm_idx).expand(B, unm_len, c), src=unm)
|
|
out.scatter_(dim=-2, index=a_idx.expand(B, a_idx.shape[1], 1).gather(dim=1, index=src_idx).expand(B, r, c), src=src)
|
|
|
|
return out
|
|
|
|
return merge, unmerge
|
|
|
|
|
|
def get_functions(x, ratio, original_shape):
|
|
b, c, original_h, original_w = original_shape
|
|
original_tokens = original_h * original_w
|
|
downsample = int(math.sqrt(original_tokens // x.shape[1]))
|
|
stride_x = 2
|
|
stride_y = 2
|
|
max_downsample = 1
|
|
|
|
if downsample <= max_downsample:
|
|
w = original_w // downsample
|
|
h = original_h // downsample
|
|
r = int(x.shape[1] * ratio)
|
|
no_rand = False
|
|
m, u = bipartite_soft_matching_random2d(x, w, h, stride_x, stride_y, r, no_rand)
|
|
return m, u
|
|
|
|
nothing = lambda y: y
|
|
return nothing, nothing
|