# Adapted from https://github.com/huggingface/diffusers/blob/main/src/diffusers/models/unet_2d_condition.py import json import os from dataclasses import dataclass from typing import Dict, List, Optional, Tuple, Union import torch import torch.nn as nn import torch.utils.checkpoint from diffusers.configuration_utils import ConfigMixin, register_to_config from diffusers.loaders import UNet2DConditionLoadersMixin from diffusers.models.attention_processor import AttentionProcessor from diffusers.models.embeddings import TimestepEmbedding, Timesteps from diffusers.models.modeling_utils import ModelMixin from diffusers.utils import BaseOutput, logging from .resnet import InflatedConv3d, InflatedGroupNorm, MappingNetwork from .unet_blocks_streaming import ( UNetMidBlock3DCrossAttnStreaming, get_down_block, get_up_block, ) logger = logging.get_logger(__name__) # pylint: disable=invalid-name @dataclass class UNet3DConditionStreamingOutput(BaseOutput): sample: torch.FloatTensor kv_cache: List[torch.FloatTensor] class UNet3DConditionStreamingModel(ModelMixin, ConfigMixin, UNet2DConditionLoadersMixin): _supports_gradient_checkpointing = True @register_to_config def __init__( self, sample_size: Optional[int] = None, in_channels: int = 4, out_channels: int = 4, center_input_sample: bool = False, flip_sin_to_cos: bool = True, freq_shift: int = 0, down_block_types: Tuple[str] = ( "CrossAttnDownBlock3D", "CrossAttnDownBlock3D", "CrossAttnDownBlock3D", "DownBlock3D", ), mid_block_type: str = "UNetMidBlock3DCrossAttn", up_block_types: Tuple[str] = ( "UpBlock3D", "CrossAttnUpBlock3D", "CrossAttnUpBlock3D", "CrossAttnUpBlock3D", ), only_cross_attention: Union[bool, Tuple[bool]] = False, block_out_channels: Tuple[int] = (320, 640, 1280, 1280), layers_per_block: int = 2, downsample_padding: int = 1, mid_block_scale_factor: float = 1, act_fn: str = "silu", norm_num_groups: int = 32, norm_eps: float = 1e-5, cross_attention_dim: int = 1280, attention_head_dim: Union[int, Tuple[int]] = 8, dual_cross_attention: bool = False, use_linear_projection: bool = False, class_embed_type: Optional[str] = None, num_class_embeds: Optional[int] = None, upcast_attention: bool = False, resnet_time_scale_shift: str = "default", use_inflated_groupnorm=False, # Additional use_motion_module=False, motion_module_resolutions=(1, 2, 4, 8), motion_module_mid_block=False, motion_module_decoder_only=False, motion_module_type=None, motion_module_kwargs={}, unet_use_cross_frame_attention=False, unet_use_temporal_attention=False, cond_mapping: bool = False, ): super().__init__() self.sample_size = sample_size time_embed_dim = block_out_channels[0] * 4 # input self.conv_in = InflatedConv3d(in_channels, block_out_channels[0], kernel_size=3, padding=(1, 1)) if cond_mapping: self.flow_conv_in = MappingNetwork( conditioning_embedding_channels=block_out_channels[0], conditioning_channels=in_channels, ) # time self.time_proj = Timesteps(block_out_channels[0], flip_sin_to_cos, freq_shift) timestep_input_dim = block_out_channels[0] self.time_embedding = TimestepEmbedding(timestep_input_dim, time_embed_dim) # class embedding if class_embed_type is None and num_class_embeds is not None: self.class_embedding = nn.Embedding(num_class_embeds, time_embed_dim) elif class_embed_type == "timestep": self.class_embedding = TimestepEmbedding(timestep_input_dim, time_embed_dim) elif class_embed_type == "identity": self.class_embedding = nn.Identity(time_embed_dim, time_embed_dim) else: self.class_embedding = None self.down_blocks = nn.ModuleList([]) self.mid_block = None self.up_blocks = nn.ModuleList([]) if isinstance(only_cross_attention, bool): only_cross_attention = [only_cross_attention] * len(down_block_types) if isinstance(attention_head_dim, int): attention_head_dim = (attention_head_dim,) * len(down_block_types) # down output_channel = block_out_channels[0] for i, down_block_type in enumerate(down_block_types): res = 2**i input_channel = output_channel output_channel = block_out_channels[i] is_final_block = i == len(block_out_channels) - 1 down_block = get_down_block( down_block_type, num_layers=layers_per_block, in_channels=input_channel, out_channels=output_channel, temb_channels=time_embed_dim, add_downsample=not is_final_block, resnet_eps=norm_eps, resnet_act_fn=act_fn, resnet_groups=norm_num_groups, cross_attention_dim=cross_attention_dim, attn_num_head_channels=attention_head_dim[i], downsample_padding=downsample_padding, dual_cross_attention=dual_cross_attention, use_linear_projection=use_linear_projection, only_cross_attention=only_cross_attention[i], upcast_attention=upcast_attention, resnet_time_scale_shift=resnet_time_scale_shift, unet_use_cross_frame_attention=unet_use_cross_frame_attention, unet_use_temporal_attention=unet_use_temporal_attention, use_inflated_groupnorm=use_inflated_groupnorm, use_motion_module=use_motion_module and (res in motion_module_resolutions) and (not motion_module_decoder_only), motion_module_type=motion_module_type, motion_module_kwargs=motion_module_kwargs, ) self.down_blocks.append(down_block) # mid if mid_block_type == "UNetMidBlock3DCrossAttn": self.mid_block = UNetMidBlock3DCrossAttnStreaming( in_channels=block_out_channels[-1], temb_channels=time_embed_dim, resnet_eps=norm_eps, resnet_act_fn=act_fn, output_scale_factor=mid_block_scale_factor, resnet_time_scale_shift=resnet_time_scale_shift, cross_attention_dim=cross_attention_dim, attn_num_head_channels=attention_head_dim[-1], resnet_groups=norm_num_groups, dual_cross_attention=dual_cross_attention, use_linear_projection=use_linear_projection, upcast_attention=upcast_attention, unet_use_cross_frame_attention=unet_use_cross_frame_attention, unet_use_temporal_attention=unet_use_temporal_attention, use_inflated_groupnorm=use_inflated_groupnorm, use_motion_module=use_motion_module and motion_module_mid_block, motion_module_type=motion_module_type, motion_module_kwargs=motion_module_kwargs, ) else: raise ValueError(f"unknown mid_block_type : {mid_block_type}") # count how many layers upsample the videos self.num_upsamplers = 0 # up reversed_block_out_channels = list(reversed(block_out_channels)) reversed_attention_head_dim = list(reversed(attention_head_dim)) only_cross_attention = list(reversed(only_cross_attention)) output_channel = reversed_block_out_channels[0] for i, up_block_type in enumerate(up_block_types): res = 2 ** (3 - i) is_final_block = i == len(block_out_channels) - 1 prev_output_channel = output_channel output_channel = reversed_block_out_channels[i] input_channel = reversed_block_out_channels[min(i + 1, len(block_out_channels) - 1)] # add upsample block for all BUT final layer if not is_final_block: add_upsample = True self.num_upsamplers += 1 else: add_upsample = False up_block = get_up_block( up_block_type, num_layers=layers_per_block + 1, in_channels=input_channel, out_channels=output_channel, prev_output_channel=prev_output_channel, temb_channels=time_embed_dim, add_upsample=add_upsample, resnet_eps=norm_eps, resnet_act_fn=act_fn, resnet_groups=norm_num_groups, cross_attention_dim=cross_attention_dim, attn_num_head_channels=reversed_attention_head_dim[i], dual_cross_attention=dual_cross_attention, use_linear_projection=use_linear_projection, only_cross_attention=only_cross_attention[i], upcast_attention=upcast_attention, resnet_time_scale_shift=resnet_time_scale_shift, unet_use_cross_frame_attention=unet_use_cross_frame_attention, unet_use_temporal_attention=unet_use_temporal_attention, use_inflated_groupnorm=use_inflated_groupnorm, use_motion_module=use_motion_module and (res in motion_module_resolutions), motion_module_type=motion_module_type, motion_module_kwargs=motion_module_kwargs, ) self.up_blocks.append(up_block) prev_output_channel = output_channel # out if use_inflated_groupnorm: self.conv_norm_out = InflatedGroupNorm( num_channels=block_out_channels[0], num_groups=norm_num_groups, eps=norm_eps ) else: self.conv_norm_out = nn.GroupNorm( num_channels=block_out_channels[0], num_groups=norm_num_groups, eps=norm_eps ) self.conv_act = nn.SiLU() self.conv_out = InflatedConv3d(block_out_channels[0], out_channels, kernel_size=3, padding=1) def set_info_for_attn(self, height: int, width: int, *args, **kwargs): """set height and width for each attention module.""" motion_module_idx = 0 def _assign_info(module: nn.Module, height: int, width: int, *args, **kwargs): nonlocal motion_module_idx for n, m in module.named_children(): if hasattr(m, "set_info"): m.set_info(height, width, *args, **kwargs) if hasattr(m, "set_index"): m.set_index(motion_module_idx) motion_module_idx += 1 else: _assign_info(m, height, width, *args, **kwargs) h_scale, w_scale = height, width for down_block in self.down_blocks: _assign_info(down_block, h_scale, w_scale, *args, **kwargs) if down_block.downsamplers is not None: h_scale = h_scale // 2 w_scale = w_scale // 2 _assign_info(self.mid_block, h_scale, w_scale, *args, **kwargs) for up_block in self.up_blocks: _assign_info(up_block, h_scale, w_scale, *args, **kwargs) if up_block.upsamplers is not None: h_scale = h_scale * 2 w_scale = w_scale * 2 def prepare_cache(self, denoising_steps_num: int): """prepare cache for temporal self attention.""" kv_cache_dict = {} # no non local, i think def _prepare_cache(module: nn.Module): for n, m in module.named_children(): if hasattr(m, "set_cache"): kv_cache = m.set_cache(denoising_steps_num) idx = m.motion_module_idx kv_cache_dict[idx] = kv_cache if hasattr(m, "prepare_pe_buffer"): m.prepare_pe_buffer() _prepare_cache(m) _prepare_cache(self) max_idx = max(list(kv_cache_dict.keys())) kv_cache_list = [kv_cache_dict[idx] for idx in range(max_idx + 1)] return kv_cache_list def set_attention_slice(self, slice_size): r""" Enable sliced attention computation. When this option is enabled, the attention module will split the input tensor in slices, to compute attention in several steps. This is useful to save some memory in exchange for a small speed decrease. Args: slice_size (`str` or `int` or `list(int)`, *optional*, defaults to `"auto"`): When `"auto"`, halves the input to the attention heads, so attention will be computed in two steps. If `"max"`, maximum amount of memory will be 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`. """ sliceable_head_dims = [] def fn_recursive_retrieve_slicable_dims(module: torch.nn.Module): if hasattr(module, "set_attention_slice"): sliceable_head_dims.append(module.sliceable_head_dim) for child in module.children(): fn_recursive_retrieve_slicable_dims(child) # retrieve number of attention layers for module in self.children(): fn_recursive_retrieve_slicable_dims(module) num_slicable_layers = len(sliceable_head_dims) if slice_size == "auto": # half the attention head size is usually a good trade-off between # speed and memory slice_size = [dim // 2 for dim in sliceable_head_dims] elif slice_size == "max": # make smallest slice possible slice_size = num_slicable_layers * [1] slice_size = num_slicable_layers * [slice_size] if not isinstance(slice_size, list) else slice_size if len(slice_size) != len(sliceable_head_dims): raise ValueError( f"You have provided {len(slice_size)}, but {self.config} has {len(sliceable_head_dims)} different" f" attention layers. Make sure to match `len(slice_size)` to be {len(sliceable_head_dims)}." ) for i in range(len(slice_size)): size = slice_size[i] dim = sliceable_head_dims[i] if size is not None and size > dim: raise ValueError(f"size {size} has to be smaller or equal to {dim}.") # Recursively walk through all the children. # Any children which exposes the set_attention_slice method # gets the message def fn_recursive_set_attention_slice(module: torch.nn.Module, slice_size: List[int]): if hasattr(module, "set_attention_slice"): module.set_attention_slice(slice_size.pop()) for child in module.children(): fn_recursive_set_attention_slice(child, slice_size) reversed_slice_size = list(reversed(slice_size)) for module in self.children(): fn_recursive_set_attention_slice(module, reversed_slice_size) def set_attn_processor( self, processor: Union[AttentionProcessor, Dict[str, AttentionProcessor]], _remove_lora=False ): 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. """ count = len(self.attn_processors.keys()) if isinstance(processor, dict) and len(processor) != count: raise ValueError( f"A dict of processors was passed, but the number of processors {len(processor)} does not match the" f" number of attention layers: {count}. Please make sure to pass {count} processor classes." ) def fn_recursive_attn_processor(name: str, module: torch.nn.Module, processor): if hasattr(module, "set_processor"): if not isinstance(processor, dict): module.set_processor(processor, _remove_lora=_remove_lora) else: module.set_processor(processor.pop(f"{name}.processor"), _remove_lora=_remove_lora) for sub_name, child in module.named_children(): fn_recursive_attn_processor(f"{name}.{sub_name}", child, processor) for name, module in self.named_children(): fn_recursive_attn_processor(name, module, processor) @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. """ # set recursively processors = {} def fn_recursive_add_processors(name: str, module: torch.nn.Module, processors: Dict[str, AttentionProcessor]): if hasattr(module, "get_processor"): processors[f"{name}.processor"] = module.get_processor(return_deprecated_lora=True) for sub_name, child in module.named_children(): fn_recursive_add_processors(f"{name}.{sub_name}", child, processors) return processors for name, module in self.named_children(): fn_recursive_add_processors(name, module, processors) return processors def forward( self, sample: torch.FloatTensor, timestep: Union[torch.Tensor, float, int], encoder_hidden_states: torch.Tensor, temporal_attention_mask: torch.Tensor, depth_sample: torch.Tensor, kv_cache: List[torch.Tensor], # support only update one element in kv-cache pe_idx: torch.Tensor, update_idx: torch.Tensor, class_labels: Optional[torch.Tensor] = None, attention_mask: Optional[torch.Tensor] = None, # support ip-adapter image_embeds: Optional[torch.Tensor] = None, # support controlnet down_block_additional_residuals: Optional[Tuple[torch.Tensor]] = None, mid_block_additional_residual: Optional[torch.Tensor] = None, return_dict: bool = True, ) -> Union[UNet3DConditionStreamingOutput, Tuple]: r""" Args: sample (`torch.FloatTensor`): (batch, channel, 1, height, width) noisy inputs tensor. timestep (`torch.FloatTensor` or `float` or `int`): (batch) timesteps encoder_hidden_states (`torch.FloatTensor`): (batch, sequence_length, feature_dim) encoder hidden states temporal_attention_maks: (batch, window_size, window_size) The attention mask for temporal self-attention. depth_sample (`torch.FloatTensor`): (batch, channel, 1, height, width) depth inputs tensor. kv_cache (`List[torch.FloatTensor]`): kv-cache for each temporal attention module. pe_idx (`torch.FloatTensor`): The positional encoding of temporal attention module for current forward pass. update_idx (`torch.LongTensor`): The index of kv-cache to update. """ # By default samples have to be AT least a multiple of the overall upsampling factor. # The overall upsampling factor is equal to 2 ** (# num of upsampling ayears). # However, the upsampling interpolation output size can be forced to fit any upsampling size # on the fly if necessary. default_overall_up_factor = 2**self.num_upsamplers # upsample size should be forwarded when sample is not a multiple of `default_overall_up_factor` forward_upsample_size = False upsample_size = None if any(s % default_overall_up_factor != 0 for s in sample.shape[-2:]): logger.info("Forward upsample size to force interpolation output size.") forward_upsample_size = True # 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) # center input if necessary if self.config.center_input_sample: sample = 2 * sample - 1.0 # time timesteps = timestep if not torch.is_tensor(timesteps): # 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 = self.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=self.dtype) emb = self.time_embedding(t_emb) if self.class_embedding is not None: if class_labels is None: raise ValueError("class_labels should be provided when num_class_embeds > 0") if self.config.class_embed_type == "timestep": class_labels = self.time_proj(class_labels) class_emb = self.class_embedding(class_labels).to(dtype=self.dtype) emb = emb + class_emb # prepare for ip-adapter if image_embeds is not None: image_embeds = self.encoder_hid_proj(image_embeds).to(encoder_hidden_states.dtype) encoder_hidden_states = torch.cat([encoder_hidden_states, image_embeds], dim=1) # pre-process sample = self.conv_in(sample) if depth_sample is not None: depth_sample = self.flow_conv_in(depth_sample) sample = depth_sample + sample # down down_block_res_samples = (sample,) for downsample_block in self.down_blocks: if hasattr(downsample_block, "has_cross_attention") and downsample_block.has_cross_attention: sample, res_samples = downsample_block( hidden_states=sample, temb=emb, encoder_hidden_states=encoder_hidden_states, attention_mask=attention_mask, temporal_attention_mask=temporal_attention_mask, kv_cache=kv_cache, pe_idx=pe_idx, update_idx=update_idx, ) else: sample, res_samples = downsample_block( hidden_states=sample, temb=emb, encoder_hidden_states=encoder_hidden_states, temporal_attention_mask=temporal_attention_mask, kv_cache=kv_cache, pe_idx=pe_idx, update_idx=update_idx, ) down_block_res_samples += res_samples # support controlnet down_block_res_samples = list(down_block_res_samples) if down_block_additional_residuals is not None: for i, down_block_additional_residual in enumerate(down_block_additional_residuals): if down_block_additional_residual.dim() == 4: # broadcast down_block_additional_residual = down_block_additional_residual.unsqueeze(2) down_block_res_samples[i] = down_block_res_samples[i] + down_block_additional_residual # mid sample = self.mid_block( sample, emb, encoder_hidden_states=encoder_hidden_states, attention_mask=attention_mask, temporal_attention_mask=temporal_attention_mask, kv_cache=kv_cache, pe_idx=pe_idx, update_idx=update_idx, ) # support controlnet if mid_block_additional_residual is not None: if mid_block_additional_residual.dim() == 4: # broadcast mid_block_additional_residual = mid_block_additional_residual.unsqueeze(2) sample = sample + mid_block_additional_residual # up for i, upsample_block in enumerate(self.up_blocks): is_final_block = i == len(self.up_blocks) - 1 res_samples = down_block_res_samples[-len(upsample_block.resnets) :] down_block_res_samples = down_block_res_samples[: -len(upsample_block.resnets)] # if we have not reached the final block and need to forward the # upsample size, we do it here if not is_final_block and forward_upsample_size: upsample_size = down_block_res_samples[-1].shape[2:] if hasattr(upsample_block, "has_cross_attention") and upsample_block.has_cross_attention: sample = upsample_block( hidden_states=sample, temb=emb, res_hidden_states_tuple=res_samples, encoder_hidden_states=encoder_hidden_states, upsample_size=upsample_size, attention_mask=attention_mask, temporal_attention_mask=temporal_attention_mask, kv_cache=kv_cache, pe_idx=pe_idx, update_idx=update_idx, ) else: sample = upsample_block( hidden_states=sample, temb=emb, res_hidden_states_tuple=res_samples, upsample_size=upsample_size, encoder_hidden_states=encoder_hidden_states, temporal_attention_mask=temporal_attention_mask, kv_cache=kv_cache, pe_idx=pe_idx, update_idx=update_idx, ) # post-process sample = self.conv_norm_out(sample) sample = self.conv_act(sample) sample = self.conv_out(sample) if not return_dict: return (sample, kv_cache) return UNet3DConditionStreamingOutput(sample=sample, kv_cache=kv_cache) @classmethod def from_pretrained_2d(cls, pretrained_model_path, subfolder=None, unet_additional_kwargs=None): if subfolder is not None: pretrained_model_path = os.path.join(pretrained_model_path, subfolder) print(f"loaded 3D unet's pretrained weights from {pretrained_model_path} ...") config_file = os.path.join(pretrained_model_path, "config.json") if not os.path.isfile(config_file): raise RuntimeError(f"{config_file} does not exist") with open(config_file, "r") as f: config = json.load(f) config["_class_name"] = cls.__name__ config["down_block_types"] = [ "CrossAttnDownBlock3D", "CrossAttnDownBlock3D", "CrossAttnDownBlock3D", "DownBlock3D", ] config["up_block_types"] = ["UpBlock3D", "CrossAttnUpBlock3D", "CrossAttnUpBlock3D", "CrossAttnUpBlock3D"] from diffusers.utils import WEIGHTS_NAME model = cls.from_config(config, **unet_additional_kwargs) model_file = os.path.join(pretrained_model_path, WEIGHTS_NAME) if not os.path.isfile(model_file): raise RuntimeError(f"{model_file} does not exist") state_dict = torch.load(model_file, map_location="cpu") m, u = model.load_state_dict(state_dict, strict=False) print(f"### missing keys: {len(m)}; \n### unexpected keys: {len(u)};") params = [p.numel() if "motion_modules." in n else 0 for n, p in model.named_parameters()] print(f"### Motion Module Parameters: {sum(params) / 1e6} M") return model