import einops import torch import torch as th import torch.nn as nn from ldm.modules.diffusionmodules.util import ( conv_nd, linear, zero_module, timestep_embedding, ) from einops import rearrange, repeat from torchvision.utils import make_grid from ldm.modules.attention import SpatialTransformer from ldm.modules.diffusionmodules.openaimodel import UNetModel, TimestepEmbedSequential, ResBlock, Downsample, AttentionBlock from ldm.models.diffusion.ddpm import LatentDiffusion from ldm.util import log_txt_as_img, exists, instantiate_from_config from ldm.models.diffusion.ddim import DDIMSampler from ldm.modules.ema import LitEma from contextlib import contextmanager, nullcontext from cldm.model import load_state_dict import numpy as np from torch.optim.lr_scheduler import LambdaLR, CosineAnnealingLR, OneCycleLR def disabled_train(self, mode=True): """Overwrite model.train with this function to make sure train/eval mode does not change anymore.""" return self class ControlledUnetModel(UNetModel): def forward(self, x, timesteps=None, context=None, control=None, only_mid_control=False, context_glyph= None, **kwargs): hs = [] with torch.no_grad(): 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, context) hs.append(h) h = self.middle_block(h, emb, context) if control is not None: h += control.pop() for i, module in enumerate(self.output_blocks): if only_mid_control or control is None: h = torch.cat([h, hs.pop()], dim=1) else: h = torch.cat([h, hs.pop() + control.pop()], dim=1) h = module(h, emb, context if context_glyph is None else context_glyph) h = h.type(x.dtype) return self.out(h) class ControlNet(nn.Module): def __init__( self, image_size, in_channels, model_channels, hint_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, 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, disable_middle_self_attn=False, use_linear_in_transformer=False, ): 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.dims = dims self.image_size = image_size self.in_channels = in_channels self.model_channels = model_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 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.") 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 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), ) self.input_blocks = nn.ModuleList( [ TimestepEmbedSequential( conv_nd(dims, in_channels, model_channels, 3, padding=1) ) ] ) self.zero_convs = nn.ModuleList([self.make_zero_conv(model_channels)]) self.input_hint_block = TimestepEmbedSequential( conv_nd(dims, hint_channels, 16, 3, padding=1), nn.SiLU(), conv_nd(dims, 16, 16, 3, padding=1), nn.SiLU(), conv_nd(dims, 16, 32, 3, padding=1, stride=2), nn.SiLU(), conv_nd(dims, 32, 32, 3, padding=1), nn.SiLU(), conv_nd(dims, 32, 96, 3, padding=1, stride=2), nn.SiLU(), conv_nd(dims, 96, 96, 3, padding=1), nn.SiLU(), conv_nd(dims, 96, 256, 3, padding=1, stride=2), nn.SiLU(), zero_module(conv_nd(dims, 256, 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 exists(disable_self_attentions): disabled_sa = disable_self_attentions[level] else: disabled_sa = False if not exists(num_attention_blocks) or nr < num_attention_blocks[level]: layers.append( 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, use_linear=use_linear_in_transformer, use_checkpoint=use_checkpoint ) ) self.input_blocks.append(TimestepEmbedSequential(*layers)) self.zero_convs.append(self.make_zero_conv(ch)) 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) self.zero_convs.append(self.make_zero_conv(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, disable_self_attn=disable_middle_self_attn, use_linear=use_linear_in_transformer, use_checkpoint=use_checkpoint ), ResBlock( ch, time_embed_dim, dropout, dims=dims, use_checkpoint=use_checkpoint, use_scale_shift_norm=use_scale_shift_norm, ), ) self.middle_block_out = self.make_zero_conv(ch) self._feature_size += ch def make_zero_conv(self, channels): return TimestepEmbedSequential(zero_module(conv_nd(self.dims, channels, channels, 1, padding=0))) def forward(self, x, hint, timesteps, context, **kwargs): t_emb = timestep_embedding(timesteps, self.model_channels, repeat_only=False) emb = self.time_embed(t_emb) guided_hint = self.input_hint_block(hint, emb, context) outs = [] h = x.type(self.dtype) for module, zero_conv in zip(self.input_blocks, self.zero_convs): if guided_hint is not None: h = module(h, emb, context) h += guided_hint guided_hint = None else: h = module(h, emb, context) outs.append(zero_conv(h, emb, context)) h = self.middle_block(h, emb, context) outs.append(self.middle_block_out(h, emb, context)) return outs class ControlLDM(LatentDiffusion): def __init__(self, control_stage_config, control_key, only_mid_control, sd_locked = True, concat_textemb = False, trans_textemb=False, trans_textemb_config = None, learnable_conscale = False, guess_mode=False, sep_lr = False, decoder_lr = 1.0**-4, add_glyph_control = False, glyph_control_config = None, glycon_wd = 0.2, glycon_lr = 1.0**-4, glycon_sched = "lambda", glyph_control_key = "centered_hint", sep_cond_txt = False, exchange_cond_txt = False, max_step = None, multiple_optimizers = False, deepspeed = False, trans_glyph_lr = 1.0**-5, *args, **kwargs ): #sep_cap_for_2b = False use_ema = kwargs.pop("use_ema", False) ckpt_path = kwargs.pop("ckpt_path", None) reset_ema = kwargs.pop("reset_ema", False) only_model= kwargs.pop("only_model", False) reset_num_ema_updates = kwargs.pop("reset_num_ema_updates", False) keep_num_ema_updates = kwargs.pop("keep_num_ema_updates", False) ignore_keys = kwargs.pop("ignore_keys", []) super().__init__(*args, use_ema=False, **kwargs) self.control_model = instantiate_from_config(control_stage_config) self.control_key = control_key self.only_mid_control = only_mid_control self.learnable_conscale = learnable_conscale conscale_init = [1.0] * 13 if not guess_mode else [(0.825 ** float(12 - i)) for i in range(13)] if learnable_conscale: # self.control_scales = nn.Parameter(torch.ones(13), requires_grad=True) self.control_scales = nn.Parameter(torch.Tensor(conscale_init), requires_grad=True) else: # TODO: register the buffer self.control_scales = conscale_init #[1.0] * 13 self.sd_locked = sd_locked self.concat_textemb = concat_textemb # update self.trans_textemb = False if trans_textemb and trans_textemb_config is not None: self.trans_textemb = True self.instantiate_trans_textemb_model(trans_textemb_config) # self.sep_cap_for_2b = sep_cap_for_2b self.sep_lr = sep_lr self.decoder_lr = decoder_lr self.sep_cond_txt = sep_cond_txt self.exchange_cond_txt = exchange_cond_txt # update (4.18) self.multiple_optimizers = multiple_optimizers self.add_glyph_control = False self.glyph_control_key = glyph_control_key self.freeze_glyph_image_encoder = True self.glyph_image_encoder_type = "CLIP" self.max_step = max_step self.trans_glyph_embed = False self.trans_glyph_lr = trans_glyph_lr if deepspeed: try: from deepspeed.ops.adam import FusedAdam, DeepSpeedCPUAdam self.optimizer = DeepSpeedCPUAdam #FusedAdam except: print("could not import FuseAdam from deepspeed") self.optimizer = torch.optim.AdamW else: self.optimizer = torch.optim.AdamW if add_glyph_control and glyph_control_config is not None: self.add_glyph_control = True self.glycon_wd = glycon_wd self.glycon_lr = glycon_lr self.glycon_sched = glycon_sched self.instantiate_glyph_control_model(glyph_control_config) if self.glyph_control_model.trans_glyph_emb_model is not None: self.trans_glyph_embed = True self.use_ema = use_ema if self.use_ema: #TODO: trainable glyph Image encoder # assert self.sd_locked == True self.model_ema = LitEma(self.control_model, init_num_updates= 0) print(f"Keeping EMAs of {len(list(self.model_ema.buffers()))}.") if not self.sd_locked: # Update self.model_diffoutblock_ema = LitEma(self.model.diffusion_model.output_blocks, init_num_updates= 0) print(f"Keeping diffoutblock EMAs of {len(list(self.model_diffoutblock_ema.buffers()))}.") self.model_diffout_ema = LitEma(self.model.diffusion_model.out, init_num_updates= 0) print(f"Keeping diffout EMAs of {len(list(self.model_diffout_ema.buffers()))}.") if not self.freeze_glyph_image_encoder: self.model_glyphcon_ema = LitEma(self.glyph_control_model.image_encoder, init_num_updates=0) print(f"Keeping glyphcon EMAs of {len(list(self.model_glyphcon_ema.buffers()))}.") if self.trans_glyph_embed: self.model_transglyph_ema = LitEma(self.glyph_control_model.trans_glyph_emb_model, init_num_updates=0) print(f"Keeping glyphcon EMAs of {len(list(self.model_transglyph_ema.buffers()))}.") if ckpt_path is not None: ema_num_updates = self.init_from_ckpt(ckpt_path, ignore_keys, only_model=only_model) self.restarted_from_ckpt = True # if reset_ema: # assert self.use_ema if self.use_ema and reset_ema: print( f"Resetting ema to pure model weights. This is useful when restoring from an ema-only checkpoint.") self.model_ema = LitEma(self.control_model, init_num_updates= ema_num_updates if keep_num_ema_updates else 0) if not self.sd_locked: # Update self.model_diffoutblock_ema = LitEma(self.model.diffusion_model.output_blocks, init_num_updates= ema_num_updates if keep_num_ema_updates else 0) self.model_diffout_ema = LitEma(self.model.diffusion_model.out, init_num_updates= ema_num_updates if keep_num_ema_updates else 0) if not self.freeze_glyph_image_encoder: self.model_glyphcon_ema = LitEma(self.glyph_control_model.image_encoder, init_num_updates= ema_num_updates if keep_num_ema_updates else 0) if self.trans_glyph_embed: self.model_transglyph_ema = LitEma(self.glyph_control_model.trans_glyph_emb_model, init_num_updates= ema_num_updates if keep_num_ema_updates else 0) if reset_num_ema_updates: print(" +++++++++++ WARNING: RESETTING NUM_EMA UPDATES TO ZERO +++++++++++ ") assert self.use_ema self.model_ema.reset_num_updates() if not self.sd_locked: # Update self.model_diffoutblock_ema.reset_num_updates() self.model_diffout_ema.reset_num_updates() if not self.freeze_glyph_image_encoder: self.model_glyphcon_ema.reset_num_updates() if self.trans_glyph_embed: self.model_transglyph_ema.reset_num_updates() # self.freeze_unet() @contextmanager def ema_scope(self, context=None): if self.use_ema: # TODO: fix the bug while adding transemb_model or trainable control scales self.model_ema.store(self.control_model.parameters()) self.model_ema.copy_to(self.control_model) if not self.sd_locked: # Update self.model_diffoutblock_ema.store(self.model.diffusion_model.output_blocks.parameters()) self.model_diffoutblock_ema.copy_to(self.model.diffusion_model.output_blocks) self.model_diffout_ema.store(self.model.diffusion_model.out.parameters()) self.model_diffout_ema.copy_to(self.model.diffusion_model.out) if not self.freeze_glyph_image_encoder: self.model_glyphcon_ema.store(self.glyph_control_model.image_encoder.parameters()) self.model_glyphcon_ema.copy_to(self.glyph_control_model.image_encoder) if self.trans_glyph_embed: self.model_transglyph_ema.store(self.glyph_control_model.trans_glyph_emb_model.parameters()) self.model_transglyph_ema.copy_to(self.glyph_control_model.trans_glyph_emb_model) if context is not None: print(f"{context}: Switched ControlNet to EMA weights") try: yield None finally: if self.use_ema: self.model_ema.restore(self.control_model.parameters()) if not self.sd_locked: # Update self.model_diffoutblock_ema.restore(self.model.diffusion_model.output_blocks.parameters()) self.model_diffout_ema.restore(self.model.diffusion_model.out.parameters()) if not self.freeze_glyph_image_encoder: self.model_glyphcon_ema.restore(self.glyph_control_model.image_encoder.parameters()) if self.trans_glyph_embed: self.model_transglyph_ema.restore(self.glyph_control_model.trans_glyph_emb_model.parameters()) if context is not None: print(f"{context}: Restored training weights of ControlNet") @torch.no_grad() def init_from_ckpt(self, path, ignore_keys=list(), only_model=False): if path.endswith("model_states.pt"): sd = torch.load(path, map_location='cpu')["module"] else: # sd = load_state_dict(path, location='cpu') # abandoned sd = torch.load(path, map_location="cpu") if "state_dict" in list(sd.keys()): sd = sd["state_dict"] keys_ = list(sd.keys())[:] for k in keys_: if k.startswith("module."): nk = k[7:] sd[nk] = sd[k] del sd[k] keys = list(sd.keys()) for k in keys: for ik in ignore_keys: if k.startswith(ik): print("Deleting key {} from state_dict.".format(k)) del sd[k] # missing, unexpected = self.load_state_dict(sd, strict=False) if not only_model else self.model.load_state_dict( # sd, strict=False) if not only_model: missing, unexpected = self.load_state_dict(sd, strict=False) elif path.endswith(".bin"): missing, unexpected = self.model.diffusion_model.load_state_dict(sd, strict=False) elif path.endswith(".ckpt"): missing, unexpected = self.model.load_state_dict(sd, strict=False) print(f"Restored from {path} with {len(missing)} missing and {len(unexpected)} unexpected keys") if len(missing) > 0: print(f"Missing Keys:\n {missing}") if len(unexpected) > 0: print(f"\nUnexpected Keys:\n {unexpected}") if "model_ema.num_updates" in sd and "model_ema.num_updates" not in unexpected: return sd["model_ema.num_updates"].item() else: return 0 def instantiate_trans_textemb_model(self, config): model = instantiate_from_config(config) params = [] for i in range(model.emb_num): if model.trans_trainable[i]: params += list(model.trans_list[i].parameters()) else: for param in model.trans_list[i].parameters(): param.requires_grad = False self.trans_textemb_model = model self.trans_textemb_params = params # add def instantiate_glyph_control_model(self, config): model = instantiate_from_config(config) # params = [] self.freeze_glyph_image_encoder = model.freeze_image_encoder #image_encoder.freeze_model self.glyph_control_model = model self.glyph_image_encoder_type = model.image_encoder_type # self.glyph_control_optim = torch.optim.AdamW([ # {"params": gain_or_bias_params, "weight_decay": 0.}, # "lr": self.glycon_lr}, # {"params": rest_params, "weight_decay": self.glycon_wd} #, "lr": self.glycon_lr}, # ], # lr = self.glycon_lr # ) # params += list(model.image_encoder.parameters()) @torch.no_grad() def get_input(self, batch, k, bs=None, *args, **kwargs): x, c = super().get_input(batch, self.first_stage_key, *args, **kwargs) control = batch[self.control_key] if bs is not None: control = control[:bs] control = control.to(self.device) control = einops.rearrange(control, 'b h w c -> b c h w') control = control.to(memory_format=torch.contiguous_format).float() if self.add_glyph_control: assert self.glyph_control_key in batch.keys() glyph_control = batch[self.glyph_control_key] if bs is not None: glyph_control = glyph_control[:bs] glycon_samples = [] for glycon_sample in glyph_control: glycon_sample = glycon_sample.to(self.device) glycon_sample = einops.rearrange(glycon_sample, 'b h w c -> b c h w') glycon_sample = glycon_sample.to(memory_format=torch.contiguous_format).float() glycon_samples.append(glycon_sample) # return x, dict(c_crossattn=[c], c_concat=[control]) return x, dict(c_crossattn=[c] if not isinstance(c, list) else c, c_concat=[control], c_glyph=glycon_samples) return x, dict(c_crossattn=[c] if not isinstance(c, list) else c, c_concat=[control]) def apply_model(self, x_noisy, t, cond, *args, **kwargs): assert isinstance(cond, dict) diffusion_model = self.model.diffusion_model #update embdim_list = [] for c in cond["c_crossattn"]: embdim_list.append(c.shape[-1]) embdim_list = np.array(embdim_list) if np.sum(embdim_list != diffusion_model.context_dim): assert self.trans_textemb if self.trans_textemb: assert self.trans_textemb_model cond_txt_list = self.trans_textemb_model(cond["c_crossattn"]) # if len(cond_txt_list) == 2: # print("cond_txt_2 max: {}".format(torch.max(torch.abs(cond_txt_list[1])))) else: cond_txt_list = cond["c_crossattn"] assert len(cond_txt_list) > 0 if self.sep_cond_txt: cond_txt = cond_txt_list[0] cond_txt_2 = None if len(cond_txt_list) == 1 else cond_txt_list[1] else: if len(cond_txt_list) > 1: cond_txt = cond_txt_list[0] # input text embedding of the pretrained SD if not self.concat_textemb: # currently len(cond_txt_list) <= 2 cond_txt_2 = torch.cat(cond_txt_list[1:], 1) # input text embedding of the ControlNet branch else: cond_txt_2 = torch.cat(cond_txt_list, 1) if self.exchange_cond_txt: txt_buffer = cond_txt cond_txt = cond_txt_2 cond_txt_2 = txt_buffer print("len cond_txt_list: {} | cond_txt_1 shape: {} | cond_txt_2 shape: {}".format(len(cond_txt_list), cond_txt.shape, cond_txt_2.shape)) else: cond_txt = torch.cat(cond_txt_list, 1) cond_txt_2 = None context_glyph = None if self.add_glyph_control: assert "c_glyph" in cond.keys() if cond["c_glyph"] is not None: context_glyph = self.glyph_control_model(cond["c_glyph"], text_embed = cond_txt_list[-1] if len(cond_txt_list) == 3 else cond_txt) else: context_glyph = cond_txt_list[-1] if len(cond_txt_list) == 3 else cond_txt # if cond_txt_2 is None: # print("cond_txt_2 is None") if cond['c_concat'] is None: eps = diffusion_model(x=x_noisy, timesteps=t, context=cond_txt, control=None, only_mid_control=self.only_mid_control, context_glyph = context_glyph) else: control = self.control_model(x=x_noisy, hint=torch.cat(cond['c_concat'], 1), timesteps=t, context=cond_txt if cond_txt_2 is None else cond_txt_2) control = [c * scale for c, scale in zip(control, self.control_scales)] eps = diffusion_model(x=x_noisy, timesteps=t, context=cond_txt, control=control, only_mid_control=self.only_mid_control, context_glyph=context_glyph) return eps @torch.no_grad() def get_unconditional_conditioning(self, N): return self.get_learned_conditioning([""] * N) # Maybe not useful: modify the codes to fit the separate input captions # @torch.no_grad() # def get_unconditional_conditioning(self, N): # return self.get_learned_conditioning([""] * N) if not self.sep_cap_for_2b else self.get_learned_conditioning([[""] * N, [""] * N]) # TODO: adapt to new model @torch.no_grad() def log_images(self, batch, N=4, n_row=2, sample=False, ddim_steps=50, ddim_eta=0.0, return_keys=None, quantize_denoised=True, inpaint=True, plot_denoise_rows=False, plot_progressive_rows=True, plot_diffusion_rows=False, unconditional_guidance_scale=9.0, unconditional_guidance_label=None, use_ema_scope=True, **kwargs): use_ddim = ddim_steps is not None log = dict() z, c = self.get_input(batch, self.first_stage_key, bs=N) c_cat, c = c["c_concat"][0][:N], c["c_crossattn"][0][:N] N = min(z.shape[0], N) n_row = min(z.shape[0], n_row) log["reconstruction"] = self.decode_first_stage(z) log["control"] = c_cat * 2.0 - 1.0 log["conditioning"] = log_txt_as_img((512, 512), batch[self.cond_stage_key], size=16) if plot_diffusion_rows: # get diffusion row diffusion_row = list() z_start = z[:n_row] for t in range(self.num_timesteps): if t % self.log_every_t == 0 or t == self.num_timesteps - 1: t = repeat(torch.tensor([t]), '1 -> b', b=n_row) t = t.to(self.device).long() noise = torch.randn_like(z_start) z_noisy = self.q_sample(x_start=z_start, t=t, noise=noise) diffusion_row.append(self.decode_first_stage(z_noisy)) diffusion_row = torch.stack(diffusion_row) # n_log_step, n_row, C, H, W diffusion_grid = rearrange(diffusion_row, 'n b c h w -> b n c h w') diffusion_grid = rearrange(diffusion_grid, 'b n c h w -> (b n) c h w') diffusion_grid = make_grid(diffusion_grid, nrow=diffusion_row.shape[0]) log["diffusion_row"] = diffusion_grid if sample: # get denoise row samples, z_denoise_row = self.sample_log(cond={"c_concat": [c_cat], "c_crossattn": [c]}, batch_size=N, ddim=use_ddim, ddim_steps=ddim_steps, eta=ddim_eta) x_samples = self.decode_first_stage(samples) log["samples"] = x_samples if plot_denoise_rows: denoise_grid = self._get_denoise_row_from_list(z_denoise_row) log["denoise_row"] = denoise_grid if unconditional_guidance_scale > 1.0: uc_cross = self.get_unconditional_conditioning(N) uc_cat = c_cat # torch.zeros_like(c_cat) uc_full = {"c_concat": [uc_cat], "c_crossattn": [uc_cross]} samples_cfg, _ = self.sample_log(cond={"c_concat": [c_cat], "c_crossattn": [c]}, batch_size=N, ddim=use_ddim, ddim_steps=ddim_steps, eta=ddim_eta, unconditional_guidance_scale=unconditional_guidance_scale, unconditional_conditioning=uc_full, ) x_samples_cfg = self.decode_first_stage(samples_cfg) log[f"samples_cfg_scale_{unconditional_guidance_scale:.2f}"] = x_samples_cfg return log # TODO: adapt to new model @torch.no_grad() def sample_log(self, cond, batch_size, ddim, ddim_steps, **kwargs): ddim_sampler = DDIMSampler(self) b, c, h, w = cond["c_concat"][0].shape shape = (self.channels, h // 8, w // 8) samples, intermediates = ddim_sampler.sample(ddim_steps, batch_size, shape, cond, verbose=False, **kwargs) return samples, intermediates # add def training_step(self, batch, batch_idx, optimizer_idx=0): loss = super().training_step(batch, batch_idx, optimizer_idx) if self.use_scheduler and not self.sd_locked and self.sep_lr: decoder_lr = self.optimizers().param_groups[1]["lr"] self.log('decoder_lr_abs', decoder_lr, prog_bar=True, logger=True, on_step=True, on_epoch=False) if self.trans_glyph_embed and self.freeze_glyph_image_encoder: trans_glyph_embed_lr = self.optimizers().param_groups[2]["lr"] self.log('trans_glyph_embed_lr_abs', trans_glyph_embed_lr, prog_bar=True, logger=True, on_step=True, on_epoch=False) return loss def configure_optimizers(self): lr = self.learning_rate params = list(self.control_model.parameters()) if self.trans_textemb: params += self.trans_textemb_params #list(self.trans_textemb_model.parameters()) if self.learnable_conscale: params += [self.control_scales] params_wlr = [] decoder_params = None if not self.sd_locked: decoder_params = list(self.model.diffusion_model.output_blocks.parameters()) decoder_params += list(self.model.diffusion_model.out.parameters()) if not self.sep_lr: params.extend(decoder_params) decoder_params = None params_wlr.append({"params": params, "lr": lr}) if decoder_params is not None: params_wlr.append({"params": decoder_params, "lr": self.decoder_lr}) # if not self.sep_lr: # opt = torch.optim.AdamW(params, lr=lr) # else: # opt = torch.optim.AdamW( # [ # {"params": params}, # {"params": decoder_params, "lr": self.decoder_lr} # ], lr=lr # ) if not self.freeze_glyph_image_encoder: if self.glyph_image_encoder_type == "CLIP": # assert self.sep_lr # follow the training codes in the OpenClip repo # https://github.com/mlfoundations/open_clip/blob/main/src/training/main.py#L303 exclude = lambda n, p: p.ndim < 2 or "bn" in n or "ln" in n or "bias" in n or 'logit_scale' in n include = lambda n, p: not exclude(n, p) # named_parameters = list(model.image_encoder.named_parameters()) named_parameters = list(self.glyph_control_model.image_encoder.named_parameters()) gain_or_bias_params = [p for n, p in named_parameters if exclude(n, p) and p.requires_grad] rest_params = [p for n, p in named_parameters if include(n, p) and p.requires_grad] self.glyph_control_params_wlr = [ {"params": gain_or_bias_params, "weight_decay": 0., "lr": self.glycon_lr}, {"params": rest_params, "weight_decay": self.glycon_wd, "lr": self.glycon_lr}, ] if not self.freeze_glyph_image_encoder and not self.multiple_optimizers: params_wlr.extend(self.glyph_control_params_wlr) if self.trans_glyph_embed: trans_glyph_params = list(self.glyph_control_model.trans_glyph_emb_model.parameters()) params_wlr.append({"params": trans_glyph_params, "lr": self.trans_glyph_lr}) # opt = torch.optim.AdamW(params_wlr) opt = self.optimizer(params_wlr) opts = [opt] if not self.freeze_glyph_image_encoder and self.multiple_optimizers: glyph_control_opt = self.optimizer(self.glyph_control_params_wlr) #torch.optim.AdamW(self.glyph_control_params_wlr) opts.append(glyph_control_opt) # updated schedulers = [] if self.use_scheduler: assert 'target' in self.scheduler_config scheduler_func = instantiate_from_config(self.scheduler_config) print("Setting up LambdaLR scheduler...") schedulers = [ { 'scheduler': LambdaLR( opt, lr_lambda= [scheduler_func.schedule] * len(params_wlr) #if not self.sep_lr else [scheduler_func.schedule, scheduler_func.schedule] ), 'interval': 'step', 'frequency': 1 }] if not self.freeze_glyph_image_encoder and self.multiple_optimizers: if self.glycon_sched == "cosine" and self.max_step is not None: glyph_scheduler = CosineAnnealingLR(glyph_control_opt, T_max=self.max_step) #: max_step elif self.glycon_sched == "onecycle" and self.max_step is not None: glyph_scheduler = OneCycleLR( glyph_control_opt, max_lr=self.glycon_lr, total_steps=self.max_step, #: max_step pct_start=0.0001, anneal_strategy="cos" #'linear' ) # elif self.glycon_sched == "lambda": else: glyph_scheduler = LambdaLR( glyph_control_opt, lr_lambda = [scheduler_func.schedule] * len(self.glyph_control_params_wlr) ) schedulers.append( { "scheduler": glyph_scheduler, "interval": 'step', 'frequency': 1 } ) return opts, schedulers # TODO: adapt to new model def low_vram_shift(self, is_diffusing): if is_diffusing: self.model = self.model.cuda() self.control_model = self.control_model.cuda() self.first_stage_model = self.first_stage_model.cpu() self.cond_stage_model = self.cond_stage_model.cpu() else: self.model = self.model.cpu() self.control_model = self.control_model.cpu() self.first_stage_model = self.first_stage_model.cuda() self.cond_stage_model = self.cond_stage_model.cuda() # ema def on_train_batch_end(self, *args, **kwargs): if self.use_ema: self.model_ema(self.control_model) if not self.sd_locked: # Update self.model_diffoutblock_ema(self.model.diffusion_model.output_blocks) self.model_diffout_ema(self.model.diffusion_model.out) if not self.freeze_glyph_image_encoder: self.model_glyphcon_ema(self.glyph_control_model.image_encoder) if self.trans_glyph_embed: self.model_transglyph_ema(self.glyph_control_model.trans_glyph_emb_model) if self.log_all_grad_norm: zeroconvs = list(self.control_model.input_hint_block.named_parameters())[-2:] zeroconvs.extend( list(self.control_model.zero_convs.named_parameters()) ) for item in zeroconvs: self.log( "zero_convs/{}_norm".format(item[0]), item[1].cpu().detach().norm().item(), prog_bar=False, logger=True, on_step=True, on_epoch=False ) self.log( "zero_convs/{}_max".format(item[0]), torch.max(item[1].cpu().detach()).item(), #TODO: lack torch.abs prog_bar=False, logger=True, on_step=True, on_epoch=False ) gradnorm_list = [] for param_group in self.trainer.optimizers[0].param_groups: for p in param_group['params']: # assert p.requires_grad and p.grad is not None if p.requires_grad and p.grad is not None: grad_norm_v = p.grad.cpu().detach().norm().item() gradnorm_list.append(grad_norm_v) # for name, p in self.named_parameters(): # if p.requires_grad and p.grad is not None: # grad_norm_v = p.grad.detach().norm().item() # gradnorm_list.append(grad_norm_v) # if "textemb_merge_model" in name: # self.log("all_gradients/{}_norm".format(name), # gradnorm_list[-1], # prog_bar=False, logger=True, on_step=True, on_epoch=False # ) # # if grad_norm_v > 0.1: # # print("the norm of gradient w.r.t {} > 0.1: {:.2f}".format # # ( # # name, grad_norm_v # # )) if len(gradnorm_list): self.log("all_gradients/grad_norm_mean", np.mean(gradnorm_list), prog_bar=False, logger=True, on_step=True, on_epoch=False ) self.log("all_gradients/grad_norm_max", np.max(gradnorm_list), prog_bar=False, logger=True, on_step=True, on_epoch=False ) self.log("all_gradients/grad_norm_min", np.min(gradnorm_list), prog_bar=False, logger=True, on_step=True, on_epoch=False ) self.log("all_gradients/param_num", len(gradnorm_list), prog_bar=False, logger=True, on_step=True, on_epoch=False ) if self.trans_textemb: for name, p in self.trans_textemb_model.named_parameters(): if p.requires_grad and p.grad is not None: self.log( "trans_textemb_gradient_norm/{}".format(name), p.grad.cpu().detach().norm().item(), prog_bar=False, logger=True, on_step=True, on_epoch=False ) self.log( "trans_textemb_params/{}_norm".format(name), p.cpu().detach().norm().item(), prog_bar=False, logger=True, on_step=True, on_epoch=False ) self.log( "trans_textemb_params/{}_abs_max".format(name), torch.max(torch.abs(p.cpu().detach())).item(), prog_bar=False, logger=True, on_step=True, on_epoch=False ) if self.trans_glyph_embed: for name, p in self.glyph_control_model.trans_glyph_emb_model.named_parameters(): if p.requires_grad and p.grad is not None: self.log( "trans_glyph_embed_gradient_norm/{}".format(name), p.grad.cpu().detach().norm().item(), prog_bar=False, logger=True, on_step=True, on_epoch=False ) self.log( "trans_glyph_embed_params/{}_norm".format(name), p.cpu().detach().norm().item(), prog_bar=False, logger=True, on_step=True, on_epoch=False ) self.log( "trans_glyph_embed_params/{}_abs_max".format(name), torch.max(torch.abs(p.cpu().detach())).item(), prog_bar=False, logger=True, on_step=True, on_epoch=False ) if self.learnable_conscale: for i in range(len(self.control_scales)): self.log( "control_scale/control_{}".format(i), self.control_scales[i], prog_bar=False, logger=True, on_step=True, on_epoch=False ) del gradnorm_list del zeroconvs # def freeze_unet(self): # # Have some bugs # self.model.eval() # # self.model.train = disabled_train # for param in self.model.parameters(): # param.requires_grad = False # if not self.sd_locked: # self.model.diffusion_model.output_blocks.train() # self.model.diffusion_model.out.train() # for param in self.model.diffusion_model.out.parameters(): # param.requires_grad = True # for param in self.model.diffusion_model.output_blocks.parameters(): # param.requires_grad = True