File size: 7,107 Bytes
913ec88 | 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 | import warnings
warnings.filterwarnings("ignore", category=FutureWarning)
import functools
from collections import OrderedDict
import spconv.pytorch as spconv
import torch
from spconv.pytorch.modules import SparseModule
from torch import nn
import torch.nn.functional as F
import torch
import torch.nn as nn
class CappedLayerNorm(nn.Module):
"""
LayerNorm with:
- capped scale: gamma ∈ (0, max_scale)
- capped bias: beta ∈ (-max_bias, max_bias)
- stable initialization: scale ≈ max_scale * 0.5 at start
"""
def __init__(self, dim, eps=1e-5, max_scale=1.8, max_bias=0.5):
super().__init__()
self.dim = dim
self.eps = eps
self.max_scale = max_scale
self.max_bias = max_bias
# raw learnable parameters
# g_raw = 0 → gamma = max_scale * sigmoid(0) = max_scale * 0.5
self.g_raw = nn.Parameter(torch.zeros(dim))
# b_raw = 0 → beta = tanh(0) * max_bias = 0
self.b_raw = nn.Parameter(torch.zeros(dim))
def forward(self, x):
# LayerNorm normalization
mean = x.mean(-1, keepdim=True)
var = x.var(-1, unbiased=False, keepdim=True)
x_norm = (x - mean) / torch.sqrt(var + self.eps)
# capped scale: gamma ∈ (0, max_scale)
gamma = torch.sigmoid(self.g_raw) * self.max_scale
# capped bias: beta ∈ (-max_bias, max_bias)
beta = torch.tanh(self.b_raw) * self.max_bias
return gamma * x_norm + beta
class ResidualConv(nn.Module):
def __init__(self, in_channels, out_channels):
super().__init__()
self.resi_ratio = 0.5
self.norm1 = nn.LayerNorm(in_channels)
self.linear1 = nn.Linear(in_channels, out_channels)
self.norm2 = nn.LayerNorm(out_channels)
self.linear2 = nn.Linear(out_channels, out_channels)
self.act = nn.ReLU()
def forward(self, x):
h = self.norm1(x)
h = self.linear1(h)
h = self.act(h)
h = self.norm2(h)
h = self.linear2(h)
return x + h * self.resi_ratio
class MLP(nn.Sequential):
def __init__(self, in_channels, out_channels, norm_fn=None, num_layers=2):
modules = []
for _ in range(num_layers - 1):
modules.append(nn.Linear(in_channels, in_channels))
if norm_fn:
modules.append(norm_fn(in_channels))
modules.append(nn.ReLU())
modules.append(nn.Linear(in_channels, out_channels))
return super().__init__(*modules)
def init_weights(self):
for m in self.modules():
if isinstance(m, nn.Linear):
nn.init.xavier_uniform_(m.weight)
nn.init.constant_(m.bias, 0)
nn.init.normal_(self[-1].weight, 0, 0.01)
nn.init.constant_(self[-1].bias, 0)
# current 1x1 conv in spconv2x has a bug. It will be removed after the bug is fixed
class Custom1x1Subm3d(spconv.SparseConv3d):
def forward(self, input):
features = torch.mm(input.features, self.weight.view(self.out_channels, self.in_channels).T)
if self.bias is not None:
features += self.bias
out_tensor = spconv.SparseConvTensor(features, input.indices, input.spatial_shape,
input.batch_size)
out_tensor.indice_dict = input.indice_dict
out_tensor.grid = input.grid
return out_tensor
class ResidualBlock(SparseModule):
def __init__(self, in_channels, out_channels, norm_fn, indice_key=None):
super().__init__()
if in_channels == out_channels:
self.i_branch = spconv.SparseSequential(nn.Identity())
else:
self.i_branch = spconv.SparseSequential(
Custom1x1Subm3d(in_channels, out_channels, kernel_size=1, bias=False))
self.conv_branch = spconv.SparseSequential(
norm_fn(in_channels), nn.ReLU(),
spconv.SubMConv3d(
in_channels,
out_channels,
kernel_size=3,
padding=1,
bias=False,
indice_key=indice_key), norm_fn(out_channels), nn.ReLU(),
spconv.SubMConv3d(
out_channels,
out_channels,
kernel_size=3,
padding=1,
bias=False,
indice_key=indice_key))
def forward(self, input):
identity = spconv.SparseConvTensor(input.features, input.indices, input.spatial_shape,
input.batch_size)
output = self.conv_branch(input)
out_feats = output.features + self.i_branch(identity).features
output = output.replace_feature(out_feats)
return output
class UBlock(nn.Module):
def __init__(self, nPlanes, norm_fn, block_reps, block, indice_key_id=1):
super().__init__()
self.nPlanes = nPlanes
blocks = {
'block{}'.format(i):
block(nPlanes[0], nPlanes[0], norm_fn, indice_key='subm{}'.format(indice_key_id))
for i in range(block_reps)
}
blocks = OrderedDict(blocks)
self.blocks = spconv.SparseSequential(blocks)
if len(nPlanes) > 1:
self.conv = spconv.SparseSequential(
norm_fn(nPlanes[0]), nn.ReLU(),
spconv.SparseConv3d(
nPlanes[0],
nPlanes[1],
kernel_size=2,
stride=2,
bias=False,
indice_key='spconv{}'.format(indice_key_id)))
self.u = UBlock(
nPlanes[1:], norm_fn, block_reps, block, indice_key_id=indice_key_id + 1)
self.deconv = spconv.SparseSequential(
norm_fn(nPlanes[1]), nn.ReLU(),
spconv.SparseInverseConv3d(
nPlanes[1],
nPlanes[0],
kernel_size=2,
bias=False,
indice_key='spconv{}'.format(indice_key_id)))
blocks_tail = {}
for i in range(block_reps):
blocks_tail['block{}'.format(i)] = block(
nPlanes[0] * (2 - i),
nPlanes[0],
norm_fn,
indice_key='subm{}'.format(indice_key_id))
blocks_tail = OrderedDict(blocks_tail)
self.blocks_tail = spconv.SparseSequential(blocks_tail)
def forward(self, input):
output = self.blocks(input)
identity = spconv.SparseConvTensor(output.features, output.indices, output.spatial_shape,
output.batch_size)
if len(self.nPlanes) > 1:
output_decoder = self.conv(output)
output_decoder = self.u(output_decoder)
output_decoder = self.deconv(output_decoder)
out_feats = torch.cat((identity.features, output_decoder.features), dim=1)
output = output.replace_feature(out_feats)
output = self.blocks_tail(output)
return output |