import logging import math from abc import abstractmethod from typing import Iterable, List, Optional, Tuple, Union import torch import torch as th import torch.nn as nn import torch.nn.functional as F from einops import rearrange from functools import partial # from torch.utils.checkpoint import checkpoint checkpoint = partial(torch.utils.checkpoint.checkpoint, use_reentrant=False) from ...modules.attention import SpatialTransformer from ...modules.diffusionmodules.util import ( avg_pool_nd, conv_nd, linear, normalization, timestep_embedding, zero_module, ) from ...modules.video_attention import SpatialVideoTransformer from ...util import exists logpy = logging.getLogger(__name__) class AttentionPool2d(nn.Module): """ Adapted from CLIP: https://github.com/openai/CLIP/blob/main/clip/model.py """ def __init__( self, spacial_dim: int, embed_dim: int, num_heads_channels: int, output_dim: Optional[int] = None, ): super().__init__() self.positional_embedding = nn.Parameter( th.randn(embed_dim, spacial_dim**2 + 1) / embed_dim**0.5 ) self.qkv_proj = conv_nd(1, embed_dim, 3 * embed_dim, 1) self.c_proj = conv_nd(1, embed_dim, output_dim or embed_dim, 1) self.num_heads = embed_dim // num_heads_channels self.attention = QKVAttention(self.num_heads) def forward(self, x: th.Tensor) -> th.Tensor: b, c, _ = x.shape x = x.reshape(b, c, -1) x = th.cat([x.mean(dim=-1, keepdim=True), x], dim=-1) x = x + self.positional_embedding[None, :, :].to(x.dtype) x = self.qkv_proj(x) x = self.attention(x) x = self.c_proj(x) return x[:, :, 0] class TimestepBlock(nn.Module): """ Any module where forward() takes timestep embeddings as a second argument. """ @abstractmethod def forward(self, x: th.Tensor, emb: th.Tensor): """ Apply the module to `x` given `emb` timestep embeddings. """ class TimestepEmbedSequential(nn.Sequential, TimestepBlock): """ A sequential module that passes timestep embeddings to the children that support it as an extra input. """ def forward( self, x: th.Tensor, emb: th.Tensor, context: Optional[th.Tensor] = None, image_only_indicator: Optional[th.Tensor] = None, time_context: Optional[int] = None, num_video_frames: Optional[int] = None, ): from ...modules.diffusionmodules.video_model import VideoResBlock for layer in self: module = layer if isinstance(module, TimestepBlock) and not isinstance( module, VideoResBlock ): x = layer(x, emb) elif isinstance(module, VideoResBlock): x = layer(x, emb, num_video_frames, image_only_indicator) elif isinstance(module, SpatialVideoTransformer): x = layer( x, context, time_context, num_video_frames, image_only_indicator, ) elif isinstance(module, SpatialTransformer): x = layer(x, context) else: x = layer(x) return x class Upsample(nn.Module): """ An upsampling layer with an optional convolution. :param channels: channels in the inputs and outputs. :param use_conv: a bool determining if a convolution is applied. :param dims: determines if the signal is 1D, 2D, or 3D. If 3D, then upsampling occurs in the inner-two dimensions. """ def __init__( self, channels: int, use_conv: bool, dims: int = 2, out_channels: Optional[int] = None, padding: int = 1, third_up: bool = False, kernel_size: int = 3, scale_factor: int = 2, ): super().__init__() self.channels = channels self.out_channels = out_channels or channels self.use_conv = use_conv self.dims = dims self.third_up = third_up self.scale_factor = scale_factor if use_conv: self.conv = conv_nd( dims, self.channels, self.out_channels, kernel_size, padding=padding ) def forward(self, x: th.Tensor) -> th.Tensor: assert x.shape[1] == self.channels if self.dims == 3: t_factor = 1 if not self.third_up else self.scale_factor x = F.interpolate( x, ( t_factor * x.shape[2], x.shape[3] * self.scale_factor, x.shape[4] * self.scale_factor, ), mode="nearest", ) else: x = F.interpolate(x, scale_factor=self.scale_factor, mode="nearest") if self.use_conv: x = self.conv(x) return x class Downsample(nn.Module): """ A downsampling layer with an optional convolution. :param channels: channels in the inputs and outputs. :param use_conv: a bool determining if a convolution is applied. :param dims: determines if the signal is 1D, 2D, or 3D. If 3D, then downsampling occurs in the inner-two dimensions. """ def __init__( self, channels: int, use_conv: bool, dims: int = 2, out_channels: Optional[int] = None, padding: int = 1, third_down: bool = False, ): super().__init__() self.channels = channels self.out_channels = out_channels or channels self.use_conv = use_conv self.dims = dims stride = 2 if dims != 3 else ((1, 2, 2) if not third_down else (2, 2, 2)) if use_conv: logpy.info(f"Building a Downsample layer with {dims} dims.") logpy.info( f" --> settings are: \n in-chn: {self.channels}, out-chn: {self.out_channels}, " f"kernel-size: 3, stride: {stride}, padding: {padding}" ) if dims == 3: logpy.info(f" --> Downsampling third axis (time): {third_down}") self.op = conv_nd( dims, self.channels, self.out_channels, 3, stride=stride, padding=padding, ) else: assert self.channels == self.out_channels self.op = avg_pool_nd(dims, kernel_size=stride, stride=stride) def forward(self, x: th.Tensor) -> th.Tensor: assert x.shape[1] == self.channels return self.op(x) class ResBlock(TimestepBlock): """ A residual block that can optionally change the number of channels. :param channels: the number of input channels. :param emb_channels: the number of timestep embedding channels. :param dropout: the rate of dropout. :param out_channels: if specified, the number of out channels. :param use_conv: if True and out_channels is specified, use a spatial convolution instead of a smaller 1x1 convolution to change the channels in the skip connection. :param dims: determines if the signal is 1D, 2D, or 3D. :param use_checkpoint: if True, use gradient checkpointing on this module. :param up: if True, use this block for upsampling. :param down: if True, use this block for downsampling. """ def __init__( self, channels: int, emb_channels: int, dropout: float, out_channels: Optional[int] = None, use_conv: bool = False, use_scale_shift_norm: bool = False, dims: int = 2, use_checkpoint: bool = False, up: bool = False, down: bool = False, kernel_size: int = 3, exchange_temb_dims: bool = False, skip_t_emb: bool = False, ): super().__init__() self.channels = channels self.emb_channels = emb_channels self.dropout = dropout self.out_channels = out_channels or channels self.use_conv = use_conv self.use_checkpoint = use_checkpoint self.use_scale_shift_norm = use_scale_shift_norm self.exchange_temb_dims = exchange_temb_dims if isinstance(kernel_size, Iterable): padding = [k // 2 for k in kernel_size] else: padding = kernel_size // 2 self.in_layers = nn.Sequential( normalization(channels), nn.SiLU(), conv_nd(dims, channels, self.out_channels, kernel_size, padding=padding), ) self.updown = up or down if up: self.h_upd = Upsample(channels, False, dims) self.x_upd = Upsample(channels, False, dims) elif down: self.h_upd = Downsample(channels, False, dims) self.x_upd = Downsample(channels, False, dims) else: self.h_upd = self.x_upd = nn.Identity() self.skip_t_emb = skip_t_emb self.emb_out_channels = ( 2 * self.out_channels if use_scale_shift_norm else self.out_channels ) if self.skip_t_emb: logpy.info(f"Skipping timestep embedding in {self.__class__.__name__}") assert not self.use_scale_shift_norm self.emb_layers = None self.exchange_temb_dims = False else: self.emb_layers = nn.Sequential( nn.SiLU(), linear( emb_channels, self.emb_out_channels, ), ) self.out_layers = nn.Sequential( normalization(self.out_channels), nn.SiLU(), nn.Dropout(p=dropout), zero_module( conv_nd( dims, self.out_channels, self.out_channels, kernel_size, padding=padding, ) ), ) if self.out_channels == channels: self.skip_connection = nn.Identity() elif use_conv: self.skip_connection = conv_nd( dims, channels, self.out_channels, kernel_size, padding=padding ) else: self.skip_connection = conv_nd(dims, channels, self.out_channels, 1) def forward(self, x: th.Tensor, emb: th.Tensor) -> th.Tensor: """ Apply the block to a Tensor, conditioned on a timestep embedding. :param x: an [N x C x ...] Tensor of features. :param emb: an [N x emb_channels] Tensor of timestep embeddings. :return: an [N x C x ...] Tensor of outputs. """ if self.use_checkpoint: return checkpoint(self._forward, x, emb) else: return self._forward(x, emb) def _forward(self, x: th.Tensor, emb: th.Tensor) -> th.Tensor: if self.updown: in_rest, in_conv = self.in_layers[:-1], self.in_layers[-1] h = in_rest(x) h = self.h_upd(h) x = self.x_upd(x) h = in_conv(h) else: h = self.in_layers(x) if self.skip_t_emb: emb_out = th.zeros_like(h) else: emb_out = self.emb_layers(emb).type(h.dtype) while len(emb_out.shape) < len(h.shape): emb_out = emb_out[..., None] if self.use_scale_shift_norm: out_norm, out_rest = self.out_layers[0], self.out_layers[1:] scale, shift = th.chunk(emb_out, 2, dim=1) h = out_norm(h) * (1 + scale) + shift h = out_rest(h) else: if self.exchange_temb_dims: emb_out = rearrange(emb_out, "b t c ... -> b c t ...") h = h + emb_out h = self.out_layers(h) return self.skip_connection(x) + h class AttentionBlock(nn.Module): """ An attention block that allows spatial positions to attend to each other. Originally ported from here, but adapted to the N-d case. https://github.com/hojonathanho/diffusion/blob/1e0dceb3b3495bbe19116a5e1b3596cd0706c543/diffusion_tf/models/unet.py#L66. """ def __init__( self, channels: int, num_heads: int = 1, num_head_channels: int = -1, use_checkpoint: bool = False, use_new_attention_order: bool = False, ): super().__init__() self.channels = channels if num_head_channels == -1: self.num_heads = num_heads else: assert ( channels % num_head_channels == 0 ), f"q,k,v channels {channels} is not divisible by num_head_channels {num_head_channels}" self.num_heads = channels // num_head_channels self.use_checkpoint = use_checkpoint self.norm = normalization(channels) self.qkv = conv_nd(1, channels, channels * 3, 1) if use_new_attention_order: # split qkv before split heads self.attention = QKVAttention(self.num_heads) else: # split heads before split qkv self.attention = QKVAttentionLegacy(self.num_heads) self.proj_out = zero_module(conv_nd(1, channels, channels, 1)) def forward(self, x: th.Tensor, **kwargs) -> th.Tensor: return checkpoint(self._forward, x) def _forward(self, x: th.Tensor) -> th.Tensor: b, c, *spatial = x.shape x = x.reshape(b, c, -1) qkv = self.qkv(self.norm(x)) h = self.attention(qkv) h = self.proj_out(h) return (x + h).reshape(b, c, *spatial) class QKVAttentionLegacy(nn.Module): """ A module which performs QKV attention. Matches legacy QKVAttention + input/ouput heads shaping """ def __init__(self, n_heads: int): super().__init__() self.n_heads = n_heads def forward(self, qkv: th.Tensor) -> th.Tensor: """ Apply QKV attention. :param qkv: an [N x (H * 3 * C) x T] tensor of Qs, Ks, and Vs. :return: an [N x (H * C) x T] tensor after attention. """ bs, width, length = qkv.shape assert width % (3 * self.n_heads) == 0 ch = width // (3 * self.n_heads) q, k, v = qkv.reshape(bs * self.n_heads, ch * 3, length).split(ch, dim=1) scale = 1 / math.sqrt(math.sqrt(ch)) weight = th.einsum( "bct,bcs->bts", q * scale, k * scale ) # More stable with f16 than dividing afterwards weight = th.softmax(weight.float(), dim=-1).type(weight.dtype) a = th.einsum("bts,bcs->bct", weight, v) return a.reshape(bs, -1, length) class QKVAttention(nn.Module): """ A module which performs QKV attention and splits in a different order. """ def __init__(self, n_heads: int): super().__init__() self.n_heads = n_heads def forward(self, qkv: th.Tensor) -> th.Tensor: """ Apply QKV attention. :param qkv: an [N x (3 * H * C) x T] tensor of Qs, Ks, and Vs. :return: an [N x (H * C) x T] tensor after attention. """ bs, width, length = qkv.shape assert width % (3 * self.n_heads) == 0 ch = width // (3 * self.n_heads) q, k, v = qkv.chunk(3, dim=1) scale = 1 / math.sqrt(math.sqrt(ch)) weight = th.einsum( "bct,bcs->bts", (q * scale).view(bs * self.n_heads, ch, length), (k * scale).view(bs * self.n_heads, ch, length), ) # More stable with f16 than dividing afterwards weight = th.softmax(weight.float(), dim=-1).type(weight.dtype) a = th.einsum("bts,bcs->bct", weight, v.reshape(bs * self.n_heads, ch, length)) return a.reshape(bs, -1, length) class Timestep(nn.Module): def __init__(self, dim: int): super().__init__() self.dim = dim def forward(self, t: th.Tensor) -> th.Tensor: return timestep_embedding(t, self.dim) class UNetModel(nn.Module): """ The full UNet model with attention and timestep embedding. :param in_channels: channels in the input Tensor. :param model_channels: base channel count for the model. :param out_channels: channels in the output Tensor. :param num_res_blocks: number of residual blocks per downsample. :param attention_resolutions: a collection of downsample rates at which attention will take place. May be a set, list, or tuple. For example, if this contains 4, then at 4x downsampling, attention will be used. :param dropout: the dropout probability. :param channel_mult: channel multiplier for each level of the UNet. :param conv_resample: if True, use learned convolutions for upsampling and downsampling. :param dims: determines if the signal is 1D, 2D, or 3D. :param num_classes: if specified (as an int), then this model will be class-conditional with `num_classes` classes. :param use_checkpoint: use gradient checkpointing to reduce memory usage. :param num_heads: the number of attention heads in each attention layer. :param num_heads_channels: if specified, ignore num_heads and instead use a fixed channel width per attention head. :param num_heads_upsample: works with num_heads to set a different number of heads for upsampling. Deprecated. :param use_scale_shift_norm: use a FiLM-like conditioning mechanism. :param resblock_updown: use residual blocks for up/downsampling. :param use_new_attention_order: use a different attention pattern for potentially increased efficiency. """ def __init__( self, in_channels: int, model_channels: int, out_channels: int, num_res_blocks: int, attention_resolutions: int, dropout: float = 0.0, channel_mult: Union[List, Tuple] = (1, 2, 4, 8), conv_resample: bool = True, dims: int = 2, num_classes: Optional[Union[int, str]] = None, use_checkpoint: bool = False, num_heads: int = -1, num_head_channels: int = -1, num_heads_upsample: int = -1, use_scale_shift_norm: bool = False, resblock_updown: bool = False, transformer_depth: int = 1, context_dim: Optional[int] = None, disable_self_attentions: Optional[List[bool]] = None, num_attention_blocks: Optional[List[int]] = None, disable_middle_self_attn: bool = False, disable_middle_transformer: bool = False, use_linear_in_transformer: bool = False, spatial_transformer_attn_type: str = "softmax", adm_in_channels: Optional[int] = None, ): super().__init__() if num_heads_upsample == -1: num_heads_upsample = num_heads if num_heads == -1: assert ( num_head_channels != -1 ), "Either num_heads or num_head_channels has to be set" if num_head_channels == -1: assert ( num_heads != -1 ), "Either num_heads or num_head_channels has to be set" self.in_channels = in_channels self.model_channels = model_channels self.out_channels = out_channels if isinstance(transformer_depth, int): transformer_depth = len(channel_mult) * [transformer_depth] transformer_depth_middle = transformer_depth[-1] if isinstance(num_res_blocks, int): self.num_res_blocks = len(channel_mult) * [num_res_blocks] else: if len(num_res_blocks) != len(channel_mult): raise ValueError( "provide num_res_blocks either as an int (globally constant) or " "as a list/tuple (per-level) with the same length as channel_mult" ) self.num_res_blocks = num_res_blocks if disable_self_attentions is not None: assert len(disable_self_attentions) == len(channel_mult) if num_attention_blocks is not None: assert len(num_attention_blocks) == len(self.num_res_blocks) assert all( map( lambda i: self.num_res_blocks[i] >= num_attention_blocks[i], range(len(num_attention_blocks)), ) ) logpy.info( f"Constructor of UNetModel received num_attention_blocks={num_attention_blocks}. " f"This option has LESS priority than attention_resolutions {attention_resolutions}, " f"i.e., in cases where num_attention_blocks[i] > 0 but 2**i not in attention_resolutions, " f"attention will still not be set." ) self.attention_resolutions = attention_resolutions self.dropout = dropout self.channel_mult = channel_mult self.conv_resample = conv_resample self.num_classes = num_classes self.use_checkpoint = use_checkpoint self.num_heads = num_heads self.num_head_channels = num_head_channels self.num_heads_upsample = num_heads_upsample time_embed_dim = model_channels * 4 self.time_embed = nn.Sequential( linear(model_channels, time_embed_dim), nn.SiLU(), linear(time_embed_dim, time_embed_dim), ) if self.num_classes is not None: if isinstance(self.num_classes, int): self.label_emb = nn.Embedding(num_classes, time_embed_dim) elif self.num_classes == "continuous": logpy.info("setting up linear c_adm embedding layer") self.label_emb = nn.Linear(1, time_embed_dim) elif self.num_classes == "timestep": self.label_emb = nn.Sequential( Timestep(model_channels), nn.Sequential( linear(model_channels, time_embed_dim), nn.SiLU(), linear(time_embed_dim, time_embed_dim), ), ) elif self.num_classes == "sequential": assert adm_in_channels is not None self.label_emb = nn.Sequential( nn.Sequential( linear(adm_in_channels, time_embed_dim), nn.SiLU(), linear(time_embed_dim, time_embed_dim), ) ) else: raise ValueError self.input_blocks = nn.ModuleList( [ TimestepEmbedSequential( conv_nd(dims, in_channels, model_channels, 3, padding=1) ) ] ) self._feature_size = model_channels input_block_chans = [model_channels] ch = model_channels ds = 1 for level, mult in enumerate(channel_mult): for nr in range(self.num_res_blocks[level]): layers = [ ResBlock( ch, time_embed_dim, dropout, out_channels=mult * model_channels, dims=dims, use_checkpoint=use_checkpoint, use_scale_shift_norm=use_scale_shift_norm, ) ] ch = mult * model_channels if ds in attention_resolutions: if num_head_channels == -1: dim_head = ch // num_heads else: num_heads = ch // num_head_channels dim_head = num_head_channels if context_dim is not None and exists(disable_self_attentions): disabled_sa = disable_self_attentions[level] else: disabled_sa = False if ( not exists(num_attention_blocks) or nr < num_attention_blocks[level] ): layers.append( SpatialTransformer( ch, num_heads, dim_head, depth=transformer_depth[level], context_dim=context_dim, disable_self_attn=disabled_sa, use_linear=use_linear_in_transformer, attn_type=spatial_transformer_attn_type, use_checkpoint=use_checkpoint, ) ) self.input_blocks.append(TimestepEmbedSequential(*layers)) self._feature_size += ch input_block_chans.append(ch) if level != len(channel_mult) - 1: out_ch = ch self.input_blocks.append( TimestepEmbedSequential( ResBlock( ch, time_embed_dim, dropout, out_channels=out_ch, dims=dims, use_checkpoint=use_checkpoint, use_scale_shift_norm=use_scale_shift_norm, down=True, ) if resblock_updown else Downsample( ch, conv_resample, dims=dims, out_channels=out_ch ) ) ) ch = out_ch input_block_chans.append(ch) ds *= 2 self._feature_size += ch if num_head_channels == -1: dim_head = ch // num_heads else: num_heads = ch // num_head_channels dim_head = num_head_channels self.middle_block = TimestepEmbedSequential( ResBlock( ch, time_embed_dim, dropout, out_channels=ch, dims=dims, use_checkpoint=use_checkpoint, use_scale_shift_norm=use_scale_shift_norm, ), SpatialTransformer( ch, num_heads, dim_head, depth=transformer_depth_middle, context_dim=context_dim, disable_self_attn=disable_middle_self_attn, use_linear=use_linear_in_transformer, attn_type=spatial_transformer_attn_type, use_checkpoint=use_checkpoint, ) if not disable_middle_transformer else th.nn.Identity(), ResBlock( ch, time_embed_dim, dropout, dims=dims, use_checkpoint=use_checkpoint, use_scale_shift_norm=use_scale_shift_norm, ), ) self._feature_size += ch self.output_blocks = nn.ModuleList([]) for level, mult in list(enumerate(channel_mult))[::-1]: for i in range(self.num_res_blocks[level] + 1): ich = input_block_chans.pop() layers = [ ResBlock( ch + ich, time_embed_dim, dropout, out_channels=model_channels * mult, dims=dims, use_checkpoint=use_checkpoint, use_scale_shift_norm=use_scale_shift_norm, ) ] ch = model_channels * mult if ds in attention_resolutions: if num_head_channels == -1: dim_head = ch // num_heads else: num_heads = ch // num_head_channels dim_head = num_head_channels if exists(disable_self_attentions): disabled_sa = disable_self_attentions[level] else: disabled_sa = False if ( not exists(num_attention_blocks) or i < num_attention_blocks[level] ): layers.append( SpatialTransformer( ch, num_heads, dim_head, depth=transformer_depth[level], context_dim=context_dim, disable_self_attn=disabled_sa, use_linear=use_linear_in_transformer, attn_type=spatial_transformer_attn_type, use_checkpoint=use_checkpoint, ) ) if level and i == self.num_res_blocks[level]: out_ch = ch layers.append( ResBlock( ch, time_embed_dim, dropout, out_channels=out_ch, dims=dims, use_checkpoint=use_checkpoint, use_scale_shift_norm=use_scale_shift_norm, up=True, ) if resblock_updown else Upsample(ch, conv_resample, dims=dims, out_channels=out_ch) ) ds //= 2 self.output_blocks.append(TimestepEmbedSequential(*layers)) self._feature_size += ch self.out = nn.Sequential( normalization(ch), nn.SiLU(), zero_module(conv_nd(dims, model_channels, out_channels, 3, padding=1)), ) def forward( self, x: th.Tensor, timesteps: Optional[th.Tensor] = None, context: Optional[th.Tensor] = None, y: Optional[th.Tensor] = None, **kwargs, ) -> th.Tensor: """ Apply the model to an input batch. :param x: an [N x C x ...] Tensor of inputs. :param timesteps: a 1-D batch of timesteps. :param context: conditioning plugged in via crossattn :param y: an [N] Tensor of labels, if class-conditional. :return: an [N x C x ...] Tensor of outputs. """ assert (y is not None) == ( self.num_classes is not None ), "must specify y if and only if the model is class-conditional" hs = [] t_emb = timestep_embedding(timesteps, self.model_channels, repeat_only=False) emb = self.time_embed(t_emb) if self.num_classes is not None: assert y.shape[0] == x.shape[0] emb = emb + self.label_emb(y) h = x for module in self.input_blocks: h = module(h, emb, context) hs.append(h) h = self.middle_block(h, emb, context) for module in self.output_blocks: h = th.cat([h, hs.pop()], dim=1) h = module(h, emb, context) h = h.type(x.dtype) return self.out(h)