from abc import abstractmethod from functools import partial import math from typing import Iterable import numpy as np import torch as th import torch.nn as nn import torch.nn.functional as F from .diffusion_utils import \ checkpoint, conv_nd, linear, avg_pool_nd, \ zero_module, normalization, timestep_embedding from .attention import SpatialTransformer from lib.model_zoo.common.get_model import get_model, register version = '0' symbol = 'openai' # dummy replace def convert_module_to_f16(x): pass def convert_module_to_f32(x): pass ## go 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: 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): b, c, *_spatial = x.shape x = x.reshape(b, c, -1) # NC(HW) x = th.cat([x.mean(dim=-1, keepdim=True), x], dim=-1) # NC(HW+1) x = x + self.positional_embedding[None, :, :].to(x.dtype) # NC(HW+1) 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, emb): """ 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, emb, context=None): for layer in self: if isinstance(layer, TimestepBlock): x = layer(x, emb) elif isinstance(layer, 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, use_conv, dims=2, out_channels=None, padding=1): super().__init__() self.channels = channels self.out_channels = out_channels or channels self.use_conv = use_conv self.dims = dims if use_conv: self.conv = conv_nd(dims, self.channels, self.out_channels, 3, padding=padding) def forward(self, x): assert x.shape[1] == self.channels if self.dims == 3: x = F.interpolate( x, (x.shape[2], x.shape[3] * 2, x.shape[4] * 2), mode="nearest" ) else: x = F.interpolate(x, scale_factor=2, mode="nearest") if self.use_conv: x = self.conv(x) return x class TransposedUpsample(nn.Module): 'Learned 2x upsampling without padding' def __init__(self, channels, out_channels=None, ks=5): super().__init__() self.channels = channels self.out_channels = out_channels or channels self.up = nn.ConvTranspose2d(self.channels,self.out_channels,kernel_size=ks,stride=2) def forward(self,x): return self.up(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, use_conv, dims=2, out_channels=None,padding=1): 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 use_conv: 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): 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, emb_channels, dropout, out_channels=None, use_conv=False, use_scale_shift_norm=False, dims=2, use_checkpoint=False, up=False, down=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.in_layers = nn.Sequential( normalization(channels), nn.SiLU(), conv_nd(dims, channels, self.out_channels, 3, padding=1), ) 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.emb_layers = nn.Sequential( nn.SiLU(), linear( emb_channels, 2 * self.out_channels if use_scale_shift_norm else self.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, 3, padding=1) ), ) if self.out_channels == channels: self.skip_connection = nn.Identity() elif use_conv: self.skip_connection = conv_nd( dims, channels, self.out_channels, 3, padding=1 ) else: self.skip_connection = conv_nd(dims, channels, self.out_channels, 1) def forward(self, x, emb): """ 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. """ return checkpoint( self._forward, (x, emb), self.parameters(), self.use_checkpoint ) def _forward(self, x, emb): 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) 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: 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, num_heads=1, num_head_channels=-1, use_checkpoint=False, use_new_attention_order=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): return checkpoint(self._forward, (x,), self.parameters(), True) # TODO: check checkpoint usage, is True # TODO: fix the .half call!!! #return pt_checkpoint(self._forward, x) # pytorch def _forward(self, x): 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) def count_flops_attn(model, _x, y): """ A counter for the `thop` package to count the operations in an attention operation. Meant to be used like: macs, params = thop.profile( model, inputs=(inputs, timestamps), custom_ops={QKVAttention: QKVAttention.count_flops}, ) """ b, c, *spatial = y[0].shape num_spatial = int(np.prod(spatial)) # We perform two matmuls with the same number of ops. # The first computes the weight matrix, the second computes # the combination of the value vectors. matmul_ops = 2 * b * (num_spatial ** 2) * c model.total_ops += th.DoubleTensor([matmul_ops]) class QKVAttentionLegacy(nn.Module): """ A module which performs QKV attention. Matches legacy QKVAttention + input/ouput heads shaping """ def __init__(self, n_heads): super().__init__() self.n_heads = n_heads def forward(self, qkv): """ 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) @staticmethod def count_flops(model, _x, y): return count_flops_attn(model, _x, y) class QKVAttention(nn.Module): """ A module which performs QKV attention and splits in a different order. """ def __init__(self, n_heads): super().__init__() self.n_heads = n_heads def forward(self, qkv): """ 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) @staticmethod def count_flops(model, _x, y): return count_flops_attn(model, _x, y) @register('openai_unet', version) 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, image_size, in_channels, model_channels, out_channels, num_res_blocks, attention_resolutions, dropout=0, channel_mult=(1, 2, 4, 8), conv_resample=True, dims=2, num_classes=None, use_checkpoint=False, use_fp16=False, num_heads=-1, num_head_channels=-1, num_heads_upsample=-1, use_scale_shift_norm=False, resblock_updown=False, use_new_attention_order=False, use_spatial_transformer=False, # custom transformer support transformer_depth=1, # custom transformer support context_dim=None, # custom transformer support n_embed=None, # custom support for prediction of discrete ids into codebook of first stage vq model legacy=True, disable_self_attentions=None, num_attention_blocks=None ): super().__init__() if use_spatial_transformer: assert context_dim is not None, 'Fool!! You forgot to include the dimension of your cross-attention conditioning...' if context_dim is not None: assert use_spatial_transformer, 'Fool!! You forgot to use the spatial transformer for your cross-attention conditioning...' from omegaconf.listconfig import ListConfig if type(context_dim) == ListConfig: context_dim = list(context_dim) 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.image_size = image_size self.in_channels = in_channels self.model_channels = model_channels self.out_channels = out_channels 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 #self.num_res_blocks = num_res_blocks if disable_self_attentions is not None: # should be a list of booleans, indicating whether to disable self-attention in TransformerBlocks or not 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)))) print(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.") # todo: convert to warning 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.dtype = th.float16 if use_fp16 else th.float32 self.num_heads = num_heads self.num_head_channels = num_head_channels self.num_heads_upsample = num_heads_upsample self.predict_codebook_ids = n_embed is not None 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: self.label_emb = nn.Embedding(num_classes, time_embed_dim) 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 legacy: #num_heads = 1 dim_head = ch // num_heads if use_spatial_transformer else num_head_channels if disable_self_attentions is not None: disabled_sa = disable_self_attentions[level] else: disabled_sa = False if num_attention_blocks is None or nr < num_attention_blocks[level]: layers.append( AttentionBlock( ch, use_checkpoint=use_checkpoint, num_heads=num_heads, num_head_channels=dim_head, use_new_attention_order=use_new_attention_order, ) if not use_spatial_transformer else SpatialTransformer( ch, num_heads, dim_head, depth=transformer_depth, context_dim=context_dim, disable_self_attn=disabled_sa ) ) 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 if legacy: #num_heads = 1 dim_head = ch // num_heads if use_spatial_transformer else num_head_channels self.middle_block = TimestepEmbedSequential( ResBlock( ch, time_embed_dim, dropout, dims=dims, use_checkpoint=use_checkpoint, use_scale_shift_norm=use_scale_shift_norm, ), AttentionBlock( ch, use_checkpoint=use_checkpoint, num_heads=num_heads, num_head_channels=dim_head, use_new_attention_order=use_new_attention_order, ) if not use_spatial_transformer else SpatialTransformer( # always uses a self-attn ch, num_heads, dim_head, depth=transformer_depth, context_dim=context_dim ), 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 legacy: #num_heads = 1 dim_head = ch // num_heads if use_spatial_transformer else num_head_channels if disable_self_attentions is not None: disabled_sa = disable_self_attentions[level] else: disabled_sa = False if num_attention_blocks is None or i < num_attention_blocks[level]: layers.append( AttentionBlock( ch, use_checkpoint=use_checkpoint, num_heads=num_heads_upsample, num_head_channels=dim_head, use_new_attention_order=use_new_attention_order, ) if not use_spatial_transformer else SpatialTransformer( ch, num_heads, dim_head, depth=transformer_depth, context_dim=context_dim, disable_self_attn=disabled_sa ) ) 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)), ) if self.predict_codebook_ids: self.id_predictor = nn.Sequential( normalization(ch), conv_nd(dims, model_channels, n_embed, 1), #nn.LogSoftmax(dim=1) # change to cross_entropy and produce non-normalized logits ) def convert_to_fp16(self): """ Convert the torso of the model to float16. """ self.input_blocks.apply(convert_module_to_f16) self.middle_block.apply(convert_module_to_f16) self.output_blocks.apply(convert_module_to_f16) def convert_to_fp32(self): """ Convert the torso of the model to float32. """ self.input_blocks.apply(convert_module_to_f32) self.middle_block.apply(convert_module_to_f32) self.output_blocks.apply(convert_module_to_f32) def forward(self, x, timesteps=None, context=None, y=None,**kwargs): """ 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 == (x.shape[0],) emb = emb + self.label_emb(y) h = x.type(self.dtype) 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) if self.predict_codebook_ids: return self.id_predictor(h) else: return self.out(h) class EncoderUNetModel(nn.Module): """ The half UNet model with attention and timestep embedding. For usage, see UNet. """ def __init__( self, image_size, in_channels, model_channels, out_channels, num_res_blocks, attention_resolutions, dropout=0, channel_mult=(1, 2, 4, 8), conv_resample=True, dims=2, use_checkpoint=False, use_fp16=False, num_heads=1, num_head_channels=-1, num_heads_upsample=-1, use_scale_shift_norm=False, resblock_updown=False, use_new_attention_order=False, pool="adaptive", *args, **kwargs ): super().__init__() if num_heads_upsample == -1: num_heads_upsample = num_heads self.in_channels = in_channels self.model_channels = model_channels self.out_channels = out_channels self.num_res_blocks = num_res_blocks self.attention_resolutions = attention_resolutions self.dropout = dropout self.channel_mult = channel_mult self.conv_resample = conv_resample self.use_checkpoint = use_checkpoint self.dtype = th.float16 if use_fp16 else th.float32 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), ) 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 _ in range(num_res_blocks): 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: layers.append( AttentionBlock( ch, use_checkpoint=use_checkpoint, num_heads=num_heads, num_head_channels=num_head_channels, use_new_attention_order=use_new_attention_order, ) ) 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 self.middle_block = TimestepEmbedSequential( ResBlock( ch, time_embed_dim, dropout, dims=dims, use_checkpoint=use_checkpoint, use_scale_shift_norm=use_scale_shift_norm, ), AttentionBlock( ch, use_checkpoint=use_checkpoint, num_heads=num_heads, num_head_channels=num_head_channels, use_new_attention_order=use_new_attention_order, ), 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.pool = pool if pool == "adaptive": self.out = nn.Sequential( normalization(ch), nn.SiLU(), nn.AdaptiveAvgPool2d((1, 1)), zero_module(conv_nd(dims, ch, out_channels, 1)), nn.Flatten(), ) elif pool == "attention": assert num_head_channels != -1 self.out = nn.Sequential( normalization(ch), nn.SiLU(), AttentionPool2d( (image_size // ds), ch, num_head_channels, out_channels ), ) elif pool == "spatial": self.out = nn.Sequential( nn.Linear(self._feature_size, 2048), nn.ReLU(), nn.Linear(2048, self.out_channels), ) elif pool == "spatial_v2": self.out = nn.Sequential( nn.Linear(self._feature_size, 2048), normalization(2048), nn.SiLU(), nn.Linear(2048, self.out_channels), ) else: raise NotImplementedError(f"Unexpected {pool} pooling") def convert_to_fp16(self): """ Convert the torso of the model to float16. """ self.input_blocks.apply(convert_module_to_f16) self.middle_block.apply(convert_module_to_f16) def convert_to_fp32(self): """ Convert the torso of the model to float32. """ self.input_blocks.apply(convert_module_to_f32) self.middle_block.apply(convert_module_to_f32) def forward(self, x, timesteps): """ 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. :return: an [N x K] Tensor of outputs. """ emb = self.time_embed(timestep_embedding(timesteps, self.model_channels)) results = [] h = x.type(self.dtype) for module in self.input_blocks: h = module(h, emb) if self.pool.startswith("spatial"): results.append(h.type(x.dtype).mean(dim=(2, 3))) h = self.middle_block(h, emb) if self.pool.startswith("spatial"): results.append(h.type(x.dtype).mean(dim=(2, 3))) h = th.cat(results, axis=-1) return self.out(h) else: h = h.type(x.dtype) return self.out(h) ####################### # Unet with self-attn # ####################### from .attention import SpatialTransformerNoContext @register('openai_unet_nocontext', version) class UNetModelNoContext(nn.Module): def __init__( self, image_size, in_channels, model_channels, out_channels, num_res_blocks, attention_resolutions, dropout=0, channel_mult=(1, 2, 4, 8), conv_resample=True, dims=2, num_classes=None, use_checkpoint=False, use_fp16=False, num_heads=-1, num_head_channels=-1, num_heads_upsample=-1, use_scale_shift_norm=False, resblock_updown=False, use_new_attention_order=False, use_spatial_transformer=False, # custom transformer support transformer_depth=1, # custom transformer support n_embed=None, # custom support for prediction of discrete ids into codebook of first stage vq model legacy=True, num_attention_blocks=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.image_size = image_size self.in_channels = in_channels self.model_channels = model_channels self.out_channels = out_channels 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 #self.num_res_blocks = num_res_blocks 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)))) print(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.") # todo: convert to warning 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.dtype = th.float16 if use_fp16 else th.float32 self.num_heads = num_heads self.num_head_channels = num_head_channels self.num_heads_upsample = num_heads_upsample self.predict_codebook_ids = n_embed is not None 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: self.label_emb = nn.Embedding(num_classes, time_embed_dim) 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 legacy: #num_heads = 1 dim_head = ch // num_heads if use_spatial_transformer else num_head_channels if not exists(num_attention_blocks) or nr < num_attention_blocks[level]: layers.append( AttentionBlock( ch, use_checkpoint=use_checkpoint, num_heads=num_heads, num_head_channels=dim_head, use_new_attention_order=use_new_attention_order, ) if not use_spatial_transformer else SpatialTransformerNoContext( ch, num_heads, dim_head, depth=transformer_depth ) ) 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 if legacy: #num_heads = 1 dim_head = ch // num_heads if use_spatial_transformer else num_head_channels self.middle_block = TimestepEmbedSequential( ResBlock( ch, time_embed_dim, dropout, dims=dims, use_checkpoint=use_checkpoint, use_scale_shift_norm=use_scale_shift_norm, ), AttentionBlock( ch, use_checkpoint=use_checkpoint, num_heads=num_heads, num_head_channels=dim_head, use_new_attention_order=use_new_attention_order, ) if not use_spatial_transformer else SpatialTransformerNoContext( # always uses a self-attn ch, num_heads, dim_head, depth=transformer_depth ), 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 legacy: #num_heads = 1 dim_head = ch // num_heads if use_spatial_transformer else num_head_channels if not exists(num_attention_blocks) or i < num_attention_blocks[level]: layers.append( AttentionBlock( ch, use_checkpoint=use_checkpoint, num_heads=num_heads_upsample, num_head_channels=dim_head, use_new_attention_order=use_new_attention_order, ) if not use_spatial_transformer else SpatialTransformerNoContext( ch, num_heads, dim_head, depth=transformer_depth, ) ) 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)), ) if self.predict_codebook_ids: self.id_predictor = nn.Sequential( normalization(ch), conv_nd(dims, model_channels, n_embed, 1), #nn.LogSoftmax(dim=1) # change to cross_entropy and produce non-normalized logits ) def forward(self, x, timesteps): assert self.num_classes is None, \ "not supported" hs = [] t_emb = timestep_embedding(timesteps, self.model_channels, repeat_only=False) emb = self.time_embed(t_emb) h = x.type(self.dtype) for module in self.input_blocks: h = module(h, emb) hs.append(h) h = self.middle_block(h, emb) for module in self.output_blocks: h = th.cat([h, hs.pop()], dim=1) h = module(h, emb) h = h.type(x.dtype) if self.predict_codebook_ids: return self.id_predictor(h) else: return self.out(h) @register('openai_unet_nocontext_noatt', version) class UNetModelNoContextNoAtt(nn.Module): def __init__( self, in_channels, model_channels, out_channels, num_res_blocks, dropout=0, channel_mult=(1, 2, 4, 8), conv_resample=True, dims=2, num_classes=None, use_checkpoint=False, use_fp16=False, use_scale_shift_norm=False, resblock_updown=False, n_embed=None,): super().__init__() self.in_channels = in_channels self.model_channels = model_channels self.out_channels = out_channels 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 #self.num_res_blocks = num_res_blocks self.dropout = dropout self.channel_mult = channel_mult self.conv_resample = conv_resample self.num_classes = num_classes self.use_checkpoint = use_checkpoint self.dtype = th.float16 if use_fp16 else th.float32 self.predict_codebook_ids = n_embed is not None 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: self.label_emb = nn.Embedding(num_classes, time_embed_dim) 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 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 self.middle_block = TimestepEmbedSequential( ResBlock( ch, time_embed_dim, dropout, dims=dims, use_checkpoint=use_checkpoint, use_scale_shift_norm=use_scale_shift_norm, ), 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 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)), ) if self.predict_codebook_ids: self.id_predictor = nn.Sequential( normalization(ch), conv_nd(dims, model_channels, n_embed, 1), #nn.LogSoftmax(dim=1) # change to cross_entropy and produce non-normalized logits ) def forward(self, x, timesteps): assert self.num_classes is None, \ "not supported" hs = [] t_emb = timestep_embedding(timesteps, self.model_channels, repeat_only=False) emb = self.time_embed(t_emb) h = x.type(self.dtype) for module in self.input_blocks: h = module(h, emb) hs.append(h) h = self.middle_block(h, emb) for module in self.output_blocks: h = th.cat([h, hs.pop()], dim=1) h = module(h, emb) h = h.type(x.dtype) if self.predict_codebook_ids: return self.id_predictor(h) else: return self.out(h) @register('openai_unet_nocontext_noatt_decoderonly', version) class UNetModelNoContextNoAttDecoderOnly(nn.Module): def __init__( self, in_channels, out_channels, model_channels, num_res_blocks, dropout=0, channel_mult=(4, 2, 1), conv_resample=True, dims=2, num_classes=None, use_checkpoint=False, use_fp16=False, use_scale_shift_norm=False, resblock_updown=False, n_embed=None,): super().__init__() self.in_channels = in_channels self.model_channels = model_channels self.out_channels = out_channels 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 #self.num_res_blocks = num_res_blocks self.dropout = dropout self.channel_mult = channel_mult self.conv_resample = conv_resample self.num_classes = num_classes self.use_checkpoint = use_checkpoint self.dtype = th.float16 if use_fp16 else th.float32 self.predict_codebook_ids = n_embed is not None 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: self.label_emb = nn.Embedding(num_classes, time_embed_dim) self._feature_size = model_channels ch = model_channels * self.channel_mult[0] self.output_blocks = nn.ModuleList( [ TimestepEmbedSequential( conv_nd(dims, in_channels, ch, 3, padding=1) ) ] ) for level, mult in enumerate(channel_mult): for i in range(self.num_res_blocks[level]): layers = [ ResBlock( ch, 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 level != len(channel_mult)-1 and (i == self.num_res_blocks[level]-1): 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) ) self.output_blocks.append(TimestepEmbedSequential(*layers)) self.out = nn.Sequential( normalization(ch), nn.SiLU(), zero_module(conv_nd(dims, model_channels, out_channels, 3, padding=1)), ) if self.predict_codebook_ids: self.id_predictor = nn.Sequential( normalization(ch), conv_nd(dims, model_channels, n_embed, 1), #nn.LogSoftmax(dim=1) # change to cross_entropy and produce non-normalized logits ) def forward(self, x, timesteps): assert self.num_classes is None, \ "not supported" hs = [] t_emb = timestep_embedding(timesteps, self.model_channels, repeat_only=False) emb = self.time_embed(t_emb) h = x.type(self.dtype) for module in self.output_blocks: h = module(h, emb) h = h.type(x.dtype) if self.predict_codebook_ids: return self.id_predictor(h) else: return self.out(h) ######################### # Double Attention Unet # ######################### from .attention import DualSpatialTransformer class TimestepEmbedSequentialExtended(nn.Sequential, TimestepBlock): def forward(self, x, emb, context=None, which_attn=None): for layer in self: if isinstance(layer, TimestepBlock): x = layer(x, emb) elif isinstance(layer, SpatialTransformer): x = layer(x, context) elif isinstance(layer, DualSpatialTransformer): x = layer(x, context, which=which_attn) else: x = layer(x) return x @register('openai_unet_dual_context', version) class UNetModelDualContext(nn.Module): def __init__( self, image_size, in_channels, model_channels, out_channels, num_res_blocks, attention_resolutions, dropout=0, channel_mult=(1, 2, 4, 8), conv_resample=True, dims=2, num_classes=None, use_checkpoint=False, use_fp16=False, num_heads=-1, num_head_channels=-1, num_heads_upsample=-1, use_scale_shift_norm=False, resblock_updown=False, use_new_attention_order=False, use_spatial_transformer=False, # custom transformer support transformer_depth=1, # custom transformer support context_dim=None, # custom transformer support n_embed=None, # custom support for prediction of discrete ids into codebook of first stage vq model legacy=True, disable_self_attentions=None, num_attention_blocks=None, ): super().__init__() if use_spatial_transformer: assert context_dim is not None, 'Fool!! You forgot to include the dimension of your cross-attention conditioning...' if context_dim is not None: assert use_spatial_transformer, 'Fool!! You forgot to use the spatial transformer for your cross-attention conditioning...' from omegaconf.listconfig import ListConfig if type(context_dim) == ListConfig: context_dim = list(context_dim) 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.image_size = image_size self.in_channels = in_channels self.model_channels = model_channels self.out_channels = out_channels 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 #self.num_res_blocks = num_res_blocks if disable_self_attentions is not None: # should be a list of booleans, indicating whether to disable self-attention in TransformerBlocks or not 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)))) print(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.") # todo: convert to warning 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.dtype = th.float16 if use_fp16 else th.float32 self.num_heads = num_heads self.num_head_channels = num_head_channels self.num_heads_upsample = num_heads_upsample self.predict_codebook_ids = n_embed is not None 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: self.label_emb = nn.Embedding(num_classes, time_embed_dim) self.input_blocks = nn.ModuleList( [ TimestepEmbedSequentialExtended( 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 legacy: #num_heads = 1 dim_head = ch // num_heads if use_spatial_transformer else num_head_channels if disable_self_attentions is not None: disabled_sa = disable_self_attentions[level] else: disabled_sa = False if num_attention_blocks is None or nr < num_attention_blocks[level]: layers.append( AttentionBlock( ch, use_checkpoint=use_checkpoint, num_heads=num_heads, num_head_channels=dim_head, use_new_attention_order=use_new_attention_order, ) if not use_spatial_transformer else DualSpatialTransformer( ch, num_heads, dim_head, depth=transformer_depth, context_dim=context_dim, disable_self_attn=disabled_sa ) ) self.input_blocks.append(TimestepEmbedSequentialExtended(*layers)) self._feature_size += ch input_block_chans.append(ch) if level != len(channel_mult) - 1: out_ch = ch self.input_blocks.append( TimestepEmbedSequentialExtended( 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 if legacy: #num_heads = 1 dim_head = ch // num_heads if use_spatial_transformer else num_head_channels self.middle_block = TimestepEmbedSequentialExtended( ResBlock( ch, time_embed_dim, dropout, dims=dims, use_checkpoint=use_checkpoint, use_scale_shift_norm=use_scale_shift_norm, ), AttentionBlock( ch, use_checkpoint=use_checkpoint, num_heads=num_heads, num_head_channels=dim_head, use_new_attention_order=use_new_attention_order, ) if not use_spatial_transformer else DualSpatialTransformer( # always uses a self-attn ch, num_heads, dim_head, depth=transformer_depth, context_dim=context_dim ), 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 legacy: #num_heads = 1 dim_head = ch // num_heads if use_spatial_transformer else num_head_channels if disable_self_attentions is not None: disabled_sa = disable_self_attentions[level] else: disabled_sa = False if num_attention_blocks is None or i < num_attention_blocks[level]: layers.append( AttentionBlock( ch, use_checkpoint=use_checkpoint, num_heads=num_heads_upsample, num_head_channels=dim_head, use_new_attention_order=use_new_attention_order, ) if not use_spatial_transformer else DualSpatialTransformer( ch, num_heads, dim_head, depth=transformer_depth, context_dim=context_dim, disable_self_attn=disabled_sa ) ) 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(TimestepEmbedSequentialExtended(*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)), ) if self.predict_codebook_ids: self.id_predictor = nn.Sequential( normalization(ch), conv_nd(dims, model_channels, n_embed, 1), #nn.LogSoftmax(dim=1) # change to cross_entropy and produce non-normalized logits ) def forward(self, x, timesteps=None, context=None, y=None, which_attn=None, **kwargs): """ 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 == (x.shape[0],) emb = emb + self.label_emb(y) h = x.type(self.dtype) for module in self.input_blocks: h = module(h, emb, context, which_attn=which_attn) hs.append(h) h = self.middle_block(h, emb, context, which_attn=which_attn) for module in self.output_blocks: h = th.cat([h, hs.pop()], dim=1) h = module(h, emb, context, which_attn=which_attn) h = h.type(x.dtype) if self.predict_codebook_ids: return self.id_predictor(h) else: return self.out(h) ########### # VD Unet # ########### from functools import partial @register('openai_unet_2d', version) class UNetModel2D(nn.Module): def __init__(self, input_channels, model_channels, output_channels, context_dim=768, num_noattn_blocks=(2, 2, 2, 2), channel_mult=(1, 2, 4, 8), with_attn=[True, True, True, False], num_heads=8, use_checkpoint=True, ): super().__init__() ResBlockPreset = partial( ResBlock, dropout=0, dims=2, use_checkpoint=use_checkpoint, use_scale_shift_norm=False) self.input_channels = input_channels self.model_channels = model_channels self.num_noattn_blocks = num_noattn_blocks self.channel_mult = channel_mult self.num_heads = num_heads ################## # Time embedding # ################## 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),) ################ # input_blocks # ################ current_channel = model_channels input_blocks = [ TimestepEmbedSequential( nn.Conv2d(input_channels, model_channels, 3, padding=1, bias=True))] input_block_channels = [current_channel] for level_idx, mult in enumerate(channel_mult): for _ in range(self.num_noattn_blocks[level_idx]): layers = [ ResBlockPreset( current_channel, time_embed_dim, out_channels = mult * model_channels,)] current_channel = mult * model_channels dim_head = current_channel // num_heads if with_attn[level_idx]: layers += [ SpatialTransformer( current_channel, num_heads, dim_head, depth=1, context_dim=context_dim, )] input_blocks += [TimestepEmbedSequential(*layers)] input_block_channels.append(current_channel) if level_idx != len(channel_mult) - 1: input_blocks += [ TimestepEmbedSequential( Downsample( current_channel, use_conv=True, dims=2, out_channels=current_channel,))] input_block_channels.append(current_channel) self.input_blocks = nn.ModuleList(input_blocks) ################# # middle_blocks # ################# middle_block = [ ResBlockPreset( current_channel, time_embed_dim,), SpatialTransformer( current_channel, num_heads, dim_head, depth=1, context_dim=context_dim, ), ResBlockPreset( current_channel, time_embed_dim,),] self.middle_block = TimestepEmbedSequential(*middle_block) ################# # output_blocks # ################# output_blocks = [] for level_idx, mult in list(enumerate(channel_mult))[::-1]: for block_idx in range(self.num_noattn_blocks[level_idx] + 1): extra_channel = input_block_channels.pop() layers = [ ResBlockPreset( current_channel + extra_channel, time_embed_dim, out_channels = model_channels * mult,) ] current_channel = model_channels * mult dim_head = current_channel // num_heads if with_attn[level_idx]: layers += [ SpatialTransformer( current_channel, num_heads, dim_head, depth=1, context_dim=context_dim,)] if level_idx!=0 and block_idx==self.num_noattn_blocks[level_idx]: layers += [ Upsample( current_channel, use_conv=True, dims=2, out_channels=current_channel)] output_blocks += [TimestepEmbedSequential(*layers)] self.output_blocks = nn.ModuleList(output_blocks) self.out = nn.Sequential( normalization(current_channel), nn.SiLU(), zero_module(nn.Conv2d(model_channels, output_channels, 3, padding=1)),) def forward(self, x, timesteps=None, context=None): hs = [] t_emb = timestep_embedding(timesteps, self.model_channels, repeat_only=False) emb = self.time_embed(t_emb) 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) return self.out(h) class FCBlock(TimestepBlock): def __init__( self, channels, emb_channels, dropout, out_channels=None, use_checkpoint=False, ): super().__init__() self.channels = channels self.emb_channels = emb_channels self.dropout = dropout self.out_channels = out_channels or channels self.use_checkpoint = use_checkpoint self.in_layers = nn.Sequential( normalization(channels), nn.SiLU(), nn.Conv2d(channels, self.out_channels, 1, padding=0),) self.emb_layers = nn.Sequential( nn.SiLU(), linear(emb_channels, self.out_channels,),) self.out_layers = nn.Sequential( normalization(self.out_channels), nn.SiLU(), nn.Dropout(p=dropout), zero_module(nn.Conv2d(self.out_channels, self.out_channels, 1, padding=0)), ) if self.out_channels == channels: self.skip_connection = nn.Identity() else: self.skip_connection = nn.Conv2d(channels, self.out_channels, 1, padding=0) def forward(self, x, emb): if len(x.shape) == 2: x = x[:, :, None, None] elif len(x.shape) == 4: pass else: raise ValueError y = checkpoint( self._forward, (x, emb), self.parameters(), self.use_checkpoint) if len(x.shape) == 2: return y[:, :, 0, 0] elif len(x.shape) == 4: return y def _forward(self, x, emb): h = self.in_layers(x) emb_out = self.emb_layers(emb).type(h.dtype) while len(emb_out.shape) < len(h.shape): emb_out = emb_out[..., None] h = h + emb_out h = self.out_layers(h) return self.skip_connection(x) + h @register('openai_unet_0d', version) class UNetModel0D(nn.Module): def __init__(self, input_channels, model_channels, output_channels, context_dim=768, num_noattn_blocks=(2, 2, 2, 2), channel_mult=(1, 2, 4, 8), with_attn=[True, True, True, False], num_heads=8, use_checkpoint=True, ): super().__init__() FCBlockPreset = partial(FCBlock, dropout=0, use_checkpoint=use_checkpoint) self.input_channels = input_channels self.model_channels = model_channels self.num_noattn_blocks = num_noattn_blocks self.channel_mult = channel_mult self.num_heads = num_heads ################## # Time embedding # ################## 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),) ################ # input_blocks # ################ current_channel = model_channels input_blocks = [ TimestepEmbedSequential( nn.Conv2d(input_channels, model_channels, 1, padding=0, bias=True))] input_block_channels = [current_channel] for level_idx, mult in enumerate(channel_mult): for _ in range(self.num_noattn_blocks[level_idx]): layers = [ FCBlockPreset( current_channel, time_embed_dim, out_channels = mult * model_channels,)] current_channel = mult * model_channels dim_head = current_channel // num_heads if with_attn[level_idx]: layers += [ SpatialTransformer( current_channel, num_heads, dim_head, depth=1, context_dim=context_dim, )] input_blocks += [TimestepEmbedSequential(*layers)] input_block_channels.append(current_channel) if level_idx != len(channel_mult) - 1: input_blocks += [ TimestepEmbedSequential( Downsample( current_channel, use_conv=True, dims=2, out_channels=current_channel,))] input_block_channels.append(current_channel) self.input_blocks = nn.ModuleList(input_blocks) ################# # middle_blocks # ################# middle_block = [ FCBlockPreset( current_channel, time_embed_dim,), SpatialTransformer( current_channel, num_heads, dim_head, depth=1, context_dim=context_dim, ), FCBlockPreset( current_channel, time_embed_dim,),] self.middle_block = TimestepEmbedSequential(*middle_block) ################# # output_blocks # ################# output_blocks = [] for level_idx, mult in list(enumerate(channel_mult))[::-1]: for block_idx in range(self.num_noattn_blocks[level_idx] + 1): extra_channel = input_block_channels.pop() layers = [ FCBlockPreset( current_channel + extra_channel, time_embed_dim, out_channels = model_channels * mult,) ] current_channel = model_channels * mult dim_head = current_channel // num_heads if with_attn[level_idx]: layers += [ SpatialTransformer( current_channel, num_heads, dim_head, depth=1, context_dim=context_dim,)] if level_idx!=0 and block_idx==self.num_noattn_blocks[level_idx]: layers += [ nn.Conv2d(current_channel, current_channel, 1, padding=0)] output_blocks += [TimestepEmbedSequential(*layers)] self.output_blocks = nn.ModuleList(output_blocks) self.out = nn.Sequential( normalization(current_channel), nn.SiLU(), zero_module(nn.Conv2d(model_channels, output_channels, 1, padding=0)),) def forward(self, x, timesteps=None, context=None): hs = [] t_emb = timestep_embedding(timesteps, self.model_channels, repeat_only=False) emb = self.time_embed(t_emb) 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) return self.out(h) class Linear_MultiDim(nn.Linear): def __init__(self, in_features, out_features, *args, **kwargs): in_features = [in_features] if isinstance(in_features, int) else list(in_features) out_features = [out_features] if isinstance(out_features, int) else list(out_features) self.in_features_multidim = in_features self.out_features_multidim = out_features super().__init__( np.array(in_features).prod(), np.array(out_features).prod(), *args, **kwargs) def forward(self, x): shape = x.shape n = len(self.in_features_multidim) x = x.view(*shape[0:-n], self.in_features) y = super().forward(x) y = y.view(*shape[0:-n], *self.out_features_multidim) return y class FCBlock_MultiDim(FCBlock): def __init__( self, channels, emb_channels, dropout, out_channels=None, use_checkpoint=False,): channels = [channels] if isinstance(channels, int) else list(channels) channels_all = np.array(channels).prod() self.channels_multidim = channels if out_channels is not None: out_channels = [out_channels] if isinstance(out_channels, int) else list(out_channels) out_channels_all = np.array(out_channels).prod() self.out_channels_multidim = out_channels else: out_channels_all = channels_all self.out_channels_multidim = self.channels_multidim self.channels = channels super().__init__( channels = channels_all, emb_channels = emb_channels, dropout = dropout, out_channels = out_channels_all, use_checkpoint = use_checkpoint,) def forward(self, x, emb): shape = x.shape n = len(self.channels_multidim) x = x.view(*shape[0:-n], self.channels, 1, 1) x = x.view(-1, self.channels, 1, 1) y = checkpoint( self._forward, (x, emb), self.parameters(), self.use_checkpoint) y = y.view(*shape[0:-n], -1) y = y.view(*shape[0:-n], *self.out_channels_multidim) return y @register('openai_unet_0dmd', version) class UNetModel0D_MultiDim(nn.Module): def __init__(self, input_channels, model_channels, output_channels, context_dim=768, num_noattn_blocks=(2, 2, 2, 2), channel_mult=(1, 2, 4, 8), second_dim=(4, 4, 4, 4), with_attn=[True, True, True, False], num_heads=8, use_checkpoint=True, ): super().__init__() FCBlockPreset = partial(FCBlock_MultiDim, dropout=0, use_checkpoint=use_checkpoint) self.input_channels = input_channels self.model_channels = model_channels self.num_noattn_blocks = num_noattn_blocks self.channel_mult = channel_mult self.second_dim = second_dim self.num_heads = num_heads ################## # Time embedding # ################## 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),) ################ # input_blocks # ################ sdim = second_dim[0] current_channel = [model_channels, sdim, 1] input_blocks = [ TimestepEmbedSequential( Linear_MultiDim([input_channels, 1, 1], current_channel, bias=True))] input_block_channels = [current_channel] for level_idx, (mult, sdim) in enumerate(zip(channel_mult, second_dim)): for _ in range(self.num_noattn_blocks[level_idx]): layers = [ FCBlockPreset( current_channel, time_embed_dim, out_channels = [mult*model_channels, sdim, 1],)] current_channel = [mult*model_channels, sdim, 1] dim_head = current_channel[0] // num_heads if with_attn[level_idx]: layers += [ SpatialTransformer( current_channel[0], num_heads, dim_head, depth=1, context_dim=context_dim, )] input_blocks += [TimestepEmbedSequential(*layers)] input_block_channels.append(current_channel) if level_idx != len(channel_mult) - 1: input_blocks += [ TimestepEmbedSequential( Linear_MultiDim(current_channel, current_channel, bias=True, ))] input_block_channels.append(current_channel) self.input_blocks = nn.ModuleList(input_blocks) ################# # middle_blocks # ################# middle_block = [ FCBlockPreset( current_channel, time_embed_dim, ), SpatialTransformer( current_channel[0], num_heads, dim_head, depth=1, context_dim=context_dim, ), FCBlockPreset( current_channel, time_embed_dim, ),] self.middle_block = TimestepEmbedSequential(*middle_block) ################# # output_blocks # ################# output_blocks = [] for level_idx, (mult, sdim) in list(enumerate(zip(channel_mult, second_dim)))[::-1]: for block_idx in range(self.num_noattn_blocks[level_idx] + 1): extra_channel = input_block_channels.pop() layers = [ FCBlockPreset( [current_channel[0] + extra_channel[0]] + current_channel[1:], time_embed_dim, out_channels = [mult*model_channels, sdim, 1], )] current_channel = [mult*model_channels, sdim, 1] dim_head = current_channel[0] // num_heads if with_attn[level_idx]: layers += [ SpatialTransformer( current_channel[0], num_heads, dim_head, depth=1, context_dim=context_dim,)] if level_idx!=0 and block_idx==self.num_noattn_blocks[level_idx]: layers += [ Linear_MultiDim(current_channel, current_channel, bias=True, )] output_blocks += [TimestepEmbedSequential(*layers)] self.output_blocks = nn.ModuleList(output_blocks) self.out = nn.Sequential( normalization(current_channel[0]), nn.SiLU(), zero_module(Linear_MultiDim(current_channel, [output_channels, 1, 1], bias=True, )),) def forward(self, x, timesteps=None, context=None): hs = [] t_emb = timestep_embedding(timesteps, self.model_channels, repeat_only=False) emb = self.time_embed(t_emb) 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) return self.out(h) @register('openai_unet_vd', version) class UNetModelVD(nn.Module): def __init__(self, unet_image_cfg, unet_test_cfg, ): super().__init__() self.unet_image = get_model()(unet_image_cfg) self.unet_text = get_model()(unet_test_cfg) self.time_embed = self.unet_image.time_embed del self.unet_image.time_embed del self.unet_text.time_embed self.model_channels = self.unet_image.model_channels def forward(self, x, timesteps, context, xtype='image', ctype='prompt'): hs = [] t_emb = timestep_embedding(timesteps, self.model_channels, repeat_only=False) emb = self.time_embed(t_emb) if xtype == 'text': x = x[:, :, None, None] h = x for i_module, t_module in zip(self.unet_image.input_blocks, self.unet_text.input_blocks): h = self.mixed_run(i_module, t_module, h, emb, context, xtype, ctype) hs.append(h) h = self.mixed_run( self.unet_image.middle_block, self.unet_text.middle_block, h, emb, context, xtype, ctype) for i_module, t_module in zip(self.unet_image.output_blocks, self.unet_text.output_blocks): h = th.cat([h, hs.pop()], dim=1) h = self.mixed_run(i_module, t_module, h, emb, context, xtype, ctype) if xtype == 'image': return self.unet_image.out(h) elif xtype == 'text': return self.unet_text.out(h).squeeze(-1).squeeze(-1) def mixed_run(self, inet, tnet, x, emb, context, xtype, ctype): h = x for ilayer, tlayer in zip(inet, tnet): if isinstance(ilayer, TimestepBlock) and xtype=='image': h = ilayer(h, emb) elif isinstance(tlayer, TimestepBlock) and xtype=='text': h = tlayer(h, emb) elif isinstance(ilayer, SpatialTransformer) and ctype=='vision': h = ilayer(h, context) elif isinstance(ilayer, SpatialTransformer) and ctype=='prompt': h = tlayer(h, context) elif xtype=='image': h = ilayer(h) elif xtype == 'text': h = tlayer(h) else: raise ValueError return h def forward_dc(self, x, timesteps, c0, c1, xtype, c0_type, c1_type, mixed_ratio): hs = [] t_emb = timestep_embedding(timesteps, self.model_channels, repeat_only=False) emb = self.time_embed(t_emb) if xtype == 'text': x = x[:, :, None, None] h = x for i_module, t_module in zip(self.unet_image.input_blocks, self.unet_text.input_blocks): h = self.mixed_run_dc(i_module, t_module, h, emb, c0, c1, xtype, c0_type, c1_type, mixed_ratio) hs.append(h) h = self.mixed_run_dc( self.unet_image.middle_block, self.unet_text.middle_block, h, emb, c0, c1, xtype, c0_type, c1_type, mixed_ratio) for i_module, t_module in zip(self.unet_image.output_blocks, self.unet_text.output_blocks): h = th.cat([h, hs.pop()], dim=1) h = self.mixed_run_dc(i_module, t_module, h, emb, c0, c1, xtype, c0_type, c1_type, mixed_ratio) if xtype == 'image': return self.unet_image.out(h) elif xtype == 'text': return self.unet_text.out(h).squeeze(-1).squeeze(-1) def mixed_run_dc(self, inet, tnet, x, emb, c0, c1, xtype, c0_type, c1_type, mixed_ratio): h = x for ilayer, tlayer in zip(inet, tnet): if isinstance(ilayer, TimestepBlock) and xtype=='image': h = ilayer(h, emb) elif isinstance(tlayer, TimestepBlock) and xtype=='text': h = tlayer(h, emb) elif isinstance(ilayer, SpatialTransformer): h0 = ilayer(h, c0)-h if c0_type=='vision' else tlayer(h, c0)-h h1 = ilayer(h, c1)-h if c1_type=='vision' else tlayer(h, c1)-h h = h0*mixed_ratio + h1*(1-mixed_ratio) + h # h = ilayer(h, c0) elif xtype=='image': h = ilayer(h) elif xtype == 'text': h = tlayer(h) else: raise ValueError return h