# Copyright 2023 The HuggingFace Team. All rights reserved. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. import math from dataclasses import dataclass from typing import Any, Dict, List, Optional, Tuple, Union import datetime import torch import torch.utils.checkpoint from torch import nn from torch.nn import functional as F from torch.nn.modules.normalization import GroupNorm import os from diffusers.configuration_utils import ConfigMixin, register_to_config from diffusers.models.attention_processor import AttentionProcessor from diffusers.utils import USE_PEFT_BACKEND from diffusers.models.autoencoders import AutoencoderKL from diffusers.models.lora import LoRACompatibleConv from diffusers.models.modeling_utils import ModelMixin from diffusers.models.unets.unet_2d_blocks import ( CrossAttnDownBlock2D, CrossAttnUpBlock2D, DownBlock2D, Downsample2D, ResnetBlock2D, Transformer2DModel, UpBlock2D, Upsample2D, ) from diffusers.models.unets.unet_2d_condition import UNet2DConditionModel from diffusers.utils import BaseOutput, logging import numpy as np from PIL import Image from safetensors import safe_open from .attention_autoencoder import AttentionAutoencoder, PositionalEncoding logger = logging.get_logger(__name__) # pylint: disable=invalid-name @dataclass class ControlNetXSOutput(BaseOutput): """ The output of [`ControlNetXSModel`]. Args: sample (`torch.FloatTensor` of shape `(batch_size, num_channels, height, width)`): The output of the `ControlNetXSModel`. Unlike `ControlNetOutput` this is NOT to be added to the base model output, but is already the final output. """ sample: torch.FloatTensor = None # copied from diffusers.models.controlnet.ControlNetConditioningEmbedding class ControlNetConditioningEmbedding(nn.Module): """ Quoting from https://arxiv.org/abs/2302.05543: "Stable Diffusion uses a pre-processing method similar to VQ-GAN [11] to convert the entire dataset of 512 × 512 images into smaller 64 × 64 “latent images” for stabilized training. This requires ControlNets to convert image-based conditions to 64 × 64 feature space to match the convolution size. We use a tiny network E(·) of four convolution layers with 4 × 4 kernels and 2 × 2 strides (activated by ReLU, channels are 16, 32, 64, 128, initialized with Gaussian weights, trained jointly with the full model) to encode image-space conditions ... into feature maps ..." """ def __init__( self, conditioning_embedding_channels: int, conditioning_channels: int = 3, block_out_channels: Tuple[int, ...] = (16, 32, 96, 256), ): super().__init__() self.conv_in = nn.Conv2d(conditioning_channels, block_out_channels[0], kernel_size=3, padding=1) self.blocks = nn.ModuleList([]) for i in range(len(block_out_channels) - 1): channel_in = block_out_channels[i] channel_out = block_out_channels[i + 1] self.blocks.append(nn.Conv2d(channel_in, channel_in, kernel_size=3, padding=1)) self.blocks.append(nn.Conv2d(channel_in, channel_out, kernel_size=3, padding=1, stride=2)) self.conv_out = zero_module( nn.Conv2d(block_out_channels[-1], conditioning_embedding_channels, kernel_size=3, padding=1) ) def forward(self, conditioning): embedding = self.conv_in(conditioning) embedding = F.silu(embedding) for block in self.blocks: embedding = block(embedding) embedding = F.silu(embedding) embedding = self.conv_out(embedding) return embedding class ControlNetConditioningEmbeddingBig(nn.Module): def __init__( self, conditioning_embedding_channels: int, conditioning_channels: int = 4, block_out_channels: Tuple[int, ...] = (16, 32, 96, 256), text_embed_dim: int = 768, ): super().__init__() self.conv_in = nn.Conv2d(conditioning_channels, block_out_channels[0], kernel_size=3, padding=1) self.cross_attention = CrossAttention(block_out_channels[0], text_embed_dim) # Encoder with increasing feature maps and more downsampling self.encoder = nn.ModuleList([ nn.Conv2d(block_out_channels[0], 64, kernel_size=3, stride=2, padding=1), nn.Conv2d(64, 128, kernel_size=3, stride=2, padding=1), nn.Conv2d(128, 256, kernel_size=3, stride=2, padding=1), nn.Conv2d(256, 320, kernel_size=3, stride=2, padding=1), nn.Conv2d(320, 512, kernel_size=3, stride=2, padding=1), nn.Conv2d(512, 640, kernel_size=3, stride=2, padding=1), ]) # Global embedding processing self.global_fc = nn.Linear(640, 640) # Bottleneck self.bottleneck_down = nn.Conv2d(640, 6, kernel_size=3, stride=1, padding=1) self.bottleneck_up = nn.Conv2d(6, 320, kernel_size=3, stride=1, padding=1) # Smaller decoder to get back to 320x64x64 self.decoder = nn.ModuleList([ nn.ConvTranspose2d(320, 320, kernel_size=4, stride=2, padding=1), # 4x4 -> 8x8 nn.ConvTranspose2d(320, 320, kernel_size=4, stride=2, padding=1), # 8x8 -> 16x16 nn.ConvTranspose2d(320, 320, kernel_size=4, stride=2, padding=1), # 16x16 -> 32x32 ]) def forward(self, x, text_embeds): x = self.conv_in(x) x = self.cross_attention(x, text_embeds) # Encoder for encoder_layer in self.encoder: x = encoder_layer(x) x = F.relu(x) # Global embedding processing b, c, h, w = x.shape x_flat = x.view(b, c, -1).mean(dim=2) # Global average pooling x_global = self.global_fc(x_flat).view(b, c, 1, 1) x = x + x_global.expand_as(x) # Add global features to local features # Bottleneck x = self.bottleneck_down(x) x = self.bottleneck_up(x) # Decoder for decoder_layer in self.decoder: x = decoder_layer(x) x = F.relu(x) #print(x.shape) return x class CrossAttention(nn.Module): def __init__(self, dim, context_dim): super().__init__() self.to_q = nn.Conv2d(dim, dim, 1) self.to_k = nn.Linear(context_dim, dim) self.to_v = nn.Linear(context_dim, dim) self.scale = dim ** -0.5 def forward(self, x, context): b, c, h, w = x.shape q = self.to_q(x).view(b, c, -1).permute(0, 2, 1) # (B, H*W, C) k = self.to_k(context) # (B, T, C) v = self.to_v(context) # (B, T, C) attn = torch.matmul(q, k.transpose(-2, -1)) * self.scale # (B, H*W, T) attn = attn.softmax(dim=-1) out = torch.matmul(attn, v) # (B, H*W, C) out = out.permute(0, 2, 1).view(b, c, h, w) # (B, C, H, W) return out + x def zero_module(module): for p in module.parameters(): nn.init.zeros_(p) return module class StyleCodesModel(ModelMixin, ConfigMixin): r""" Based off ControlNet-XS """ @classmethod def init_original(cls, base_model: UNet2DConditionModel, is_sdxl=True): """ Create a ControlNetXS model with the same parameters as in the original paper (https://github.com/vislearn/ControlNet-XS). Parameters: base_model (`UNet2DConditionModel`): Base UNet model. Needs to be either StableDiffusion or StableDiffusion-XL. is_sdxl (`bool`, defaults to `True`): Whether passed `base_model` is a StableDiffusion-XL model. """ def get_dim_attn_heads(base_model: UNet2DConditionModel, size_ratio: float, num_attn_heads: int): """ Currently, diffusers can only set the dimension of attention heads (see https://github.com/huggingface/diffusers/issues/2011#issuecomment-1547958131 for why). The original ControlNet-XS model, however, define the number of attention heads. That's why compute the dimensions needed to get the correct number of attention heads. """ block_out_channels = [int(size_ratio * c) for c in base_model.config.block_out_channels] dim_attn_heads = [math.ceil(c / num_attn_heads) for c in block_out_channels] return dim_attn_heads if is_sdxl: return StyleCodesModel.from_unet( base_model, time_embedding_mix=0.95, learn_embedding=True, size_ratio=0.1, conditioning_embedding_out_channels=(16, 32, 96, 256), num_attention_heads=get_dim_attn_heads(base_model, 0.1, 64), ) else: return StyleCodesModel.from_unet( base_model, time_embedding_mix=1.0, learn_embedding=True, size_ratio=0.0125, conditioning_embedding_out_channels=(16, 32, 96, 256), num_attention_heads=get_dim_attn_heads(base_model, 0.0125, 8), ) @classmethod def _gather_subblock_sizes(cls, unet: UNet2DConditionModel, base_or_control: str): """To create correctly sized connections between base and control model, we need to know the input and output channels of each subblock. Parameters: unet (`UNet2DConditionModel`): Unet of which the subblock channels sizes are to be gathered. base_or_control (`str`): Needs to be either "base" or "control". If "base", decoder is also considered. """ if base_or_control not in ["base", "control"]: raise ValueError("`base_or_control` needs to be either `base` or `control`") channel_sizes = {"down": [], "mid": [], "up": []} # input convolution channel_sizes["down"].append((unet.conv_in.in_channels, unet.conv_in.out_channels)) # encoder blocks for module in unet.down_blocks: if isinstance(module, (CrossAttnDownBlock2D, DownBlock2D)): for r in module.resnets: channel_sizes["down"].append((r.in_channels, r.out_channels)) if module.downsamplers: channel_sizes["down"].append( (module.downsamplers[0].channels, module.downsamplers[0].out_channels) ) else: raise ValueError(f"Encountered unknown module of type {type(module)} while creating ControlNet-XS.") # middle block channel_sizes["mid"].append((unet.mid_block.resnets[0].in_channels, unet.mid_block.resnets[0].out_channels)) # decoder blocks #if base_or_control == "base": for module in unet.up_blocks: if isinstance(module, (CrossAttnUpBlock2D, UpBlock2D)): for r in module.resnets: channel_sizes["up"].append((r.in_channels, r.out_channels)) else: raise ValueError( f"Encountered unknown module of type {type(module)} while creating ControlNet-XS." ) return channel_sizes def _make_colab_linear_layer(self, in_channels, out_channels): # Create a Linear layer where in_features = in_channels + out_channels #in_features = in_channels + out_channels linear_layer = nn.Linear(in_channels, out_channels) # Initialize weights as identity with torch.no_grad(): linear_layer.weight.copy_(torch.eye(in_channels)) return linear_layer @register_to_config def __init__( self, conditioning_channels: int = 3, conditioning_embedding_out_channels: Tuple[int] = (16, 32, 96, 256), controlnet_conditioning_channel_order: str = "rgb", time_embedding_input_dim: int = 320, time_embedding_dim: int = 1280, time_embedding_mix: float = 1.0, learn_embedding: bool = False, base_model_channel_sizes: Dict[str, List[Tuple[int]]] = { "down": [ (4, 320), (320, 320), (320, 320), (320, 320), (320, 640), (640, 640), (640, 640), (640, 1280), (1280, 1280), ], "mid": [(1280, 1280)], "up": [ (2560, 1280), (2560, 1280), (1920, 1280), (1920, 640), (1280, 640), (960, 640), (960, 320), (640, 320), (640, 320), ], }, sample_size: Optional[int] = None, down_block_types: Tuple[str] = ( "CrossAttnDownBlock2D", "CrossAttnDownBlock2D", "CrossAttnDownBlock2D", "DownBlock2D", ), up_block_types: Tuple[str] = ("UpBlock2D", "CrossAttnUpBlock2D", "CrossAttnUpBlock2D", "CrossAttnUpBlock2D"), block_out_channels: Tuple[int] = (320, 640, 1280, 1280), norm_num_groups: Optional[int] = 32, cross_attention_dim: Union[int, Tuple[int]] = 1280, transformer_layers_per_block: Union[int, Tuple[int], Tuple[Tuple]] = 1, num_attention_heads: Optional[Union[int, Tuple[int]]] = 8, upcast_attention: bool = False, ): super().__init__() # 1 - Create control unet self.control_model = UNet2DConditionModel( sample_size=sample_size, down_block_types=down_block_types, up_block_types=up_block_types, block_out_channels=block_out_channels, norm_num_groups=norm_num_groups, cross_attention_dim=cross_attention_dim, transformer_layers_per_block=transformer_layers_per_block, attention_head_dim=num_attention_heads, use_linear_projection=True, upcast_attention=upcast_attention, time_embedding_dim=time_embedding_dim, ) # 2 - Do model surgery on control model # 2.1 - Allow to use the same time information as the base model adjust_time_dims(self.control_model, time_embedding_input_dim, time_embedding_dim) # 2.2 - Allow for information infusion from base model # We concat the output of each base encoder subblocks to the input of the next control encoder subblock # (We ignore the 1st element, as it represents the `conv_in`.) extra_input_channels = [input_channels for input_channels, _ in base_model_channel_sizes["down"][1:]] it_extra_input_channels = iter(extra_input_channels) # print(extra_input_channels) # for b, block in enumerate(self.control_model.down_blocks): # for r in range(len(block.resnets)): # increase_block_input_in_encoder_resnet( # self.control_model, block_no=b, resnet_idx=r, by=next(it_extra_input_channels) # ) # if block.downsamplers: # increase_block_input_in_encoder_downsampler( # self.control_model, block_no=b, by=next(it_extra_input_channels) # ) # increase_block_input_in_mid_resnet(self.control_model, by=extra_input_channels[-1]) def get_flat_subblock_channel_sizes_down(model): subblock_channel_sizes = [] for block in model.down_blocks: # Iterate through ResnetBlock2D subblocks for resnet in block.resnets: # Only handle the first convolution for ResnetBlock2D if hasattr(resnet, 'conv1'): input_channels = resnet.conv1.in_channels output_channels = resnet.conv1.out_channels subblock_channel_sizes.append((input_channels, output_channels)) # Check and iterate through Upsample2D subblocks only if they exist if hasattr(block, 'upsamplers') and block.upsamplers: for upsampler in block.upsamplers: if hasattr(upsampler, 'conv'): input_channels = upsampler.conv.in_channels output_channels = upsampler.conv.out_channels subblock_channel_sizes.append((input_channels, output_channels)) print("down" ,subblock_channel_sizes) return subblock_channel_sizes def get_flat_subblock_channel_sizes(model): subblock_channel_sizes = [] for block in model.up_blocks: # Iterate through ResnetBlock2D subblocks for resnet in block.resnets: # Only handle the first convolution for ResnetBlock2D if hasattr(resnet, 'conv1'): input_channels = resnet.conv1.in_channels output_channels = resnet.conv1.out_channels subblock_channel_sizes.append((input_channels, output_channels)) # Check and iterate through Upsample2D subblocks only if they exist if hasattr(block, 'upsamplers') and block.upsamplers: for upsampler in block.upsamplers: if hasattr(upsampler, 'conv'): input_channels = upsampler.conv.in_channels output_channels = upsampler.conv.out_channels # subblock_channel_sizes.append((input_channels, output_channels)) print("up", subblock_channel_sizes) return subblock_channel_sizes get_flat_subblock_channel_sizes_down(self.control_model) # Now use this function to dynamically get the extra input channels #extra_input_channels_up = [t[1] for t in get_flat_subblock_channel_sizes(self.control_model)] #all_channels_up = get_flat_subblock_channel_sizes(self.control_model) #print(extra_input_channels_up) # it_extra_input_channels = iter(extra_input_channels_up) # #print(self.control_model.up_blocks) # for b, block in enumerate(self.control_model.up_blocks): # for r in range(len(block.resnets)): # increase_block_input_in_decoder_resnet( # self.control_model, block_no=b, resnet_idx=r, by=next(it_extra_input_channels) # ) # print(len(block.resnets)) # # if block.upsamplers: # #increase_block_input_in_decoder_downsampler( # # self.control_model, block_no=b, by=next(it_extra_input_channels) # #) # 2.3 - Make group norms work with modified channel sizes adjust_group_norms(self.control_model) # 3 - Gather Channel Sizes self.ch_inout_ctrl = StyleCodesModel._gather_subblock_sizes(self.control_model, base_or_control="control") self.ch_inout_base = base_model_channel_sizes # 4 - Build connections between base and control model self.control_model.down_zero_convs_in = nn.ModuleList([]) self.control_model.middle_block_out = nn.ModuleList([]) #self.control_model.middle_block_in = nn.ModuleList([]) self.control_model.up_zero_convs_out = nn.ModuleList([]) #self.control_model.up_zero_convs_in = nn.ModuleList([]) #for ch_io_base in self.ch_inout_base["down"]: # for i in range(len(self.ch_inout_base["down"])): # if i < len(self.ch_inout_ctrl["down"]) - 1: # ch_io_base = self.ch_inout_base["down"][i] # self.control_model.down_zero_convs_in.append(self._make_zero_conv(in_channels=ch_io_base[1], out_channels=ch_io_base[1])) #self.control_model.down_zero_convs_in.append(self._make_zero_conv(in_channels=ch_io_base[1], out_channels=ch_io_base[1])) linear_shape = self.ch_inout_ctrl["mid"][-1][1] + self.ch_inout_ctrl["mid"][-1][1] self.middle_block_out = self._make_colab_linear_layer(in_channels=linear_shape, out_channels=linear_shape) #self.up_zero_convs_out.append( # self._make_zero_conv(self.ch_inout_ctrl["down"][-1][1], self.ch_inout_base["mid"][-1][1]) #) #skip connections i dont care about these #for i in range(1, len(self.ch_inout_ctrl["down"])): # self.up_zero_convs_out.append( # self._make_zero_conv(self.ch_inout_ctrl["down"][-(i + 1)][1], self.ch_inout_base["up"][i - 1][1]) # ) #up blocks for output #need to check the input sizes #need to implement the increased input size for the up blocks as done already with the down blocks base_last_out_channels = [1280,1280, 1280, 1280, 1280, 1280, 1280, 640, 640, 640, 320, 320,320] base_current_in_channels = [1280, 1280, 1280, 1280, 1280, 1280, 640, 640, 640, 320, 320,320] #JANK WARNING REMEMBER TO FIX LATER BEFORE ACTUALLY PUTTING THIS CODE ANYWHERE print(f"subblock up sizes {self.ch_inout_ctrl}") # for i in range(len(base_current_in_channels)): # self.control_model.up_zero_convs_in.append( # self._make_zero_conv(base_last_out_channels[i], base_current_in_channels[i]) # ) for i in range(len(self.ch_inout_base["up"])): #for ch_io_base in self.ch_inout_base["up"]: ch_io_base = self.ch_inout_base["up"][i] if i < len(self.ch_inout_ctrl["up"]): linear_shape = ch_io_base[1] + ch_io_base[1] self.control_model.up_zero_convs_out.append( self._make_colab_linear_layer(in_channels=linear_shape, out_channels=linear_shape) ) # for i in range(len(self.ch_inout_ctrl["up"])): # self.control_model.up_zero_convs_out.append( # self._make_zero_conv(self.ch_inout_ctrl["up"][i][1], self.ch_inout_base["up"][i][1]) # ) # 5 - Create conditioning hint embedding # self.controlnet_cond_embedding = ControlNetConditioningEmbedding( # conditioning_embedding_channels=block_out_channels[0], # block_out_channels=conditioning_embedding_out_channels, # conditioning_channels=conditioning_channels, # ) self.sref_autoencoder = AttentionAutoencoder().to(device='cuda') # In the mininal implementation setting, we only need the control model up to the mid block #del self.control_model.up_blocks del self.control_model.down_blocks del self.control_model.conv_norm_out del self.control_model.conv_out del self.control_model.time_embedding del self.control_model.conv_in def load_model(self, path: str): """Load the model from the given path. Parameters: path (`str`): Path to the model checkpoint. """ if os.path.splitext(path)[-1] == ".safetensors": state_dict = {"image_proj": {}, "ip_adapter": {}, "controlnet": {}} with safe_open(path, framework="pt", device="cpu") as f: for key in f.keys(): if key.startswith("image_proj."): state_dict["image_proj"][key.replace("image_proj.", "")] = f.get_tensor(key) elif key.startswith("ip_adapter."): state_dict["ip_adapter"][key.replace("ip_adapter.", "")] = f.get_tensor(key) elif key.startswith("controlnet."): state_dict["controlnet"][key.replace("controlnet.", "")] = f.get_tensor(key) else: state_dict = torch.load(path, map_location="cpu") print("load controlnet", self.load_state_dict(state_dict["controlnet"],strict=False)) @classmethod def from_unet( cls, unet: UNet2DConditionModel, conditioning_channels: int = 3, conditioning_embedding_out_channels: Tuple[int] = (16, 32, 96, 256), controlnet_conditioning_channel_order: str = "rgb", learn_embedding: bool = False, time_embedding_mix: float = 1.0, block_out_channels: Optional[Tuple[int]] = None, size_ratio: Optional[float] = None, num_attention_heads: Optional[Union[int, Tuple[int]]] = 8, norm_num_groups: Optional[int] = None, ): r""" Instantiate a [`ControlNetXSModel`] from [`UNet2DConditionModel`]. Parameters: unet (`UNet2DConditionModel`): The UNet model we want to control. The dimensions of the ControlNetXSModel will be adapted to it. conditioning_channels (`int`, defaults to 3): Number of channels of conditioning input (e.g. an image) conditioning_embedding_out_channels (`tuple[int]`, defaults to `(16, 32, 96, 256)`): The tuple of output channel for each block in the `controlnet_cond_embedding` layer. controlnet_conditioning_channel_order (`str`, defaults to `"rgb"`): The channel order of conditional image. Will convert to `rgb` if it's `bgr`. learn_embedding (`bool`, defaults to `False`): Wether to use time embedding of the control model. If yes, the time embedding is a linear interpolation of the time embeddings of the control and base model with interpolation parameter `time_embedding_mix**3`. time_embedding_mix (`float`, defaults to 1.0): Linear interpolation parameter used if `learn_embedding` is `True`. block_out_channels (`Tuple[int]`, *optional*): Down blocks output channels in control model. Either this or `size_ratio` must be given. size_ratio (float, *optional*): When given, block_out_channels is set to a relative fraction of the base model's block_out_channels. Either this or `block_out_channels` must be given. num_attention_heads (`Union[int, Tuple[int]]`, *optional*): The dimension of the attention heads. The naming seems a bit confusing and it is, see https://github.com/huggingface/diffusers/issues/2011#issuecomment-1547958131 for why. norm_num_groups (int, *optional*, defaults to `None`): The number of groups to use for the normalization of the control unet. If `None`, `int(unet.config.norm_num_groups * size_ratio)` is taken. """ # Check input fixed_size = block_out_channels is not None relative_size = size_ratio is not None if not (fixed_size ^ relative_size): raise ValueError( "Pass exactly one of `block_out_channels` (for absolute sizing) or `control_model_ratio` (for relative sizing)." ) # Create model if block_out_channels is None: block_out_channels = [int(size_ratio * c) for c in unet.config.block_out_channels] # Check that attention heads and group norms match channel sizes # - attention heads def attn_heads_match_channel_sizes(attn_heads, channel_sizes): if isinstance(attn_heads, (tuple, list)): return all(c % a == 0 for a, c in zip(attn_heads, channel_sizes)) else: return all(c % attn_heads == 0 for c in channel_sizes) num_attention_heads = num_attention_heads or unet.config.attention_head_dim if not attn_heads_match_channel_sizes(num_attention_heads, block_out_channels): raise ValueError( f"The dimension of attention heads ({num_attention_heads}) must divide `block_out_channels` ({block_out_channels}). If you didn't set `num_attention_heads` the default settings don't match your model. Set `num_attention_heads` manually." ) # - group norms def group_norms_match_channel_sizes(num_groups, channel_sizes): return all(c % num_groups == 0 for c in channel_sizes) if norm_num_groups is None: if group_norms_match_channel_sizes(unet.config.norm_num_groups, block_out_channels): norm_num_groups = unet.config.norm_num_groups else: norm_num_groups = min(block_out_channels) if group_norms_match_channel_sizes(norm_num_groups, block_out_channels): print( f"`norm_num_groups` was set to `min(block_out_channels)` (={norm_num_groups}) so it divides all block_out_channels` ({block_out_channels}). Set it explicitly to remove this information." ) else: raise ValueError( f"`block_out_channels` ({block_out_channels}) don't match the base models `norm_num_groups` ({unet.config.norm_num_groups}). Setting `norm_num_groups` to `min(block_out_channels)` ({norm_num_groups}) didn't fix this. Pass `norm_num_groups` explicitly so it divides all block_out_channels." ) def get_time_emb_input_dim(unet: UNet2DConditionModel): return unet.time_embedding.linear_1.in_features def get_time_emb_dim(unet: UNet2DConditionModel): return unet.time_embedding.linear_2.out_features # Clone params from base unet if # (i) it's required to build SD or SDXL, and # (ii) it's not used for the time embedding (as time embedding of control model is never used), and # (iii) it's not set further below anyway to_keep = [ "cross_attention_dim", "down_block_types", "sample_size", "transformer_layers_per_block", "up_block_types", "upcast_attention", ] kwargs = {k: v for k, v in dict(unet.config).items() if k in to_keep} kwargs.update(block_out_channels=block_out_channels) kwargs.update(num_attention_heads=num_attention_heads) kwargs.update(norm_num_groups=norm_num_groups) # Add controlnetxs-specific params kwargs.update( conditioning_channels=conditioning_channels, controlnet_conditioning_channel_order=controlnet_conditioning_channel_order, time_embedding_input_dim=get_time_emb_input_dim(unet), time_embedding_dim=get_time_emb_dim(unet), time_embedding_mix=time_embedding_mix, learn_embedding=learn_embedding, base_model_channel_sizes=StyleCodesModel._gather_subblock_sizes(unet, base_or_control="base"), conditioning_embedding_out_channels=conditioning_embedding_out_channels, ) return cls(**kwargs) @property def attn_processors(self) -> Dict[str, AttentionProcessor]: r""" Returns: `dict` of attention processors: A dictionary containing all attention processors used in the model with indexed by its weight name. """ return self.control_model.attn_processors def set_attn_processor(self, processor: Union[AttentionProcessor, Dict[str, AttentionProcessor]]): r""" Sets the attention processor to use to compute attention. Parameters: processor (`dict` of `AttentionProcessor` or only `AttentionProcessor`): The instantiated processor class or a dictionary of processor classes that will be set as the processor for **all** `Attention` layers. If `processor` is a dict, the key needs to define the path to the corresponding cross attention processor. This is strongly recommended when setting trainable attention processors. """ self.control_model.set_attn_processor(processor) def set_default_attn_processor(self): """ Disables custom attention processors and sets the default attention implementation. """ self.control_model.set_default_attn_processor() def set_attention_slice(self, slice_size): r""" Enable sliced attention computation. When this option is enabled, the attention module splits the input tensor in slices to compute attention in several steps. This is useful for saving some memory in exchange for a small decrease in speed. Args: slice_size (`str` or `int` or `list(int)`, *optional*, defaults to `"auto"`): When `"auto"`, input to the attention heads is halved, so attention is computed in two steps. If `"max"`, maximum amount of memory is saved by running only one slice at a time. If a number is provided, uses as many slices as `attention_head_dim // slice_size`. In this case, `attention_head_dim` must be a multiple of `slice_size`. """ self.control_model.set_attention_slice(slice_size) def _set_gradient_checkpointing(self, module, value=False): if isinstance(module, (UNet2DConditionModel)): if value: module.enable_gradient_checkpointing() else: module.disable_gradient_checkpointing() def forward( self, base_model: UNet2DConditionModel, sample: torch.FloatTensor, timestep: Union[torch.Tensor, float, int], encoder_hidden_states: torch.Tensor, encoder_hidden_states_controlnet: torch.Tensor, controlnet_cond: torch.Tensor, conditioning_scale: float = 1.0, class_labels: Optional[torch.Tensor] = None, timestep_cond: Optional[torch.Tensor] = None, attention_mask: Optional[torch.Tensor] = None, cross_attention_kwargs: Optional[Dict[str, Any]] = None, added_cond_kwargs: Optional[Dict[str, torch.Tensor]] = None, return_dict: bool = True, stylecode=None, ) -> Union[ControlNetXSOutput, Tuple]: """ The [`ControlNetModel`] forward method. Args: base_model (`UNet2DConditionModel`): The base unet model we want to control. sample (`torch.FloatTensor`): The noisy input tensor. timestep (`Union[torch.Tensor, float, int]`): The number of timesteps to denoise an input. encoder_hidden_states (`torch.Tensor`): The encoder hidden states. controlnet_cond (`torch.FloatTensor`): The conditional input tensor of shape `(batch_size, sequence_length, hidden_size)`. conditioning_scale (`float`, defaults to `1.0`): How much the control model affects the base model outputs. class_labels (`torch.Tensor`, *optional*, defaults to `None`): Optional class labels for conditioning. Their embeddings will be summed with the timestep embeddings. timestep_cond (`torch.Tensor`, *optional*, defaults to `None`): Additional conditional embeddings for timestep. If provided, the embeddings will be summed with the timestep_embedding passed through the `self.time_embedding` layer to obtain the final timestep embeddings. attention_mask (`torch.Tensor`, *optional*, defaults to `None`): An attention mask of shape `(batch, key_tokens)` is applied to `encoder_hidden_states`. If `1` the mask is kept, otherwise if `0` it is discarded. Mask will be converted into a bias, which adds large negative values to the attention scores corresponding to "discard" tokens. added_cond_kwargs (`dict`): Additional conditions for the Stable Diffusion XL UNet. cross_attention_kwargs (`dict[str]`, *optional*, defaults to `None`): A kwargs dictionary that if specified is passed along to the `AttnProcessor`. return_dict (`bool`, defaults to `True`): Whether or not to return a [`~models.controlnet.ControlNetOutput`] instead of a plain tuple. Returns: [`~models.controlnetxs.ControlNetXSOutput`] **or** `tuple`: If `return_dict` is `True`, a [`~models.controlnetxs.ControlNetXSOutput`] is returned, otherwise a tuple is returned where the first element is the sample tensor. """ # check channel order channel_order = self.config.controlnet_conditioning_channel_order if channel_order == "rgb": # in rgb order by default ... elif channel_order == "bgr": controlnet_cond = torch.flip(controlnet_cond, dims=[1]) else: raise ValueError(f"unknown `controlnet_conditioning_channel_order`: {channel_order}") # scale control strength n_connections = 0 + 1 + len(self.control_model.up_zero_convs_out) scale_list = torch.full((n_connections,), conditioning_scale) # prepare attention_mask if attention_mask is not None: attention_mask = (1 - attention_mask.to(sample.dtype)) * -10000.0 attention_mask = attention_mask.unsqueeze(1) # 1. time timesteps = timestep if not torch.is_tensor(timesteps): # TODO: this requires sync between CPU and GPU. So try to pass timesteps as tensors if you can # This would be a good case for the `match` statement (Python 3.10+) is_mps = sample.device.type == "mps" if isinstance(timestep, float): dtype = torch.float32 if is_mps else torch.float64 else: dtype = torch.int32 if is_mps else torch.int64 timesteps = torch.tensor([timesteps], dtype=dtype, device=sample.device) elif len(timesteps.shape) == 0: timesteps = timesteps[None].to(sample.device) # broadcast to batch dimension in a way that's compatible with ONNX/Core ML timesteps = timesteps.expand(sample.shape[0]) t_emb = base_model.time_proj(timesteps) # timesteps does not contain any weights and will always return f32 tensors # but time_embedding might actually be running in fp16. so we need to cast here. # there might be better ways to encapsulate this. t_emb = t_emb.to(dtype=sample.dtype) if self.config.learn_embedding: ctrl_temb = self.control_model.time_embedding(t_emb, timestep_cond) base_temb = base_model.time_embedding(t_emb, timestep_cond) interpolation_param = self.config.time_embedding_mix**0.3 temb = ctrl_temb * interpolation_param + base_temb * (1 - interpolation_param) else: temb = base_model.time_embedding(t_emb) # added time & text embeddings aug_emb = None aug_emb_ctrl = None if base_model.class_embedding is not None: if class_labels is None: raise ValueError("class_labels should be provided when num_class_embeds > 0") if base_model.config.class_embed_type == "timestep": class_labels = base_model.time_proj(class_labels) class_emb = base_model.class_embedding(class_labels).to(dtype=self.dtype) temb = temb + class_emb if base_model.config.addition_embed_type is not None: if base_model.config.addition_embed_type == "text": aug_emb = base_model.add_embedding(encoder_hidden_states) aug_emb_ctrl = base_model.add_embedding(encoder_hidden_states_controlnet) elif base_model.config.addition_embed_type == "text_image": raise NotImplementedError() elif base_model.config.addition_embed_type == "text_time": # SDXL - style if "text_embeds" not in added_cond_kwargs: raise ValueError( f"{self.__class__} has the config param `addition_embed_type` set to 'text_time' which requires the keyword argument `text_embeds` to be passed in `added_cond_kwargs`" ) text_embeds = added_cond_kwargs.get("text_embeds") if "time_ids" not in added_cond_kwargs: raise ValueError( f"{self.__class__} has the config param `addition_embed_type` set to 'text_time' which requires the keyword argument `time_ids` to be passed in `added_cond_kwargs`" ) time_ids = added_cond_kwargs.get("time_ids") time_embeds = base_model.add_time_proj(time_ids.flatten()) time_embeds = time_embeds.reshape((text_embeds.shape[0], -1)) add_embeds = torch.concat([text_embeds, time_embeds], dim=-1) add_embeds = add_embeds.to(temb.dtype) aug_emb = base_model.add_embedding(add_embeds) elif base_model.config.addition_embed_type == "image": raise NotImplementedError() elif base_model.config.addition_embed_type == "image_hint": raise NotImplementedError() temb = temb + aug_emb if aug_emb is not None else temb #temb_ctrl = torch.zeros_like(temb) temb_ctrl = temb + aug_emb_ctrl if aug_emb_ctrl is not None else temb # text embeddings #note when i have more time actually skip the cross attention layers cemb = encoder_hidden_states #cemb_ctrl = torch.zeros_like(encoder_hidden_states) cemb_ctrl = encoder_hidden_states # Preparation #print("1:cond, 2: embeddings",controlnet_cond.shape,encoder_hidden_states_controlnet.shape) #save_debug_image(controlnet_cond[0]) #guided_hint = self.controlnet_cond_embedding(controlnet_cond) #guided_hint=None h_ctrl = h_base = sample hs_base, hs_ctrl = [], [] it_up_convs_out = iter (self.control_model.up_zero_convs_out) scales = iter(scale_list) base_down_subblocks = self.to_sub_blocks(base_model.down_blocks) #ctrl_down_subblocks = self.to_sub_blocks(self.control_model.down_blocks) base_mid_subblocks = self.to_sub_blocks([base_model.mid_block]) ctrl_mid_subblocks = self.to_sub_blocks([self.control_model.mid_block]) base_up_subblocks = self.to_sub_blocks(base_model.up_blocks) ctrl_up_subblocks = self.to_sub_blocks(self.control_model.up_blocks) # Cross Control # 0 - conv in h_base = base_model.conv_in(h_base) #h_ctrl = self.control_model.conv_in(h_ctrl) #if guided_hint is not None: h_ctrl = controlnet_cond # h_base = h_base + next(it_down_convs_out)(h_ctrl) * next(scales) # D - add ctrl -> base hs_base.append(h_base) #hs_ctrl.append(h_ctrl) # 1 - down for m_base in base_down_subblocks: #h_ctrl = torch.cat([h_ctrl, next(it_down_convs_in)(h_base)], dim=1) # A - concat base -> ctrl h_base = m_base(h_base, temb, cemb, attention_mask, cross_attention_kwargs) # B - apply base subblock #h_ctrl = m_ctrl(h_ctrl, temb_ctrl, cemb_ctrl, attention_mask, cross_attention_kwargs) # C - apply ctrl subblock #h_base = h_base + next(it_down_convs_out)(h_ctrl) * next(scales) # D - add ctrl -> base hs_base.append(h_base) #hs_ctrl.append(h_ctrl) if stylecode is None: h_ctrl,encoded_strings = self.sref_autoencoder.forward_encoding(h_ctrl,h_base.shape[2],h_base.shape[3]) else: h_ctrl = self.sref_autoencoder.forward_from_stylecode(stylecode,h_base.shape[2],h_base.shape[3],h_base.dtype, h_base.device) # 2 - mid #h_ctrl = torch.cat([h_ctrl, next(it_down_convs_in)(h_base)], dim=1) # A - concat base -> ctrl for m_base, m_ctrl in zip(base_mid_subblocks, ctrl_mid_subblocks): h_base = m_base(h_base, temb, cemb, attention_mask, cross_attention_kwargs) # B - apply base subblock h_ctrl = m_ctrl(h_ctrl, temb_ctrl, cemb_ctrl, attention_mask, cross_attention_kwargs) # C - apply ctrl subblock #taken from https://github.com/dvlab-research/ControlNeXt/blob/main/ControlNeXt-SD1.5/models/unet.py #mid_block_additional_residual = self.middle_block_out(h_ctrl) # mid_block_additional_residual = mid_block_out # mid_block_additional_residual=nn.functional.adaptive_avg_pool2d(mid_block_additional_residual, h_base.shape[-2:]) # mid_block_additional_residual = mid_block_additional_residual.to(h_base) # mean_latents, std_latents = torch.mean(h_base, dim=(1, 2, 3), keepdim=True), torch.std(h_base, dim=(1, 2, 3), keepdim=True) # mean_control, std_control = torch.mean(mid_block_additional_residual, dim=(1, 2, 3), keepdim=True), torch.std(mid_block_additional_residual, dim=(1, 2, 3), keepdim=True) # mid_block_additional_residual = (mid_block_additional_residual - mean_control) * (std_latents / (std_control + 1e-12)) + mean_latents # h_base = h_base + mid_block_additional_residual * next(scales) batch_size, channels, height, width = h_ctrl.shape colab_input = torch.cat([h_ctrl, h_base], dim=1).view(batch_size, channels * 2, height * width).permute(0, 2, 1) colab_output = self.middle_block_out(colab_input) sequence_len = height * width colab_output = colab_output.permute(0, 2, 1).view(batch_size, channels * 2, height, width) # Reshape back h_ctrl, h_base_output = torch.chunk(colab_output, 2, dim=1) #mix using cond scale h_base = h_base * (1 - conditioning_scale) + h_base_output * conditioning_scale #h_base = h_base + mid_block_additional_residual * next(scales) # D - add ctrl -> base # 3 - up for m_base,m_ctrl in zip(base_up_subblocks,ctrl_up_subblocks): hs_base_new = hs_base.pop() h_base_with_skip = torch.cat([h_base, hs_base_new], dim=1) # concat info from base encoder+ctrl encoder empty = torch.zeros_like(hs_base_new) h_ctrl = torch.cat([h_ctrl, empty], dim=1) # concat info from ctrl encoder + skip connections h_ctrl = m_ctrl(h_ctrl, temb_ctrl, cemb_ctrl, attention_mask, cross_attention_kwargs) # C - apply ctrl subblock h_base = m_base(h_base_with_skip, temb, cemb, attention_mask, cross_attention_kwargs) batch_size, channels, height, width = h_ctrl.shape colab_input = torch.cat([h_ctrl, h_base], dim=1).view(batch_size, channels * 2, height * width).permute(0, 2, 1) colab_output = next(it_up_convs_out)(colab_input) colab_output = colab_output.permute(0, 2, 1).view(batch_size, channels * 2, height, width) h_ctrl, h_base_output = torch.chunk(colab_output, 2, dim=1) h_base = h_base * (1 - conditioning_scale) + h_base_output * conditioning_scale #hn_ctrl = next(it_up_convs_out)(h_ctrl) #print(hn_ctrl) #h_base = h_base + hn_ctrl * next(scales) # D - add ctrl -> base h_base = base_model.conv_norm_out(h_base) h_base = base_model.conv_act(h_base) h_base = base_model.conv_out(h_base) if not return_dict: return h_base return ControlNetXSOutput(sample=h_base) #needs new stuff to work correctly # def pre_process( # self, # base_model: UNet2DConditionModel, # sample: torch.FloatTensor, # timestep: Union[torch.Tensor, float, int], # encoder_hidden_states: torch.Tensor, # controlnet_cond: torch.Tensor, # conditioning_scale: float = 1.0, # class_labels: Optional[torch.Tensor] = None, # timestep_cond: Optional[torch.Tensor] = None, # attention_mask: Optional[torch.Tensor] = None, # cross_attention_kwargs: Optional[Dict[str, Any]] = None, # added_cond_kwargs: Optional[Dict[str, torch.Tensor]] = None, # return_dict: bool = True # ): # """ # The [`ControlNetModel`] forward method. # Args: # base_model (`UNet2DConditionModel`): # The base unet model we want to control. # sample (`torch.FloatTensor`): # The noisy input tensor. # timestep (`Union[torch.Tensor, float, int]`): # The number of timesteps to denoise an input. # encoder_hidden_states (`torch.Tensor`): # The encoder hidden states. # controlnet_cond (`torch.FloatTensor`): # The conditional input tensor of shape `(batch_size, sequence_length, hidden_size)`. # conditioning_scale (`float`, defaults to `1.0`): # How much the control model affects the base model outputs. # class_labels (`torch.Tensor`, *optional*, defaults to `None`): # Optional class labels for conditioning. Their embeddings will be summed with the timestep embeddings. # timestep_cond (`torch.Tensor`, *optional*, defaults to `None`): # Additional conditional embeddings for timestep. If provided, the embeddings will be summed with the # timestep_embedding passed through the `self.time_embedding` layer to obtain the final timestep # embeddings. # attention_mask (`torch.Tensor`, *optional*, defaults to `None`): # An attention mask of shape `(batch, key_tokens)` is applied to `encoder_hidden_states`. If `1` the mask # is kept, otherwise if `0` it is discarded. Mask will be converted into a bias, which adds large # negative values to the attention scores corresponding to "discard" tokens. # added_cond_kwargs (`dict`): # Additional conditions for the Stable Diffusion XL UNet. # cross_attention_kwargs (`dict[str]`, *optional*, defaults to `None`): # A kwargs dictionary that if specified is passed along to the `AttnProcessor`. # return_dict (`bool`, defaults to `True`): # Whether or not to return a [`~models.controlnet.ControlNetOutput`] instead of a plain tuple. # Returns: # [`~models.controlnetxs.ControlNetXSOutput`] **or** `tuple`: # If `return_dict` is `True`, a [`~models.controlnetxs.ControlNetXSOutput`] is returned, otherwise a # tuple is returned where the first element is the sample tensor. # """ # # check channel order # channel_order = self.config.controlnet_conditioning_channel_order # if channel_order == "rgb": # # in rgb order by default # ... # elif channel_order == "bgr": # controlnet_cond = torch.flip(controlnet_cond, dims=[1]) # else: # raise ValueError(f"unknown `controlnet_conditioning_channel_order`: {channel_order}") # # scale control strength # n_connections = len(self.control_model.down_zero_convs_out) + 1 + len(self.control_model.up_zero_convs_out) # scale_list = torch.full((n_connections,), conditioning_scale) # # prepare attention_mask # if attention_mask is not None: # attention_mask = (1 - attention_mask.to(sample.dtype)) * -10000.0 # attention_mask = attention_mask.unsqueeze(1) # # 1. time # timesteps = timestep # if not torch.is_tensor(timesteps): # # TODO: this requires sync between CPU and GPU. So try to pass timesteps as tensors if you can # # This would be a good case for the `match` statement (Python 3.10+) # is_mps = sample.device.type == "mps" # if isinstance(timestep, float): # dtype = torch.float32 if is_mps else torch.float64 # else: # dtype = torch.int32 if is_mps else torch.int64 # timesteps = torch.tensor([timesteps], dtype=dtype, device=sample.device) # elif len(timesteps.shape) == 0: # timesteps = timesteps[None].to(sample.device) # # broadcast to batch dimension in a way that's compatible with ONNX/Core ML # timesteps = timesteps.expand(sample.shape[0]) # t_emb = base_model.time_proj(timesteps) # # timesteps does not contain any weights and will always return f32 tensors # # but time_embedding might actually be running in fp16. so we need to cast here. # # there might be better ways to encapsulate this. # t_emb = t_emb.to(dtype=sample.dtype) # if self.config.learn_embedding: # ctrl_temb = self.control_model.time_embedding(t_emb, timestep_cond) # base_temb = base_model.time_embedding(t_emb, timestep_cond) # interpolation_param = self.config.time_embedding_mix**0.3 # temb = ctrl_temb * interpolation_param + base_temb * (1 - interpolation_param) # else: # temb = base_model.time_embedding(t_emb) # # added time & text embeddings # aug_emb = None # if base_model.class_embedding is not None: # if class_labels is None: # raise ValueError("class_labels should be provided when num_class_embeds > 0") # if base_model.config.class_embed_type == "timestep": # class_labels = base_model.time_proj(class_labels) # class_emb = base_model.class_embedding(class_labels).to(dtype=self.dtype) # temb = temb + class_emb # if base_model.config.addition_embed_type is not None: # if base_model.config.addition_embed_type == "text": # aug_emb = base_model.add_embedding(encoder_hidden_states) # elif base_model.config.addition_embed_type == "text_image": # raise NotImplementedError() # elif base_model.config.addition_embed_type == "text_time": # # SDXL - style # if "text_embeds" not in added_cond_kwargs: # raise ValueError( # f"{self.__class__} has the config param `addition_embed_type` set to 'text_time' which requires the keyword argument `text_embeds` to be passed in `added_cond_kwargs`" # ) # text_embeds = added_cond_kwargs.get("text_embeds") # if "time_ids" not in added_cond_kwargs: # raise ValueError( # f"{self.__class__} has the config param `addition_embed_type` set to 'text_time' which requires the keyword argument `time_ids` to be passed in `added_cond_kwargs`" # ) # time_ids = added_cond_kwargs.get("time_ids") # time_embeds = base_model.add_time_proj(time_ids.flatten()) # time_embeds = time_embeds.reshape((text_embeds.shape[0], -1)) # add_embeds = torch.concat([text_embeds, time_embeds], dim=-1) # add_embeds = add_embeds.to(temb.dtype) # aug_emb = base_model.add_embedding(add_embeds) # elif base_model.config.addition_embed_type == "image": # raise NotImplementedError() # elif base_model.config.addition_embed_type == "image_hint": # raise NotImplementedError() # temb = temb + aug_emb if aug_emb is not None else temb # # text embeddings # cemb = encoder_hidden_states # # Preparation # guided_hint = self.controlnet_cond_embedding(controlnet_cond) # #guided_hint=None # # h_ctrl = h_base = sample # # hs_base, hs_ctrl = [], [] # # it_down_convs_in, it_down_convs_out, it_up_convs_in, it_up_convs_out = map( # # iter, (self.control_model.down_zero_convs_in, self.control_model.down_zero_convs_out, self.control_model.up_zero_convs_in, self.control_model.up_zero_convs_out) # # ) # scales = iter(scale_list) # return temb,cemb,scales,guided_hint def _make_zero_conv(self, in_channels, out_channels=None): # keep running track of channels sizes #self.in_channels = in_channels #self.out_channels = out_channels or in_channels # return zero_module(nn.Conv2d(in_channels, out_channels, 1, padding=0)) def _make_identity_conv(self, in_channels, out_channels=None): #out_channels = out_channels or in_channels return nn.Conv2d(in_channels, out_channels, kernel_size=1, padding=0, bias=False) @torch.no_grad() def _check_if_vae_compatible(self, vae: AutoencoderKL): condition_downscale_factor = 2 ** (len(self.config.conditioning_embedding_out_channels) - 1) vae_downscale_factor = 2 ** (len(vae.config.block_out_channels) - 1) compatible = condition_downscale_factor == vae_downscale_factor return compatible, condition_downscale_factor, vae_downscale_factor def to_sub_blocks(self,blocks): if not is_iterable(blocks): blocks = [blocks] sub_blocks = [] for b in blocks: if hasattr(b, "resnets"): if hasattr(b, "attentions") and b.attentions is not None: for r, a in zip(b.resnets, b.attentions): sub_blocks.append([r, a]) num_resnets = len(b.resnets) num_attns = len(b.attentions) if num_resnets > num_attns: # we can have more resnets than attentions, so add each resnet as separate subblock for i in range(num_attns, num_resnets): sub_blocks.append([b.resnets[i]]) else: for r in b.resnets: sub_blocks.append([r]) # upsamplers are part of the same subblock if hasattr(b, "upsamplers") and b.upsamplers is not None: for u in b.upsamplers: sub_blocks[-1].extend([u]) # downsamplers are own subblock if hasattr(b, "downsamplers") and b.downsamplers is not None: for d in b.downsamplers: sub_blocks.append([d]) return list(map(SubBlock, sub_blocks)) class SubBlock(nn.ModuleList): """A SubBlock is the largest piece of either base or control model, that is executed independently of the other model respectively. Before each subblock, information is concatted from base to control. And after each subblock, information is added from control to base. """ def __init__(self, ms, *args, **kwargs): if not is_iterable(ms): ms = [ms] super().__init__(ms, *args, **kwargs) def forward( self, x: torch.Tensor, temb: torch.Tensor, cemb: torch.Tensor, attention_mask: Optional[torch.Tensor] = None, cross_attention_kwargs: Optional[Dict[str, Any]] = None, ): """Iterate through children and pass correct information to each.""" for m in self: if isinstance(m, ResnetBlock2D): x = m(x, temb) elif isinstance(m, Transformer2DModel): x = m(x, cemb, attention_mask=attention_mask, cross_attention_kwargs=cross_attention_kwargs).sample elif isinstance(m, Downsample2D): x = m(x) elif isinstance(m, Upsample2D): x = m(x) else: raise ValueError( f"Type of m is {type(m)} but should be `ResnetBlock2D`, `Transformer2DModel`, `Downsample2D` or `Upsample2D`" ) return x def adjust_time_dims(unet: UNet2DConditionModel, in_dim: int, out_dim: int): unet.time_embedding.linear_1 = nn.Linear(in_dim, out_dim) def increase_block_input_in_encoder_resnet(unet: UNet2DConditionModel, block_no, resnet_idx, by): """Increase channels sizes to allow for additional concatted information from base model""" r = unet.down_blocks[block_no].resnets[resnet_idx] old_norm1, old_conv1 = r.norm1, r.conv1 # norm norm_args = "num_groups num_channels eps affine".split(" ") for a in norm_args: assert hasattr(old_norm1, a) norm_kwargs = {a: getattr(old_norm1, a) for a in norm_args} norm_kwargs["num_channels"] += by # surgery done here # conv1 conv1_args = [ "in_channels", "out_channels", "kernel_size", "stride", "padding", "dilation", "groups", "bias", "padding_mode", ] #if not USE_PEFT_BACKEND: # conv1_args.append("lora_layer") for a in conv1_args: assert hasattr(old_conv1, a) conv1_kwargs = {a: getattr(old_conv1, a) for a in conv1_args} conv1_kwargs["bias"] = "bias" in conv1_kwargs # as param, bias is a boolean, but as attr, it's a tensor. conv1_kwargs["in_channels"] += by # surgery done here # conv_shortcut # as we changed the input size of the block, the input and output sizes are likely different, # therefore we need a conv_shortcut (simply adding won't work) conv_shortcut_args_kwargs = { "in_channels": conv1_kwargs["in_channels"], "out_channels": conv1_kwargs["out_channels"], # default arguments from resnet.__init__ "kernel_size": 1, "stride": 1, "padding": 0, "bias": True, } # swap old with new modules unet.down_blocks[block_no].resnets[resnet_idx].norm1 = GroupNorm(**norm_kwargs) unet.down_blocks[block_no].resnets[resnet_idx].conv1 = ( nn.Conv2d(**conv1_kwargs) if USE_PEFT_BACKEND else LoRACompatibleConv(**conv1_kwargs) ) unet.down_blocks[block_no].resnets[resnet_idx].conv_shortcut = ( nn.Conv2d(**conv_shortcut_args_kwargs) if USE_PEFT_BACKEND else LoRACompatibleConv(**conv_shortcut_args_kwargs) ) print(f"increasing down {unet.down_blocks[block_no].resnets[resnet_idx].in_channels} by {by}") unet.down_blocks[block_no].resnets[resnet_idx].in_channels += by # surgery done here def increase_block_input_in_decoder_resnet(unet: UNet2DConditionModel, block_no, resnet_idx, by): """Increase channels sizes to allow for additional concatted information from base model""" r = unet.up_blocks[block_no].resnets[resnet_idx] old_norm1, old_conv1 = r.norm1, r.conv1 # norm norm_args = "num_groups num_channels eps affine".split(" ") for a in norm_args: assert hasattr(old_norm1, a) norm_kwargs = {a: getattr(old_norm1, a) for a in norm_args} norm_kwargs["num_channels"] += by # surgery done here # conv1 conv1_args = [ "in_channels", "out_channels", "kernel_size", "stride", "padding", "dilation", "groups", "bias", "padding_mode", ] #if not USE_PEFT_BACKEND: # conv1_args.append("lora_layer") for a in conv1_args: assert hasattr(old_conv1, a) conv1_kwargs = {a: getattr(old_conv1, a) for a in conv1_args} conv1_kwargs["bias"] = "bias" in conv1_kwargs # as param, bias is a boolean, but as attr, it's a tensor. conv1_kwargs["in_channels"] += by # surgery done here # conv_shortcut # as we changed the input size of the block, the input and output sizes are likely different, # therefore we need a conv_shortcut (simply adding won't work) conv_shortcut_args_kwargs = { "in_channels": conv1_kwargs["in_channels"], "out_channels": conv1_kwargs["out_channels"], # default arguments from resnet.__init__ "kernel_size": 1, "stride": 1, "padding": 0, "bias": True, } # swap old with new modules unet.up_blocks[block_no].resnets[resnet_idx].norm1 = GroupNorm(**norm_kwargs) unet.up_blocks[block_no].resnets[resnet_idx].conv1 = ( nn.Conv2d(**conv1_kwargs) if USE_PEFT_BACKEND else LoRACompatibleConv(**conv1_kwargs) ) unet.up_blocks[block_no].resnets[resnet_idx].conv_shortcut = ( nn.Conv2d(**conv_shortcut_args_kwargs) if USE_PEFT_BACKEND else LoRACompatibleConv(**conv_shortcut_args_kwargs) ) #by =unet.up_blocks[block_no].resnets[resnet_idx].out_channels print(f"increasing up {unet.up_blocks[block_no].resnets[resnet_idx].in_channels} by {by}") unet.up_blocks[block_no].resnets[resnet_idx].in_channels += by # surgery done here def increase_block_input_in_encoder_downsampler(unet: UNet2DConditionModel, block_no, by): """Increase channels sizes to allow for additional concatted information from base model""" old_down = unet.down_blocks[block_no].downsamplers[0].conv args = [ "in_channels", "out_channels", "kernel_size", "stride", "padding", "dilation", "groups", "bias", "padding_mode", ] #if not USE_PEFT_BACKEND: # args.append("lora_layer") for a in args: assert hasattr(old_down, a) kwargs = {a: getattr(old_down, a) for a in args} kwargs["bias"] = "bias" in kwargs # as param, bias is a boolean, but as attr, it's a tensor. kwargs["in_channels"] += by # surgery done here # swap old with new modules unet.down_blocks[block_no].downsamplers[0].conv = ( nn.Conv2d(**kwargs) if USE_PEFT_BACKEND else LoRACompatibleConv(**kwargs) ) unet.down_blocks[block_no].downsamplers[0].channels += by # surgery done here def increase_block_input_in_decoder_downsampler(unet: UNet2DConditionModel, block_no, by): """Increase channels sizes to allow for additional concatted information from base model""" old_down = unet.up_blocks[block_no].upsamplers[0].conv args = [ "in_channels", "out_channels", "kernel_size", "stride", "padding", "dilation", "groups", "bias", "padding_mode", ] if not USE_PEFT_BACKEND: args.append("lora_layer") for a in args: assert hasattr(old_down, a) kwargs = {a: getattr(old_down, a) for a in args} kwargs["bias"] = "bias" in kwargs # as param, bias is a boolean, but as attr, it's a tensor. kwargs["in_channels"] += by # surgery done here # swap old with new modules unet.up_blocks[block_no].upsamplers[0].conv = ( nn.Conv2d(**kwargs) if USE_PEFT_BACKEND else LoRACompatibleConv(**kwargs) ) unet.up_blocks[block_no].upsamplers[0].channels += by # surgery done here def increase_block_input_in_mid_resnet(unet: UNet2DConditionModel, by): """Increase channels sizes to allow for additional concatted information from base model""" m = unet.mid_block.resnets[0] old_norm1, old_conv1 = m.norm1, m.conv1 # norm norm_args = "num_groups num_channels eps affine".split(" ") for a in norm_args: assert hasattr(old_norm1, a) norm_kwargs = {a: getattr(old_norm1, a) for a in norm_args} norm_kwargs["num_channels"] += by # surgery done here conv1_args = [ "in_channels", "out_channels", "kernel_size", "stride", "padding", "dilation", "groups", "bias", "padding_mode", ] #if not USE_PEFT_BACKEND: # conv1_args.append("lora_layer") conv1_kwargs = {a: getattr(old_conv1, a) for a in conv1_args} conv1_kwargs["bias"] = "bias" in conv1_kwargs # as param, bias is a boolean, but as attr, it's a tensor. conv1_kwargs["in_channels"] += by # surgery done here # conv_shortcut # as we changed the input size of the block, the input and output sizes are likely different, # therefore we need a conv_shortcut (simply adding won't work) conv_shortcut_args_kwargs = { "in_channels": conv1_kwargs["in_channels"], "out_channels": conv1_kwargs["out_channels"], # default arguments from resnet.__init__ "kernel_size": 1, "stride": 1, "padding": 0, "bias": True, } # swap old with new modules unet.mid_block.resnets[0].norm1 = GroupNorm(**norm_kwargs) unet.mid_block.resnets[0].conv1 = ( nn.Conv2d(**conv1_kwargs) if USE_PEFT_BACKEND else LoRACompatibleConv(**conv1_kwargs) ) unet.mid_block.resnets[0].conv_shortcut = ( nn.Conv2d(**conv_shortcut_args_kwargs) if USE_PEFT_BACKEND else LoRACompatibleConv(**conv_shortcut_args_kwargs) ) unet.mid_block.resnets[0].in_channels += by # surgery done here def adjust_group_norms(unet: UNet2DConditionModel, max_num_group: int = 32): def find_denominator(number, start): if start >= number: return number while start != 0: residual = number % start if residual == 0: return start start -= 1 for block in [*unet.down_blocks, unet.mid_block]: # resnets for r in block.resnets: if r.norm1.num_groups < max_num_group: r.norm1.num_groups = find_denominator(r.norm1.num_channels, start=max_num_group) if r.norm2.num_groups < max_num_group: r.norm2.num_groups = find_denominator(r.norm2.num_channels, start=max_num_group) # transformers if hasattr(block, "attentions"): for a in block.attentions: if a.norm.num_groups < max_num_group: a.norm.num_groups = find_denominator(a.norm.num_channels, start=max_num_group) def is_iterable(o): if isinstance(o, str): return False try: iter(o) return True except TypeError: return False def save_debug_image(image, folder='debug_images', noise_threshold=0.1): os.makedirs(folder, exist_ok=True) timestamp = datetime.datetime.now().strftime("%Y%m%d_%H%M%S_%f") filename = f"debug_image_{timestamp}.png" filepath = os.path.join(folder, filename) print("Debugging image information:") print(f"Type of image: {type(image)}") if isinstance(image, torch.Tensor): print(f"Image tensor shape: {image.shape}") print(f"Image tensor dtype: {image.dtype}") print(f"Image tensor device: {image.device}") print(f"Image tensor min: {image.min()}, max: {image.max()}") image_np = image.cpu().detach().numpy() elif isinstance(image, np.ndarray): image_np = image else: print(f"Unexpected image type: {type(image)}") return print(f"Numpy array shape: {image_np.shape}") print(f"Numpy array dtype: {image_np.dtype}") print(f"Numpy array min: {image_np.min()}, max: {image_np.max()}") # Handle different array shapes if image_np.ndim == 4: image_np = np.squeeze(image_np, axis=0) image_np = np.transpose(image_np, (1, 2, 0)) elif image_np.ndim == 3: if image_np.shape[0] in [1, 3, 4]: image_np = np.transpose(image_np, (1, 2, 0)) elif image_np.ndim == 2: image_np = np.expand_dims(image_np, axis=-1) print(f"Processed numpy array shape: {image_np.shape}") # Normalize the image, accounting for noise if image_np.dtype != np.uint8: if image_np.max() <= 1 + noise_threshold: # Assume the image is in [0, 1] range with some noise image_np = np.clip(image_np, 0, 1) image_np = (image_np * 255).astype(np.uint8) else: # Assume the image is in a wider range, possibly due to noise lower_percentile = np.percentile(image_np, 1) upper_percentile = np.percentile(image_np, 99) image_np = np.clip(image_np, lower_percentile, upper_percentile) image_np = ((image_np - lower_percentile) / (upper_percentile - lower_percentile) * 255).astype(np.uint8) print(f"Normalized array min: {image_np.min()}, max: {image_np.max()}") try: image_pil = Image.fromarray(image_np.squeeze() if image_np.shape[-1] == 1 else image_np) image_pil.save(filepath) print(f"Debug image saved as '{filepath}'") except Exception as e: print(f"Error saving image: {str(e)}") print("Attempting to save as numpy array...") np_filepath = filepath.replace('.png', '.npy') np.save(np_filepath, image_np) print(f"Numpy array saved as '{np_filepath}'") def zero_module(module): for p in module.parameters(): nn.init.zeros_(p) return module