MDT / models.py
shgao's picture
updaate mdt demo
723d961
raw
history blame
23.3 kB
# Copyright (c) Meta Platforms, Inc. and affiliates.
# All rights reserved.
# This source code is licensed under the license found in the
# LICENSE file in the root directory of this source tree.
# --------------------------------------------------------
# References:
# GLIDE: https://github.com/openai/glide-text2im
# MAE: https://github.com/facebookresearch/mae/blob/main/models_mae.py
# --------------------------------------------------------
import torch
import torch.nn as nn
import numpy as np
import math
from timm.models.vision_transformer import PatchEmbed, Mlp
from timm.models.layers import trunc_normal_
import math
def modulate(x, shift, scale):
return x * (1 + scale.unsqueeze(1)) + shift.unsqueeze(1)
class Attention(nn.Module):
def __init__(self, dim, num_heads=8, qkv_bias=False, qk_scale=None, attn_drop=0., proj_drop=0., num_patches=None):
super().__init__()
self.num_heads = num_heads
head_dim = dim // num_heads
# NOTE scale factor was wrong in my original version, can set manually to be compat with prev weights
self.scale = qk_scale or head_dim ** -0.5
self.qkv = nn.Linear(dim, dim * 3, bias=qkv_bias)
self.attn_drop = nn.Dropout(attn_drop)
self.proj = nn.Linear(dim, dim)
self.proj_drop = nn.Dropout(proj_drop)
self.rel_pos_bias = RelativePositionBias(
window_size=[int(num_patches**0.5), int(num_patches**0.5)], num_heads=num_heads)
def get_masked_rel_bias(self, B, ids_keep):
# get masked rel_pos_bias
rel_pos_bias = self.rel_pos_bias()
rel_pos_bias = rel_pos_bias.unsqueeze(dim=0).repeat(B, 1, 1, 1)
rel_pos_bias_masked = torch.gather(
rel_pos_bias, dim=2, index=ids_keep.unsqueeze(dim=1).unsqueeze(dim=-1).repeat(1, rel_pos_bias.shape[1], 1, rel_pos_bias.shape[-1]))
rel_pos_bias_masked = torch.gather(
rel_pos_bias_masked, dim=3, index=ids_keep.unsqueeze(dim=1).unsqueeze(dim=2).repeat(1, rel_pos_bias.shape[1], ids_keep.shape[1], 1))
return rel_pos_bias_masked
def forward(self, x, ids_keep=None):
B, N, C = x.shape
qkv = self.qkv(x).reshape(B, N, 3, self.num_heads, C //
self.num_heads).permute(2, 0, 3, 1, 4)
# make torchscript happy (cannot use tensor as tuple)
q, k, v = qkv[0], qkv[1], qkv[2]
attn = (q @ k.transpose(-2, -1)) * self.scale
if ids_keep is not None:
rp_bias = self.get_masked_rel_bias(B, ids_keep)
else:
rp_bias = self.rel_pos_bias()
attn += rp_bias
attn = attn.softmax(dim=-1)
attn = self.attn_drop(attn)
x = (attn @ v).transpose(1, 2).reshape(B, N, C)
x = self.proj(x)
x = self.proj_drop(x)
return x
class RelativePositionBias(nn.Module):
# https://github.com/microsoft/unilm/blob/master/beit/modeling_finetune.py
def __init__(self, window_size, num_heads):
super().__init__()
self.window_size = window_size
self.num_relative_distance = (
2 * window_size[0] - 1) * (2 * window_size[1] - 1) + 3
self.relative_position_bias_table = nn.Parameter(
torch.zeros(self.num_relative_distance, num_heads))
# get pair-wise relative position index for each token inside the window
coords_h = torch.arange(window_size[0])
coords_w = torch.arange(window_size[1])
coords = torch.stack(torch.meshgrid([coords_h, coords_w]))
coords_flatten = torch.flatten(coords, 1)
relative_coords = coords_flatten[:, :, None] - \
coords_flatten[:, None, :]
relative_coords = relative_coords.permute(
1, 2, 0).contiguous()
relative_coords[:, :, 0] += window_size[0] - 1
relative_coords[:, :, 1] += window_size[1] - 1
relative_coords[:, :, 0] *= 2 * window_size[1] - 1
relative_position_index = \
torch.zeros(
size=(window_size[0] * window_size[1],) * 2, dtype=relative_coords.dtype)
relative_position_index = relative_coords.sum(-1)
self.register_buffer("relative_position_index",
relative_position_index)
trunc_normal_(self.relative_position_bias_table, std=.02)
def forward(self):
relative_position_bias = \
self.relative_position_bias_table[self.relative_position_index.view(-1)].view(
self.window_size[0] * self.window_size[1],
self.window_size[0] * self.window_size[1], -1) # Wh*Ww,Wh*Ww,nH
# nH, Wh*Ww, Wh*Ww
return relative_position_bias.permute(2, 0, 1).contiguous()
#################################################################################
# Embedding Layers for Timesteps and Class Labels #
#################################################################################
class TimestepEmbedder(nn.Module):
"""
Embeds scalar timesteps into vector representations.
"""
def __init__(self, hidden_size, frequency_embedding_size=256):
super().__init__()
self.mlp = nn.Sequential(
nn.Linear(frequency_embedding_size, hidden_size, bias=True),
nn.SiLU(),
nn.Linear(hidden_size, hidden_size, bias=True),
)
self.frequency_embedding_size = frequency_embedding_size
@staticmethod
def timestep_embedding(t, dim, max_period=10000):
"""
Create sinusoidal timestep embeddings.
:param t: a 1-D Tensor of N indices, one per batch element.
These may be fractional.
:param dim: the dimension of the output.
:param max_period: controls the minimum frequency of the embeddings.
:return: an (N, D) Tensor of positional embeddings.
"""
# https://github.com/openai/glide-text2im/blob/main/glide_text2im/nn.py
half = dim // 2
freqs = torch.exp(
-math.log(max_period) * torch.arange(start=0,
end=half, dtype=torch.float32) / half
).to(device=t.device)
args = t[:, None].float() * freqs[None]
embedding = torch.cat([torch.cos(args), torch.sin(args)], dim=-1)
if dim % 2:
embedding = torch.cat(
[embedding, torch.zeros_like(embedding[:, :1])], dim=-1)
return embedding
def forward(self, t):
t_freq = self.timestep_embedding(t, self.frequency_embedding_size)
t_emb = self.mlp(t_freq)
return t_emb
class LabelEmbedder(nn.Module):
"""
Embeds class labels into vector representations. Also handles label dropout for classifier-free guidance.
"""
def __init__(self, num_classes, hidden_size, dropout_prob):
super().__init__()
use_cfg_embedding = dropout_prob > 0
self.embedding_table = nn.Embedding(
num_classes + use_cfg_embedding, hidden_size)
self.num_classes = num_classes
self.dropout_prob = dropout_prob
def token_drop(self, labels, force_drop_ids=None):
"""
Drops labels to enable classifier-free guidance.
"""
if force_drop_ids is None:
drop_ids = torch.rand(labels.shape[0]) < self.dropout_prob
else:
drop_ids = force_drop_ids == 1
labels = torch.where(drop_ids.to(labels.device),
self.num_classes, labels)
return labels
def forward(self, labels, train, force_drop_ids=None):
use_dropout = self.dropout_prob > 0
if (train and use_dropout) or (force_drop_ids is not None):
labels = self.token_drop(labels, force_drop_ids)
embeddings = self.embedding_table(labels)
return embeddings
#################################################################################
# Core MDT Model #
#################################################################################
class MDTBlock(nn.Module):
"""
A MDT block with adaptive layer norm zero (adaLN-Zero) conMDTioning.
"""
def __init__(self, hidden_size, num_heads, mlp_ratio=4.0, **block_kwargs):
super().__init__()
self.norm1 = nn.LayerNorm(
hidden_size, elementwise_affine=False, eps=1e-6)
self.attn = Attention(
hidden_size, num_heads=num_heads, qkv_bias=True, **block_kwargs)
self.norm2 = nn.LayerNorm(
hidden_size, elementwise_affine=False, eps=1e-6)
mlp_hidden_dim = int(hidden_size * mlp_ratio)
def approx_gelu(): return nn.GELU(approximate="tanh")
self.mlp = Mlp(in_features=hidden_size,
hidden_features=mlp_hidden_dim, act_layer=approx_gelu, drop=0)
self.adaLN_modulation = nn.Sequential(
nn.SiLU(),
nn.Linear(hidden_size, 6 * hidden_size, bias=True)
)
def forward(self, x, c, ids_keep=None):
shift_msa, scale_msa, gate_msa, shift_mlp, scale_mlp, gate_mlp = self.adaLN_modulation(
c).chunk(6, dim=1)
x = x + gate_msa.unsqueeze(1) * self.attn(
modulate(self.norm1(x), shift_msa, scale_msa), ids_keep=ids_keep)
x = x + \
gate_mlp.unsqueeze(
1) * self.mlp(modulate(self.norm2(x), shift_mlp, scale_mlp))
return x
class FinalLayer(nn.Module):
"""
The final layer of MDT.
"""
def __init__(self, hidden_size, patch_size, out_channels):
super().__init__()
self.norm_final = nn.LayerNorm(
hidden_size, elementwise_affine=False, eps=1e-6)
self.linear = nn.Linear(
hidden_size, patch_size * patch_size * out_channels, bias=True)
self.adaLN_modulation = nn.Sequential(
nn.SiLU(),
nn.Linear(hidden_size, 2 * hidden_size, bias=True)
)
def forward(self, x, c):
shift, scale = self.adaLN_modulation(c).chunk(2, dim=1)
x = modulate(self.norm_final(x), shift, scale)
x = self.linear(x)
return x
class MDT(nn.Module):
"""
Diffusion model with a Transformer backbone.
"""
def __init__(
self,
input_size=32,
patch_size=2,
in_channels=4,
hidden_size=1152,
depth=28,
num_heads=16,
mlp_ratio=4.0,
class_dropout_prob=0.1,
num_classes=1000,
learn_sigma=True,
mask_ratio=None,
decode_layer=None,
):
super().__init__()
self.learn_sigma = learn_sigma
self.in_channels = in_channels
self.out_channels = in_channels * 2 if learn_sigma else in_channels
self.patch_size = patch_size
self.num_heads = num_heads
self.x_embedder = PatchEmbed(
input_size, patch_size, in_channels, hidden_size, bias=True)
self.t_embedder = TimestepEmbedder(hidden_size)
self.y_embedder = LabelEmbedder(
num_classes, hidden_size, class_dropout_prob)
num_patches = self.x_embedder.num_patches
# Will use learnbale sin-cos embedding:
self.pos_embed = nn.Parameter(torch.zeros(
1, num_patches, hidden_size), requires_grad=True)
self.blocks = nn.ModuleList([
MDTBlock(hidden_size, num_heads, mlp_ratio=mlp_ratio, num_patches=num_patches) for _ in range(depth)
])
self.sideblocks = nn.ModuleList([
MDTBlock(hidden_size, num_heads, mlp_ratio=mlp_ratio, num_patches=num_patches) for _ in range(1)
])
self.final_layer = FinalLayer(
hidden_size, patch_size, self.out_channels)
self.decoder_pos_embed = nn.Parameter(torch.zeros(
1, num_patches, hidden_size), requires_grad=True)
if mask_ratio is not None:
self.mask_token = nn.Parameter(torch.zeros(1, 1, hidden_size))
self.mask_ratio = float(mask_ratio)
self.decode_layer = int(decode_layer)
else:
self.mask_token = nn.Parameter(torch.zeros(
1, 1, hidden_size), requires_grad=False)
self.mask_ratio = None
self.decode_layer = int(decode_layer)
print("mask ratio:", self.mask_ratio,
"decode_layer:", self.decode_layer)
self.initialize_weights()
def initialize_weights(self):
# Initialize transformer layers:
def _basic_init(module):
if isinstance(module, nn.Linear):
torch.nn.init.xavier_uniform_(module.weight)
if module.bias is not None:
nn.init.constant_(module.bias, 0)
self.apply(_basic_init)
# Initialize pos_embed by sin-cos embedding:
pos_embed = get_2d_sincos_pos_embed(
self.pos_embed.shape[-1], int(self.x_embedder.num_patches ** 0.5))
self.pos_embed.data.copy_(
torch.from_numpy(pos_embed).float().unsqueeze(0))
decoder_pos_embed = get_2d_sincos_pos_embed(
self.decoder_pos_embed.shape[-1], int(self.x_embedder.num_patches ** 0.5))
self.decoder_pos_embed.data.copy_(
torch.from_numpy(decoder_pos_embed).float().unsqueeze(0))
# Initialize patch_embed like nn.Linear (instead of nn.Conv2d):
w = self.x_embedder.proj.weight.data
nn.init.xavier_uniform_(w.view([w.shape[0], -1]))
nn.init.constant_(self.x_embedder.proj.bias, 0)
# Initialize label embedding table:
nn.init.normal_(self.y_embedder.embedding_table.weight, std=0.02)
# Initialize timestep embedding MLP:
nn.init.normal_(self.t_embedder.mlp[0].weight, std=0.02)
nn.init.normal_(self.t_embedder.mlp[2].weight, std=0.02)
# Zero-out adaLN modulation layers in MDT blocks:
for block in self.blocks:
nn.init.constant_(block.adaLN_modulation[-1].weight, 0)
nn.init.constant_(block.adaLN_modulation[-1].bias, 0)
for block in self.sideblocks:
nn.init.constant_(block.adaLN_modulation[-1].weight, 0)
nn.init.constant_(block.adaLN_modulation[-1].bias, 0)
# Zero-out output layers:
nn.init.constant_(self.final_layer.adaLN_modulation[-1].weight, 0)
nn.init.constant_(self.final_layer.adaLN_modulation[-1].bias, 0)
nn.init.constant_(self.final_layer.linear.weight, 0)
nn.init.constant_(self.final_layer.linear.bias, 0)
if self.mask_ratio is not None:
torch.nn.init.normal_(self.mask_token, std=.02)
def unpatchify(self, x):
"""
x: (N, T, patch_size**2 * C)
imgs: (N, H, W, C)
"""
c = self.out_channels
p = self.x_embedder.patch_size[0]
h = w = int(x.shape[1] ** 0.5)
assert h * w == x.shape[1]
x = x.reshape(shape=(x.shape[0], h, w, p, p, c))
x = torch.einsum('nhwpqc->nchpwq', x)
imgs = x.reshape(shape=(x.shape[0], c, h * p, h * p))
return imgs
def random_masking(self, x, mask_ratio):
"""
Perform per-sample random masking by per-sample shuffling.
Per-sample shuffling is done by argsort random noise.
x: [N, L, D], sequence
"""
N, L, D = x.shape # batch, length, dim
len_keep = int(L * (1 - mask_ratio))
noise = torch.rand(N, L, device=x.device) # noise in [0, 1]
# sort noise for each sample
# ascend: small is keep, large is remove
ids_shuffle = torch.argsort(noise, dim=1)
ids_restore = torch.argsort(ids_shuffle, dim=1)
# keep the first subset
ids_keep = ids_shuffle[:, :len_keep]
x_masked = torch.gather(
x, dim=1, index=ids_keep.unsqueeze(-1).repeat(1, 1, D))
# generate the binary mask: 0 is keep, 1 is remove
mask = torch.ones([N, L], device=x.device)
mask[:, :len_keep] = 0
# unshuffle to get the binary mask
mask = torch.gather(mask, dim=1, index=ids_restore)
return x_masked, mask, ids_restore, ids_keep
def forward_side_interpolater(self, x, c, mask, ids_restore):
# append mask tokens to sequence
mask_tokens = self.mask_token.repeat(
x.shape[0], ids_restore.shape[1] - x.shape[1], 1)
x_ = torch.cat([x, mask_tokens], dim=1)
x = torch.gather(
x_, dim=1, index=ids_restore.unsqueeze(-1).repeat(1, 1, x.shape[2])) # unshuffle
# add pos embed
x = x + self.decoder_pos_embed
# pass to the basic block
x_before = x
for sideblock in self.sideblocks:
x = sideblock(x, c, ids_keep=None)
# masked shortcut
mask = mask.unsqueeze(dim=-1)
x = x*mask + (1-mask)*x_before
return x
def forward(self, x, t, y, enable_mask=False):
"""
Forward pass of MDT.
x: (N, C, H, W) tensor of spatial inputs (images or latent representations of images)
t: (N,) tensor of diffusion timesteps
y: (N,) tensor of class labels
enable_mask: Use mask latent modeling
"""
x = self.x_embedder(
x) + self.pos_embed # (N, T, D), where T = H * W / patch_size ** 2
t = self.t_embedder(t) # (N, D)
y = self.y_embedder(y, self.training) # (N, D)
c = t + y # (N, D)
masked_stage = False
# masking op for training
if self.mask_ratio is not None and enable_mask:
# masking: length -> length * mask_ratio
x, mask, ids_restore, ids_keep = self.random_masking(
x, self.mask_ratio)
masked_stage = True
for i in range(len(self.blocks)):
if i == (len(self.blocks) - self.decode_layer):
if self.mask_ratio is not None and enable_mask:
x = self.forward_side_interpolater(x, c, mask, ids_restore)
masked_stage = False
else:
# add pos embed
x = x + self.decoder_pos_embed
block = self.blocks[i]
if masked_stage:
x = block(x, c, ids_keep=ids_keep)
else:
x = block(x, c, ids_keep=None)
# (N, T, patch_size ** 2 * out_channels)
x = self.final_layer(x, c)
x = self.unpatchify(x) # (N, out_channels, H, W)
return x
def forward_with_cfg(self, x, t, y, cfg_scale=None, diffusion_steps=1000, scale_pow=4.0):
"""
Forward pass of MDT, but also batches the unconditional forward pass for classifier-free guidance.
"""
# https://github.com/openai/glide-text2im/blob/main/notebooks/text2im.ipynb
if cfg_scale is not None:
half = x[: len(x) // 2]
combined = torch.cat([half, half], dim=0)
model_out = self.forward(combined, t, y)
eps, rest = model_out[:, :3], model_out[:, 3:]
cond_eps, uncond_eps = torch.split(eps, len(eps) // 2, dim=0)
scale_step = (
1-torch.cos(((1-t/diffusion_steps)**scale_pow)*math.pi))*1/2 # power-cos scaling
real_cfg_scale = (cfg_scale-1)*scale_step + 1
real_cfg_scale = real_cfg_scale[: len(x) // 2].view(-1, 1, 1, 1)
half_eps = uncond_eps + real_cfg_scale * (cond_eps - uncond_eps)
eps = torch.cat([half_eps, half_eps], dim=0)
return torch.cat([eps, rest], dim=1)
else:
model_out = self.forward(x, t, y)
eps, rest = model_out[:, :3], model_out[:, 3:]
return torch.cat([eps, rest], dim=1)
#################################################################################
# Sine/Cosine Positional Embedding Functions #
#################################################################################
# https://github.com/facebookresearch/mae/blob/main/util/pos_embed.py
def get_2d_sincos_pos_embed(embed_dim, grid_size, cls_token=False, extra_tokens=0):
"""
grid_size: int of the grid height and width
return:
pos_embed: [grid_size*grid_size, embed_dim] or [1+grid_size*grid_size, embed_dim] (w/ or w/o cls_token)
"""
grid_h = np.arange(grid_size, dtype=np.float32)
grid_w = np.arange(grid_size, dtype=np.float32)
grid = np.meshgrid(grid_w, grid_h) # here w goes first
grid = np.stack(grid, axis=0)
grid = grid.reshape([2, 1, grid_size, grid_size])
pos_embed = get_2d_sincos_pos_embed_from_grid(embed_dim, grid)
if cls_token and extra_tokens > 0:
pos_embed = np.concatenate(
[np.zeros([extra_tokens, embed_dim]), pos_embed], axis=0)
return pos_embed
def get_2d_sincos_pos_embed_from_grid(embed_dim, grid):
assert embed_dim % 2 == 0
# use half of dimensions to encode grid_h
emb_h = get_1d_sincos_pos_embed_from_grid(
embed_dim // 2, grid[0]) # (H*W, D/2)
emb_w = get_1d_sincos_pos_embed_from_grid(
embed_dim // 2, grid[1]) # (H*W, D/2)
emb = np.concatenate([emb_h, emb_w], axis=1) # (H*W, D)
return emb
def get_1d_sincos_pos_embed_from_grid(embed_dim, pos):
"""
embed_dim: output dimension for each position
pos: a list of positions to be encoded: size (M,)
out: (M, D)
"""
assert embed_dim % 2 == 0
omega = np.arange(embed_dim // 2, dtype=np.float64)
omega /= embed_dim / 2.
omega = 1. / 10000**omega # (D/2,)
pos = pos.reshape(-1) # (M,)
out = np.einsum('m,d->md', pos, omega) # (M, D/2), outer product
emb_sin = np.sin(out) # (M, D/2)
emb_cos = np.cos(out) # (M, D/2)
emb = np.concatenate([emb_sin, emb_cos], axis=1) # (M, D)
return emb
#################################################################################
# MDT Configs #
#################################################################################
def MDT_XL_2(**kwargs):
return MDT(depth=28, hidden_size=1152, patch_size=2, num_heads=16, **kwargs)
def MDT_XL_4(**kwargs):
return MDT(depth=28, hidden_size=1152, patch_size=4, num_heads=16, **kwargs)
def MDT_XL_8(**kwargs):
return MDT(depth=28, hidden_size=1152, patch_size=8, num_heads=16, **kwargs)
def MDT_L_2(**kwargs):
return MDT(depth=24, hidden_size=1024, patch_size=2, num_heads=16, **kwargs)
def MDT_L_4(**kwargs):
return MDT(depth=24, hidden_size=1024, patch_size=4, num_heads=16, **kwargs)
def MDT_L_8(**kwargs):
return MDT(depth=24, hidden_size=1024, patch_size=8, num_heads=16, **kwargs)
def MDT_B_2(**kwargs):
return MDT(depth=12, hidden_size=768, patch_size=2, num_heads=12, **kwargs)
def MDT_B_4(**kwargs):
return MDT(depth=12, hidden_size=768, patch_size=4, num_heads=12, **kwargs)
def MDT_B_8(**kwargs):
return MDT(depth=12, hidden_size=768, patch_size=8, num_heads=12, **kwargs)
def MDT_S_2(**kwargs):
return MDT(depth=12, hidden_size=384, patch_size=2, num_heads=6, **kwargs)
def MDT_S_4(**kwargs):
return MDT(depth=12, hidden_size=384, patch_size=4, num_heads=6, **kwargs)
def MDT_S_8(**kwargs):
return MDT(depth=12, hidden_size=384, patch_size=8, num_heads=6, **kwargs)