File size: 15,451 Bytes
491eded |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 |
"""
sparse_structure_vae.py
This file implements a Variational Autoencoder (VAE) for 3D sparse structural representations.
It's part of the TRELLIS framework and contains components for encoding volumetric data
into a latent space and decoding it back to volumetric representation.
The implementation includes:
- 3D normalization layers
- 3D residual blocks for feature extraction
- 3D downsampling and upsampling blocks for resolution changes
- Encoder (SparseStructureEncoder) that maps input volumes to a latent distribution
- Decoder (SparseStructureDecoder) that reconstructs volumes from latent codes
This VAE architecture is specifically designed for capturing structural information
in a compressed latent representation that can be sampled probabilistically.
"""
from typing import *
import torch
import torch.nn as nn
import torch.nn.functional as F
from ..modules.norm import GroupNorm32, ChannelLayerNorm32
from ..modules.spatial import pixel_shuffle_3d
from ..modules.utils import zero_module, convert_module_to_f16, convert_module_to_f32
def norm_layer(norm_type: str, *args, **kwargs) -> nn.Module:
"""
Return a normalization layer based on the specified type.
Args:
norm_type: Either "group" for GroupNorm or "layer" for LayerNorm
*args, **kwargs: Arguments passed to the normalization layer
Returns:
An instance of the requested normalization layer
"""
if norm_type == "group":
return GroupNorm32(32, *args, **kwargs)
elif norm_type == "layer":
return ChannelLayerNorm32(*args, **kwargs)
else:
raise ValueError(f"Invalid norm type {norm_type}")
class ResBlock3d(nn.Module):
"""
3D Residual Block with two convolutions and a skip connection.
The block applies normalization, activation, and convolution twice,
with a skip connection from the input to the output.
"""
def __init__(
self,
channels: int,
out_channels: Optional[int] = None,
norm_type: Literal["group", "layer"] = "layer",
):
"""
Initialize a 3D ResBlock.
Args:
channels: Number of input channels
out_channels: Number of output channels (defaults to input channels)
norm_type: Type of normalization to use
"""
super().__init__()
self.channels = channels
self.out_channels = out_channels or channels
# First normalization and convolution
self.norm1 = norm_layer(norm_type, channels)
self.norm2 = norm_layer(norm_type, self.out_channels)
self.conv1 = nn.Conv3d(channels, self.out_channels, 3, padding=1)
# Second convolution is initialized with zeros for stable training
self.conv2 = zero_module(nn.Conv3d(self.out_channels, self.out_channels, 3, padding=1))
# Skip connection: identity if channels match, otherwise 1x1 conv
self.skip_connection = nn.Conv3d(channels, self.out_channels, 1) if channels != self.out_channels else nn.Identity()
def forward(self, x: torch.Tensor) -> torch.Tensor:
"""
Forward pass for the ResBlock.
Args:
x: Input tensor of shape [B, C, D, H, W]
Returns:
Output tensor after residual computation
"""
h = self.norm1(x)
h = F.silu(h)
h = self.conv1(h)
h = self.norm2(h)
h = F.silu(h)
h = self.conv2(h)
h = h + self.skip_connection(x) # Residual connection
return h
class DownsampleBlock3d(nn.Module):
"""
3D downsampling block to reduce spatial dimensions by a factor of 2.
Supports downsampling via strided convolution or average pooling.
"""
def __init__(
self,
in_channels: int,
out_channels: int,
mode: Literal["conv", "avgpool"] = "conv",
):
"""
Initialize a 3D downsampling block.
Args:
in_channels: Number of input channels
out_channels: Number of output channels
mode: Downsampling method ("conv" or "avgpool")
"""
assert mode in ["conv", "avgpool"], f"Invalid mode {mode}"
super().__init__()
self.in_channels = in_channels
self.out_channels = out_channels
if mode == "conv":
self.conv = nn.Conv3d(in_channels, out_channels, 2, stride=2)
elif mode == "avgpool":
assert in_channels == out_channels, "Pooling mode requires in_channels to be equal to out_channels"
def forward(self, x: torch.Tensor) -> torch.Tensor:
"""
Forward pass for the downsampling block.
Args:
x: Input tensor of shape [B, C, D, H, W]
Returns:
Downsampled tensor
"""
if hasattr(self, "conv"):
return self.conv(x)
else:
return F.avg_pool3d(x, 2)
class UpsampleBlock3d(nn.Module):
"""
3D upsampling block to increase spatial dimensions by a factor of 2.
Supports upsampling via transposed convolution or nearest-neighbor interpolation.
"""
def __init__(
self,
in_channels: int,
out_channels: int,
mode: Literal["conv", "nearest"] = "conv",
):
"""
Initialize a 3D upsampling block.
Args:
in_channels: Number of input channels
out_channels: Number of output channels
mode: Upsampling method ("conv" or "nearest")
"""
assert mode in ["conv", "nearest"], f"Invalid mode {mode}"
super().__init__()
self.in_channels = in_channels
self.out_channels = out_channels
if mode == "conv":
# For pixel shuffle upsampling, we need 8x channels (2³ = 8)
self.conv = nn.Conv3d(in_channels, out_channels*8, 3, padding=1)
elif mode == "nearest":
assert in_channels == out_channels, "Nearest mode requires in_channels to be equal to out_channels"
def forward(self, x: torch.Tensor) -> torch.Tensor:
"""
Forward pass for the upsampling block.
Args:
x: Input tensor of shape [B, C, D, H, W]
Returns:
Upsampled tensor
"""
if hasattr(self, "conv"):
x = self.conv(x)
return pixel_shuffle_3d(x, 2) # 3D pixel shuffle for upsampling
else:
return F.interpolate(x, scale_factor=2, mode="nearest")
class SparseStructureEncoder(nn.Module):
"""
Encoder for Sparse Structure (\mathcal{E}_S in the paper Sec. 3.3).
Takes a 3D volume as input and encodes it into a latent distribution (mean and logvar).
Can sample from this distribution to get a latent representation.
Args:
in_channels (int): Channels of the input.
latent_channels (int): Channels of the latent representation.
num_res_blocks (int): Number of residual blocks at each resolution.
channels (List[int]): Channels of the encoder blocks.
num_res_blocks_middle (int): Number of residual blocks in the middle.
norm_type (Literal["group", "layer"]): Type of normalization layer.
use_fp16 (bool): Whether to use FP16.
"""
def __init__(
self,
in_channels: int,
latent_channels: int,
num_res_blocks: int,
channels: List[int],
num_res_blocks_middle: int = 2,
norm_type: Literal["group", "layer"] = "layer",
use_fp16: bool = False,
):
"""
Initialize the encoder for sparse structure.
"""
super().__init__()
self.in_channels = in_channels
self.latent_channels = latent_channels
self.num_res_blocks = num_res_blocks
self.channels = channels
self.num_res_blocks_middle = num_res_blocks_middle
self.norm_type = norm_type
self.use_fp16 = use_fp16
self.dtype = torch.float16 if use_fp16 else torch.float32
# Initial projection from input to feature space
self.input_layer = nn.Conv3d(in_channels, channels[0], 3, padding=1)
# Encoder blocks with progressive downsampling
self.blocks = nn.ModuleList([])
for i, ch in enumerate(channels):
# Add residual blocks at the current resolution
self.blocks.extend([
ResBlock3d(ch, ch)
for _ in range(num_res_blocks)
])
# Add downsampling block if not at the final resolution
if i < len(channels) - 1:
self.blocks.append(
DownsampleBlock3d(ch, channels[i+1])
)
# Middle blocks at the lowest resolution
self.middle_block = nn.Sequential(*[
ResBlock3d(channels[-1], channels[-1])
for _ in range(num_res_blocks_middle)
])
# Output layer produces both mean and logvar for the latent distribution
self.out_layer = nn.Sequential(
norm_layer(norm_type, channels[-1]),
nn.SiLU(),
nn.Conv3d(channels[-1], latent_channels*2, 3, padding=1)
)
if use_fp16:
self.convert_to_fp16()
@property
def device(self) -> torch.device:
"""
Return the device of the model.
"""
return next(self.parameters()).device
def convert_to_fp16(self) -> None:
"""
Convert the torso of the model to float16.
"""
self.use_fp16 = True
self.dtype = torch.float16
self.blocks.apply(convert_module_to_f16)
self.middle_block.apply(convert_module_to_f16)
def convert_to_fp32(self) -> None:
"""
Convert the torso of the model to float32.
"""
self.use_fp16 = False
self.dtype = torch.float32
self.blocks.apply(convert_module_to_f32)
self.middle_block.apply(convert_module_to_f32)
def forward(self, x: torch.Tensor, sample_posterior: bool = False, return_raw: bool = False) -> torch.Tensor:
"""
Forward pass through the encoder.
Args:
x: Input tensor of shape [B, C, D, H, W]
sample_posterior: Whether to sample from the posterior distribution or just return mean
return_raw: Whether to return the raw outputs (z, mean, logvar) instead of just z
Returns:
Either the latent representation or a tuple of (z, mean, logvar) if return_raw=True
"""
h = self.input_layer(x)
h = h.type(self.dtype) # Convert to FP16 if needed
# Process through encoder blocks
for block in self.blocks:
h = block(h)
h = self.middle_block(h)
h = h.type(x.dtype) # Convert back to input dtype
h = self.out_layer(h)
# Split output into mean and log variance
mean, logvar = h.chunk(2, dim=1)
# Sample from the posterior if requested
if sample_posterior:
std = torch.exp(0.5 * logvar)
z = mean + std * torch.randn_like(std) # Reparameterization trick
else:
z = mean
if return_raw:
return z, mean, logvar
return z
class SparseStructureDecoder(nn.Module):
"""
Decoder for Sparse Structure (\mathcal{D}_S in the paper Sec. 3.3).
Takes a latent representation and decodes it back to a 3D volume.
Uses a symmetric architecture to the encoder with upsampling instead of downsampling.
Args:
out_channels (int): Channels of the output.
latent_channels (int): Channels of the latent representation.
num_res_blocks (int): Number of residual blocks at each resolution.
channels (List[int]): Channels of the decoder blocks.
num_res_blocks_middle (int): Number of residual blocks in the middle.
norm_type (Literal["group", "layer"]): Type of normalization layer.
use_fp16 (bool): Whether to use FP16.
"""
def __init__(
self,
out_channels: int,
latent_channels: int,
num_res_blocks: int,
channels: List[int],
num_res_blocks_middle: int = 2,
norm_type: Literal["group", "layer"] = "layer",
use_fp16: bool = False,
):
"""
Initialize the decoder for sparse structure.
"""
super().__init__()
self.out_channels = out_channels
self.latent_channels = latent_channels
self.num_res_blocks = num_res_blocks
self.channels = channels
self.num_res_blocks_middle = num_res_blocks_middle
self.norm_type = norm_type
self.use_fp16 = use_fp16
self.dtype = torch.float16 if use_fp16 else torch.float32
# Initial projection from latent space to feature space
self.input_layer = nn.Conv3d(latent_channels, channels[0], 3, padding=1)
# Middle blocks at the lowest resolution
self.middle_block = nn.Sequential(*[
ResBlock3d(channels[0], channels[0])
for _ in range(num_res_blocks_middle)
])
# Decoder blocks with progressive upsampling
self.blocks = nn.ModuleList([])
for i, ch in enumerate(channels):
# Add residual blocks at the current resolution
self.blocks.extend([
ResBlock3d(ch, ch)
for _ in range(num_res_blocks)
])
# Add upsampling block if not at the final resolution
if i < len(channels) - 1:
self.blocks.append(
UpsampleBlock3d(ch, channels[i+1])
)
# Final output layer to generate the desired output channels
self.out_layer = nn.Sequential(
norm_layer(norm_type, channels[-1]),
nn.SiLU(),
nn.Conv3d(channels[-1], out_channels, 3, padding=1)
)
if use_fp16:
self.convert_to_fp16()
@property
def device(self) -> torch.device:
"""
Return the device of the model.
"""
return next(self.parameters()).device
def convert_to_fp16(self) -> None:
"""
Convert the torso of the model to float16.
"""
self.use_fp16 = True
self.dtype = torch.float16
self.blocks.apply(convert_module_to_f16)
self.middle_block.apply(convert_module_to_f16)
def convert_to_fp32(self) -> None:
"""
Convert the torso of the model to float32.
"""
self.use_fp16 = False
self.dtype = torch.float32
self.blocks.apply(convert_module_to_f32)
self.middle_block.apply(convert_module_to_f32)
def forward(self, x: torch.Tensor) -> torch.Tensor:
"""
Forward pass through the decoder.
Args:
x: Latent representation tensor of shape [B, C, D, H, W]
Returns:
Reconstructed output tensor
"""
h = self.input_layer(x)
h = h.type(self.dtype) # Convert to FP16 if needed
h = self.middle_block(h)
# Process through decoder blocks
for block in self.blocks:
h = block(h)
h = h.type(x.dtype) # Convert back to input dtype
h = self.out_layer(h)
return h
|