diff options
35 files changed, 2187 insertions, 128 deletions
@@ -135,6 +135,7 @@ The documentation was moved from this README over to the project's [wiki](https: - SwinIR - https://github.com/JingyunLiang/SwinIR
- Swin2SR - https://github.com/mv-lab/swin2sr
- LDSR - https://github.com/Hafiidz/latent-diffusion
+- MiDaS - https://github.com/isl-org/MiDaS
- Ideas for optimizations - https://github.com/basujindal/stable-diffusion
- Cross Attention layer optimization - Doggettx - https://github.com/Doggettx/stable-diffusion, original idea for prompt editing.
- Cross Attention layer optimization - InvokeAI, lstein - https://github.com/invoke-ai/InvokeAI (originally http://github.com/lstein/stable-diffusion)
diff --git a/modules/ldsr_model_arch.py b/extensions-builtin/LDSR/ldsr_model_arch.py index 90e0a2f0..a87d1ef9 100644 --- a/modules/ldsr_model_arch.py +++ b/extensions-builtin/LDSR/ldsr_model_arch.py @@ -22,6 +22,7 @@ class LDSR: pl_sd = torch.load(self.modelPath, map_location="cpu") sd = pl_sd["state_dict"] config = OmegaConf.load(self.yamlPath) + config.model.target = "ldm.models.diffusion.ddpm.LatentDiffusionV1" model = instantiate_from_config(config.model) model.load_state_dict(sd, strict=False) model.cuda() diff --git a/extensions-builtin/LDSR/preload.py b/extensions-builtin/LDSR/preload.py new file mode 100644 index 00000000..d746007c --- /dev/null +++ b/extensions-builtin/LDSR/preload.py @@ -0,0 +1,6 @@ +import os
+from modules import paths
+
+
+def preload(parser):
+ parser.add_argument("--ldsr-models-path", type=str, help="Path to directory with LDSR model file(s).", default=os.path.join(paths.models_path, 'LDSR'))
diff --git a/modules/ldsr_model.py b/extensions-builtin/LDSR/scripts/ldsr_model.py index 8c4db44a..5c96037d 100644 --- a/modules/ldsr_model.py +++ b/extensions-builtin/LDSR/scripts/ldsr_model.py @@ -5,8 +5,9 @@ import traceback from basicsr.utils.download_util import load_file_from_url from modules.upscaler import Upscaler, UpscalerData -from modules.ldsr_model_arch import LDSR -from modules import shared +from ldsr_model_arch import LDSR +from modules import shared, script_callbacks +import sd_hijack_autoencoder, sd_hijack_ddpm_v1 class UpscalerLDSR(Upscaler): @@ -52,3 +53,12 @@ class UpscalerLDSR(Upscaler): return img ddim_steps = shared.opts.ldsr_steps return ldsr.super_resolution(img, ddim_steps, self.scale) + + +def on_ui_settings(): + import gradio as gr + + shared.opts.add_option("ldsr_steps", shared.OptionInfo(100, "LDSR processing steps. Lower = faster", gr.Slider, {"minimum": 1, "maximum": 200, "step": 1}, section=('upscaling', "Upscaling"))) + + +script_callbacks.on_ui_settings(on_ui_settings) diff --git a/extensions-builtin/LDSR/sd_hijack_autoencoder.py b/extensions-builtin/LDSR/sd_hijack_autoencoder.py new file mode 100644 index 00000000..8e03c7f8 --- /dev/null +++ b/extensions-builtin/LDSR/sd_hijack_autoencoder.py @@ -0,0 +1,286 @@ +# The content of this file comes from the ldm/models/autoencoder.py file of the compvis/stable-diffusion repo +# The VQModel & VQModelInterface were subsequently removed from ldm/models/autoencoder.py when we moved to the stability-ai/stablediffusion repo +# As the LDSR upscaler relies on VQModel & VQModelInterface, the hijack aims to put them back into the ldm.models.autoencoder + +import torch +import pytorch_lightning as pl +import torch.nn.functional as F +from contextlib import contextmanager +from taming.modules.vqvae.quantize import VectorQuantizer2 as VectorQuantizer +from ldm.modules.diffusionmodules.model import Encoder, Decoder +from ldm.util import instantiate_from_config + +import ldm.models.autoencoder + +class VQModel(pl.LightningModule): + def __init__(self, + ddconfig, + lossconfig, + n_embed, + embed_dim, + ckpt_path=None, + ignore_keys=[], + image_key="image", + colorize_nlabels=None, + monitor=None, + batch_resize_range=None, + scheduler_config=None, + lr_g_factor=1.0, + remap=None, + sane_index_shape=False, # tell vector quantizer to return indices as bhw + use_ema=False + ): + super().__init__() + self.embed_dim = embed_dim + self.n_embed = n_embed + self.image_key = image_key + self.encoder = Encoder(**ddconfig) + self.decoder = Decoder(**ddconfig) + self.loss = instantiate_from_config(lossconfig) + self.quantize = VectorQuantizer(n_embed, embed_dim, beta=0.25, + remap=remap, + sane_index_shape=sane_index_shape) + self.quant_conv = torch.nn.Conv2d(ddconfig["z_channels"], embed_dim, 1) + self.post_quant_conv = torch.nn.Conv2d(embed_dim, ddconfig["z_channels"], 1) + if colorize_nlabels is not None: + assert type(colorize_nlabels)==int + self.register_buffer("colorize", torch.randn(3, colorize_nlabels, 1, 1)) + if monitor is not None: + self.monitor = monitor + self.batch_resize_range = batch_resize_range + if self.batch_resize_range is not None: + print(f"{self.__class__.__name__}: Using per-batch resizing in range {batch_resize_range}.") + + self.use_ema = use_ema + if self.use_ema: + self.model_ema = LitEma(self) + print(f"Keeping EMAs of {len(list(self.model_ema.buffers()))}.") + + if ckpt_path is not None: + self.init_from_ckpt(ckpt_path, ignore_keys=ignore_keys) + self.scheduler_config = scheduler_config + self.lr_g_factor = lr_g_factor + + @contextmanager + def ema_scope(self, context=None): + if self.use_ema: + self.model_ema.store(self.parameters()) + self.model_ema.copy_to(self) + if context is not None: + print(f"{context}: Switched to EMA weights") + try: + yield None + finally: + if self.use_ema: + self.model_ema.restore(self.parameters()) + if context is not None: + print(f"{context}: Restored training weights") + + def init_from_ckpt(self, path, ignore_keys=list()): + sd = torch.load(path, map_location="cpu")["state_dict"] + 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) + print(f"Restored from {path} with {len(missing)} missing and {len(unexpected)} unexpected keys") + if len(missing) > 0: + print(f"Missing Keys: {missing}") + print(f"Unexpected Keys: {unexpected}") + + def on_train_batch_end(self, *args, **kwargs): + if self.use_ema: + self.model_ema(self) + + def encode(self, x): + h = self.encoder(x) + h = self.quant_conv(h) + quant, emb_loss, info = self.quantize(h) + return quant, emb_loss, info + + def encode_to_prequant(self, x): + h = self.encoder(x) + h = self.quant_conv(h) + return h + + def decode(self, quant): + quant = self.post_quant_conv(quant) + dec = self.decoder(quant) + return dec + + def decode_code(self, code_b): + quant_b = self.quantize.embed_code(code_b) + dec = self.decode(quant_b) + return dec + + def forward(self, input, return_pred_indices=False): + quant, diff, (_,_,ind) = self.encode(input) + dec = self.decode(quant) + if return_pred_indices: + return dec, diff, ind + return dec, diff + + def get_input(self, batch, k): + x = batch[k] + if len(x.shape) == 3: + x = x[..., None] + x = x.permute(0, 3, 1, 2).to(memory_format=torch.contiguous_format).float() + if self.batch_resize_range is not None: + lower_size = self.batch_resize_range[0] + upper_size = self.batch_resize_range[1] + if self.global_step <= 4: + # do the first few batches with max size to avoid later oom + new_resize = upper_size + else: + new_resize = np.random.choice(np.arange(lower_size, upper_size+16, 16)) + if new_resize != x.shape[2]: + x = F.interpolate(x, size=new_resize, mode="bicubic") + x = x.detach() + return x + + def training_step(self, batch, batch_idx, optimizer_idx): + # https://github.com/pytorch/pytorch/issues/37142 + # try not to fool the heuristics + x = self.get_input(batch, self.image_key) + xrec, qloss, ind = self(x, return_pred_indices=True) + + if optimizer_idx == 0: + # autoencode + aeloss, log_dict_ae = self.loss(qloss, x, xrec, optimizer_idx, self.global_step, + last_layer=self.get_last_layer(), split="train", + predicted_indices=ind) + + self.log_dict(log_dict_ae, prog_bar=False, logger=True, on_step=True, on_epoch=True) + return aeloss + + if optimizer_idx == 1: + # discriminator + discloss, log_dict_disc = self.loss(qloss, x, xrec, optimizer_idx, self.global_step, + last_layer=self.get_last_layer(), split="train") + self.log_dict(log_dict_disc, prog_bar=False, logger=True, on_step=True, on_epoch=True) + return discloss + + def validation_step(self, batch, batch_idx): + log_dict = self._validation_step(batch, batch_idx) + with self.ema_scope(): + log_dict_ema = self._validation_step(batch, batch_idx, suffix="_ema") + return log_dict + + def _validation_step(self, batch, batch_idx, suffix=""): + x = self.get_input(batch, self.image_key) + xrec, qloss, ind = self(x, return_pred_indices=True) + aeloss, log_dict_ae = self.loss(qloss, x, xrec, 0, + self.global_step, + last_layer=self.get_last_layer(), + split="val"+suffix, + predicted_indices=ind + ) + + discloss, log_dict_disc = self.loss(qloss, x, xrec, 1, + self.global_step, + last_layer=self.get_last_layer(), + split="val"+suffix, + predicted_indices=ind + ) + rec_loss = log_dict_ae[f"val{suffix}/rec_loss"] + self.log(f"val{suffix}/rec_loss", rec_loss, + prog_bar=True, logger=True, on_step=False, on_epoch=True, sync_dist=True) + self.log(f"val{suffix}/aeloss", aeloss, + prog_bar=True, logger=True, on_step=False, on_epoch=True, sync_dist=True) + if version.parse(pl.__version__) >= version.parse('1.4.0'): + del log_dict_ae[f"val{suffix}/rec_loss"] + self.log_dict(log_dict_ae) + self.log_dict(log_dict_disc) + return self.log_dict + + def configure_optimizers(self): + lr_d = self.learning_rate + lr_g = self.lr_g_factor*self.learning_rate + print("lr_d", lr_d) + print("lr_g", lr_g) + opt_ae = torch.optim.Adam(list(self.encoder.parameters())+ + list(self.decoder.parameters())+ + list(self.quantize.parameters())+ + list(self.quant_conv.parameters())+ + list(self.post_quant_conv.parameters()), + lr=lr_g, betas=(0.5, 0.9)) + opt_disc = torch.optim.Adam(self.loss.discriminator.parameters(), + lr=lr_d, betas=(0.5, 0.9)) + + if self.scheduler_config is not None: + scheduler = instantiate_from_config(self.scheduler_config) + + print("Setting up LambdaLR scheduler...") + scheduler = [ + { + 'scheduler': LambdaLR(opt_ae, lr_lambda=scheduler.schedule), + 'interval': 'step', + 'frequency': 1 + }, + { + 'scheduler': LambdaLR(opt_disc, lr_lambda=scheduler.schedule), + 'interval': 'step', + 'frequency': 1 + }, + ] + return [opt_ae, opt_disc], scheduler + return [opt_ae, opt_disc], [] + + def get_last_layer(self): + return self.decoder.conv_out.weight + + def log_images(self, batch, only_inputs=False, plot_ema=False, **kwargs): + log = dict() + x = self.get_input(batch, self.image_key) + x = x.to(self.device) + if only_inputs: + log["inputs"] = x + return log + xrec, _ = self(x) + if x.shape[1] > 3: + # colorize with random projection + assert xrec.shape[1] > 3 + x = self.to_rgb(x) + xrec = self.to_rgb(xrec) + log["inputs"] = x + log["reconstructions"] = xrec + if plot_ema: + with self.ema_scope(): + xrec_ema, _ = self(x) + if x.shape[1] > 3: xrec_ema = self.to_rgb(xrec_ema) + log["reconstructions_ema"] = xrec_ema + return log + + def to_rgb(self, x): + assert self.image_key == "segmentation" + if not hasattr(self, "colorize"): + self.register_buffer("colorize", torch.randn(3, x.shape[1], 1, 1).to(x)) + x = F.conv2d(x, weight=self.colorize) + x = 2.*(x-x.min())/(x.max()-x.min()) - 1. + return x + + +class VQModelInterface(VQModel): + def __init__(self, embed_dim, *args, **kwargs): + super().__init__(embed_dim=embed_dim, *args, **kwargs) + self.embed_dim = embed_dim + + def encode(self, x): + h = self.encoder(x) + h = self.quant_conv(h) + return h + + def decode(self, h, force_not_quantize=False): + # also go through quantization layer + if not force_not_quantize: + quant, emb_loss, info = self.quantize(h) + else: + quant = h + quant = self.post_quant_conv(quant) + dec = self.decoder(quant) + return dec + +setattr(ldm.models.autoencoder, "VQModel", VQModel) +setattr(ldm.models.autoencoder, "VQModelInterface", VQModelInterface) diff --git a/extensions-builtin/LDSR/sd_hijack_ddpm_v1.py b/extensions-builtin/LDSR/sd_hijack_ddpm_v1.py new file mode 100644 index 00000000..5c0488e5 --- /dev/null +++ b/extensions-builtin/LDSR/sd_hijack_ddpm_v1.py @@ -0,0 +1,1449 @@ +# This script is copied from the compvis/stable-diffusion repo (aka the SD V1 repo) +# Original filename: ldm/models/diffusion/ddpm.py +# The purpose to reinstate the old DDPM logic which works with VQ, whereas the V2 one doesn't +# Some models such as LDSR require VQ to work correctly +# The classes are suffixed with "V1" and added back to the "ldm.models.diffusion.ddpm" module + +import torch +import torch.nn as nn +import numpy as np +import pytorch_lightning as pl +from torch.optim.lr_scheduler import LambdaLR +from einops import rearrange, repeat +from contextlib import contextmanager +from functools import partial +from tqdm import tqdm +from torchvision.utils import make_grid +from pytorch_lightning.utilities.distributed import rank_zero_only + +from ldm.util import log_txt_as_img, exists, default, ismap, isimage, mean_flat, count_params, instantiate_from_config +from ldm.modules.ema import LitEma +from ldm.modules.distributions.distributions import normal_kl, DiagonalGaussianDistribution +from ldm.models.autoencoder import VQModelInterface, IdentityFirstStage, AutoencoderKL +from ldm.modules.diffusionmodules.util import make_beta_schedule, extract_into_tensor, noise_like +from ldm.models.diffusion.ddim import DDIMSampler + +import ldm.models.diffusion.ddpm + +__conditioning_keys__ = {'concat': 'c_concat', + 'crossattn': 'c_crossattn', + 'adm': 'y'} + + +def disabled_train(self, mode=True): + """Overwrite model.train with this function to make sure train/eval mode + does not change anymore.""" + return self + + +def uniform_on_device(r1, r2, shape, device): + return (r1 - r2) * torch.rand(*shape, device=device) + r2 + + +class DDPMV1(pl.LightningModule): + # classic DDPM with Gaussian diffusion, in image space + def __init__(self, + unet_config, + timesteps=1000, + beta_schedule="linear", + loss_type="l2", + ckpt_path=None, + ignore_keys=[], + load_only_unet=False, + monitor="val/loss", + use_ema=True, + first_stage_key="image", + image_size=256, + channels=3, + log_every_t=100, + clip_denoised=True, + linear_start=1e-4, + linear_end=2e-2, + cosine_s=8e-3, + given_betas=None, + original_elbo_weight=0., + v_posterior=0., # weight for choosing posterior variance as sigma = (1-v) * beta_tilde + v * beta + l_simple_weight=1., + conditioning_key=None, + parameterization="eps", # all assuming fixed variance schedules + scheduler_config=None, + use_positional_encodings=False, + learn_logvar=False, + logvar_init=0., + ): + super().__init__() + assert parameterization in ["eps", "x0"], 'currently only supporting "eps" and "x0"' + self.parameterization = parameterization + print(f"{self.__class__.__name__}: Running in {self.parameterization}-prediction mode") + self.cond_stage_model = None + self.clip_denoised = clip_denoised + self.log_every_t = log_every_t + self.first_stage_key = first_stage_key + self.image_size = image_size # try conv? + self.channels = channels + self.use_positional_encodings = use_positional_encodings + self.model = DiffusionWrapperV1(unet_config, conditioning_key) + count_params(self.model, verbose=True) + self.use_ema = use_ema + if self.use_ema: + self.model_ema = LitEma(self.model) + print(f"Keeping EMAs of {len(list(self.model_ema.buffers()))}.") + + self.use_scheduler = scheduler_config is not None + if self.use_scheduler: + self.scheduler_config = scheduler_config + + self.v_posterior = v_posterior + self.original_elbo_weight = original_elbo_weight + self.l_simple_weight = l_simple_weight + + if monitor is not None: + self.monitor = monitor + if ckpt_path is not None: + self.init_from_ckpt(ckpt_path, ignore_keys=ignore_keys, only_model=load_only_unet) + + self.register_schedule(given_betas=given_betas, beta_schedule=beta_schedule, timesteps=timesteps, + linear_start=linear_start, linear_end=linear_end, cosine_s=cosine_s) + + self.loss_type = loss_type + + self.learn_logvar = learn_logvar + self.logvar = torch.full(fill_value=logvar_init, size=(self.num_timesteps,)) + if self.learn_logvar: + self.logvar = nn.Parameter(self.logvar, requires_grad=True) + + + def register_schedule(self, given_betas=None, beta_schedule="linear", timesteps=1000, + linear_start=1e-4, linear_end=2e-2, cosine_s=8e-3): + if exists(given_betas): + betas = given_betas + else: + betas = make_beta_schedule(beta_schedule, timesteps, linear_start=linear_start, linear_end=linear_end, + cosine_s=cosine_s) + alphas = 1. - betas + alphas_cumprod = np.cumprod(alphas, axis=0) + alphas_cumprod_prev = np.append(1., alphas_cumprod[:-1]) + + timesteps, = betas.shape + self.num_timesteps = int(timesteps) + self.linear_start = linear_start + self.linear_end = linear_end + assert alphas_cumprod.shape[0] == self.num_timesteps, 'alphas have to be defined for each timestep' + + to_torch = partial(torch.tensor, dtype=torch.float32) + + self.register_buffer('betas', to_torch(betas)) + self.register_buffer('alphas_cumprod', to_torch(alphas_cumprod)) + self.register_buffer('alphas_cumprod_prev', to_torch(alphas_cumprod_prev)) + + # calculations for diffusion q(x_t | x_{t-1}) and others + self.register_buffer('sqrt_alphas_cumprod', to_torch(np.sqrt(alphas_cumprod))) + self.register_buffer('sqrt_one_minus_alphas_cumprod', to_torch(np.sqrt(1. - alphas_cumprod))) + self.register_buffer('log_one_minus_alphas_cumprod', to_torch(np.log(1. - alphas_cumprod))) + self.register_buffer('sqrt_recip_alphas_cumprod', to_torch(np.sqrt(1. / alphas_cumprod))) + self.register_buffer('sqrt_recipm1_alphas_cumprod', to_torch(np.sqrt(1. / alphas_cumprod - 1))) + + # calculations for posterior q(x_{t-1} | x_t, x_0) + posterior_variance = (1 - self.v_posterior) * betas * (1. - alphas_cumprod_prev) / ( + 1. - alphas_cumprod) + self.v_posterior * betas + # above: equal to 1. / (1. / (1. - alpha_cumprod_tm1) + alpha_t / beta_t) + self.register_buffer('posterior_variance', to_torch(posterior_variance)) + # below: log calculation clipped because the posterior variance is 0 at the beginning of the diffusion chain + self.register_buffer('posterior_log_variance_clipped', to_torch(np.log(np.maximum(posterior_variance, 1e-20)))) + self.register_buffer('posterior_mean_coef1', to_torch( + betas * np.sqrt(alphas_cumprod_prev) / (1. - alphas_cumprod))) + self.register_buffer('posterior_mean_coef2', to_torch( + (1. - alphas_cumprod_prev) * np.sqrt(alphas) / (1. - alphas_cumprod))) + + if self.parameterization == "eps": + lvlb_weights = self.betas ** 2 / ( + 2 * self.posterior_variance * to_torch(alphas) * (1 - self.alphas_cumprod)) + elif self.parameterization == "x0": + lvlb_weights = 0.5 * np.sqrt(torch.Tensor(alphas_cumprod)) / (2. * 1 - torch.Tensor(alphas_cumprod)) + else: + raise NotImplementedError("mu not supported") + # TODO how to choose this term + lvlb_weights[0] = lvlb_weights[1] + self.register_buffer('lvlb_weights', lvlb_weights, persistent=False) + assert not torch.isnan(self.lvlb_weights).all() + + @contextmanager + def ema_scope(self, context=None): + if self.use_ema: + self.model_ema.store(self.model.parameters()) + self.model_ema.copy_to(self.model) + if context is not None: + print(f"{context}: Switched to EMA weights") + try: + yield None + finally: + if self.use_ema: + self.model_ema.restore(self.model.parameters()) + if context is not None: + print(f"{context}: Restored training weights") + + def init_from_ckpt(self, path, ignore_keys=list(), only_model=False): + 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: + 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) + print(f"Restored from {path} with {len(missing)} missing and {len(unexpected)} unexpected keys") + if len(missing) > 0: + print(f"Missing Keys: {missing}") + if len(unexpected) > 0: + print(f"Unexpected Keys: {unexpected}") + + def q_mean_variance(self, x_start, t): + """ + Get the distribution q(x_t | x_0). + :param x_start: the [N x C x ...] tensor of noiseless inputs. + :param t: the number of diffusion steps (minus 1). Here, 0 means one step. + :return: A tuple (mean, variance, log_variance), all of x_start's shape. + """ + mean = (extract_into_tensor(self.sqrt_alphas_cumprod, t, x_start.shape) * x_start) + variance = extract_into_tensor(1.0 - self.alphas_cumprod, t, x_start.shape) + log_variance = extract_into_tensor(self.log_one_minus_alphas_cumprod, t, x_start.shape) + return mean, variance, log_variance + + def predict_start_from_noise(self, x_t, t, noise): + return ( + extract_into_tensor(self.sqrt_recip_alphas_cumprod, t, x_t.shape) * x_t - + extract_into_tensor(self.sqrt_recipm1_alphas_cumprod, t, x_t.shape) * noise + ) + + def q_posterior(self, x_start, x_t, t): + posterior_mean = ( + extract_into_tensor(self.posterior_mean_coef1, t, x_t.shape) * x_start + + extract_into_tensor(self.posterior_mean_coef2, t, x_t.shape) * x_t + ) + posterior_variance = extract_into_tensor(self.posterior_variance, t, x_t.shape) + posterior_log_variance_clipped = extract_into_tensor(self.posterior_log_variance_clipped, t, x_t.shape) + return posterior_mean, posterior_variance, posterior_log_variance_clipped + + def p_mean_variance(self, x, t, clip_denoised: bool): + model_out = self.model(x, t) + if self.parameterization == "eps": + x_recon = self.predict_start_from_noise(x, t=t, noise=model_out) + elif self.parameterization == "x0": + x_recon = model_out + if clip_denoised: + x_recon.clamp_(-1., 1.) + + model_mean, posterior_variance, posterior_log_variance = self.q_posterior(x_start=x_recon, x_t=x, t=t) + return model_mean, posterior_variance, posterior_log_variance + + @torch.no_grad() + def p_sample(self, x, t, clip_denoised=True, repeat_noise=False): + b, *_, device = *x.shape, x.device + model_mean, _, model_log_variance = self.p_mean_variance(x=x, t=t, clip_denoised=clip_denoised) + noise = noise_like(x.shape, device, repeat_noise) + # no noise when t == 0 + nonzero_mask = (1 - (t == 0).float()).reshape(b, *((1,) * (len(x.shape) - 1))) + return model_mean + nonzero_mask * (0.5 * model_log_variance).exp() * noise + + @torch.no_grad() + def p_sample_loop(self, shape, return_intermediates=False): + device = self.betas.device + b = shape[0] + img = torch.randn(shape, device=device) + intermediates = [img] + for i in tqdm(reversed(range(0, self.num_timesteps)), desc='Sampling t', total=self.num_timesteps): + img = self.p_sample(img, torch.full((b,), i, device=device, dtype=torch.long), + clip_denoised=self.clip_denoised) + if i % self.log_every_t == 0 or i == self.num_timesteps - 1: + intermediates.append(img) + if return_intermediates: + return img, intermediates + return img + + @torch.no_grad() + def sample(self, batch_size=16, return_intermediates=False): + image_size = self.image_size + channels = self.channels + return self.p_sample_loop((batch_size, channels, image_size, image_size), + return_intermediates=return_intermediates) + + def q_sample(self, x_start, t, noise=None): + noise = default(noise, lambda: torch.randn_like(x_start)) + return (extract_into_tensor(self.sqrt_alphas_cumprod, t, x_start.shape) * x_start + + extract_into_tensor(self.sqrt_one_minus_alphas_cumprod, t, x_start.shape) * noise) + + def get_loss(self, pred, target, mean=True): + if self.loss_type == 'l1': + loss = (target - pred).abs() + if mean: + loss = loss.mean() + elif self.loss_type == 'l2': + if mean: + loss = torch.nn.functional.mse_loss(target, pred) + else: + loss = torch.nn.functional.mse_loss(target, pred, reduction='none') + else: + raise NotImplementedError("unknown loss type '{loss_type}'") + + return loss + + def p_losses(self, x_start, t, noise=None): + noise = default(noise, lambda: torch.randn_like(x_start)) + x_noisy = self.q_sample(x_start=x_start, t=t, noise=noise) + model_out = self.model(x_noisy, t) + + loss_dict = {} + if self.parameterization == "eps": + target = noise + elif self.parameterization == "x0": + target = x_start + else: + raise NotImplementedError(f"Paramterization {self.parameterization} not yet supported") + + loss = self.get_loss(model_out, target, mean=False).mean(dim=[1, 2, 3]) + + log_prefix = 'train' if self.training else 'val' + + loss_dict.update({f'{log_prefix}/loss_simple': loss.mean()}) + loss_simple = loss.mean() * self.l_simple_weight + + loss_vlb = (self.lvlb_weights[t] * loss).mean() + loss_dict.update({f'{log_prefix}/loss_vlb': loss_vlb}) + + loss = loss_simple + self.original_elbo_weight * loss_vlb + + loss_dict.update({f'{log_prefix}/loss': loss}) + + return loss, loss_dict + + def forward(self, x, *args, **kwargs): + # b, c, h, w, device, img_size, = *x.shape, x.device, self.image_size + # assert h == img_size and w == img_size, f'height and width of image must be {img_size}' + t = torch.randint(0, self.num_timesteps, (x.shape[0],), device=self.device).long() + return self.p_losses(x, t, *args, **kwargs) + + def get_input(self, batch, k): + x = batch[k] + if len(x.shape) == 3: + x = x[..., None] + x = rearrange(x, 'b h w c -> b c h w') + x = x.to(memory_format=torch.contiguous_format).float() + return x + + def shared_step(self, batch): + x = self.get_input(batch, self.first_stage_key) + loss, loss_dict = self(x) + return loss, loss_dict + + def training_step(self, batch, batch_idx): + loss, loss_dict = self.shared_step(batch) + + self.log_dict(loss_dict, prog_bar=True, + logger=True, on_step=True, on_epoch=True) + + self.log("global_step", self.global_step, + prog_bar=True, logger=True, on_step=True, on_epoch=False) + + if self.use_scheduler: + lr = self.optimizers().param_groups[0]['lr'] + self.log('lr_abs', lr, prog_bar=True, logger=True, on_step=True, on_epoch=False) + + return loss + + @torch.no_grad() + def validation_step(self, batch, batch_idx): + _, loss_dict_no_ema = self.shared_step(batch) + with self.ema_scope(): + _, loss_dict_ema = self.shared_step(batch) + loss_dict_ema = {key + '_ema': loss_dict_ema[key] for key in loss_dict_ema} + self.log_dict(loss_dict_no_ema, prog_bar=False, logger=True, on_step=False, on_epoch=True) + self.log_dict(loss_dict_ema, prog_bar=False, logger=True, on_step=False, on_epoch=True) + + def on_train_batch_end(self, *args, **kwargs): + if self.use_ema: + self.model_ema(self.model) + + def _get_rows_from_list(self, samples): + n_imgs_per_row = len(samples) + denoise_grid = rearrange(samples, 'n b c h w -> b n c h w') + denoise_grid = rearrange(denoise_grid, 'b n c h w -> (b n) c h w') + denoise_grid = make_grid(denoise_grid, nrow=n_imgs_per_row) + return denoise_grid + + @torch.no_grad() + def log_images(self, batch, N=8, n_row=2, sample=True, return_keys=None, **kwargs): + log = dict() + x = self.get_input(batch, self.first_stage_key) + N = min(x.shape[0], N) + n_row = min(x.shape[0], n_row) + x = x.to(self.device)[:N] + log["inputs"] = x + + # get diffusion row + diffusion_row = list() + x_start = x[: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(x_start) + x_noisy = self.q_sample(x_start=x_start, t=t, noise=noise) + diffusion_row.append(x_noisy) + + log["diffusion_row"] = self._get_rows_from_list(diffusion_row) + + if sample: + # get denoise row + with self.ema_scope("Plotting"): + samples, denoise_row = self.sample(batch_size=N, return_intermediates=True) + + log["samples"] = samples + log["denoise_row"] = self._get_rows_from_list(denoise_row) + + if return_keys: + if np.intersect1d(list(log.keys()), return_keys).shape[0] == 0: + return log + else: + return {key: log[key] for key in return_keys} + return log + + def configure_optimizers(self): + lr = self.learning_rate + params = list(self.model.parameters()) + if self.learn_logvar: + params = params + [self.logvar] + opt = torch.optim.AdamW(params, lr=lr) + return opt + + +class LatentDiffusionV1(DDPMV1): + """main class""" + def __init__(self, + first_stage_config, + cond_stage_config, + num_timesteps_cond=None, + cond_stage_key="image", + cond_stage_trainable=False, + concat_mode=True, + cond_stage_forward=None, + conditioning_key=None, + scale_factor=1.0, + scale_by_std=False, + *args, **kwargs): + self.num_timesteps_cond = default(num_timesteps_cond, 1) + self.scale_by_std = scale_by_std + assert self.num_timesteps_cond <= kwargs['timesteps'] + # for backwards compatibility after implementation of DiffusionWrapper + if conditioning_key is None: + conditioning_key = 'concat' if concat_mode else 'crossattn' + if cond_stage_config == '__is_unconditional__': + conditioning_key = None + ckpt_path = kwargs.pop("ckpt_path", None) + ignore_keys = kwargs.pop("ignore_keys", []) + super().__init__(conditioning_key=conditioning_key, *args, **kwargs) + self.concat_mode = concat_mode + self.cond_stage_trainable = cond_stage_trainable + self.cond_stage_key = cond_stage_key + try: + self.num_downs = len(first_stage_config.params.ddconfig.ch_mult) - 1 + except: + self.num_downs = 0 + if not scale_by_std: + self.scale_factor = scale_factor + else: + self.register_buffer('scale_factor', torch.tensor(scale_factor)) + self.instantiate_first_stage(first_stage_config) + self.instantiate_cond_stage(cond_stage_config) + self.cond_stage_forward = cond_stage_forward + self.clip_denoised = False + self.bbox_tokenizer = None + + self.restarted_from_ckpt = False + if ckpt_path is not None: + self.init_from_ckpt(ckpt_path, ignore_keys) + self.restarted_from_ckpt = True + + def make_cond_schedule(self, ): + self.cond_ids = torch.full(size=(self.num_timesteps,), fill_value=self.num_timesteps - 1, dtype=torch.long) + ids = torch.round(torch.linspace(0, self.num_timesteps - 1, self.num_timesteps_cond)).long() + self.cond_ids[:self.num_timesteps_cond] = ids + + @rank_zero_only + @torch.no_grad() + def on_train_batch_start(self, batch, batch_idx, dataloader_idx): + # only for very first batch + if self.scale_by_std and self.current_epoch == 0 and self.global_step == 0 and batch_idx == 0 and not self.restarted_from_ckpt: + assert self.scale_factor == 1., 'rather not use custom rescaling and std-rescaling simultaneously' + # set rescale weight to 1./std of encodings + print("### USING STD-RESCALING ###") + x = super().get_input(batch, self.first_stage_key) + x = x.to(self.device) + encoder_posterior = self.encode_first_stage(x) + z = self.get_first_stage_encoding(encoder_posterior).detach() + del self.scale_factor + self.register_buffer('scale_factor', 1. / z.flatten().std()) + print(f"setting self.scale_factor to {self.scale_factor}") + print("### USING STD-RESCALING ###") + + def register_schedule(self, + given_betas=None, beta_schedule="linear", timesteps=1000, + linear_start=1e-4, linear_end=2e-2, cosine_s=8e-3): + super().register_schedule(given_betas, beta_schedule, timesteps, linear_start, linear_end, cosine_s) + + self.shorten_cond_schedule = self.num_timesteps_cond > 1 + if self.shorten_cond_schedule: + self.make_cond_schedule() + + def instantiate_first_stage(self, config): + model = instantiate_from_config(config) + self.first_stage_model = model.eval() + self.first_stage_model.train = disabled_train + for param in self.first_stage_model.parameters(): + param.requires_grad = False + + def instantiate_cond_stage(self, config): + if not self.cond_stage_trainable: + if config == "__is_first_stage__": + print("Using first stage also as cond stage.") + self.cond_stage_model = self.first_stage_model + elif config == "__is_unconditional__": + print(f"Training {self.__class__.__name__} as an unconditional model.") + self.cond_stage_model = None + # self.be_unconditional = True + else: + model = instantiate_from_config(config) + self.cond_stage_model = model.eval() + self.cond_stage_model.train = disabled_train + for param in self.cond_stage_model.parameters(): + param.requires_grad = False + else: + assert config != '__is_first_stage__' + assert config != '__is_unconditional__' + model = instantiate_from_config(config) + self.cond_stage_model = model + + def _get_denoise_row_from_list(self, samples, desc='', force_no_decoder_quantization=False): + denoise_row = [] + for zd in tqdm(samples, desc=desc): + denoise_row.append(self.decode_first_stage(zd.to(self.device), + force_not_quantize=force_no_decoder_quantization)) + n_imgs_per_row = len(denoise_row) + denoise_row = torch.stack(denoise_row) # n_log_step, n_row, C, H, W + denoise_grid = rearrange(denoise_row, 'n b c h w -> b n c h w') + denoise_grid = rearrange(denoise_grid, 'b n c h w -> (b n) c h w') + denoise_grid = make_grid(denoise_grid, nrow=n_imgs_per_row) + return denoise_grid + + def get_first_stage_encoding(self, encoder_posterior): + if isinstance(encoder_posterior, DiagonalGaussianDistribution): + z = encoder_posterior.sample() + elif isinstance(encoder_posterior, torch.Tensor): + z = encoder_posterior + else: + raise NotImplementedError(f"encoder_posterior of type '{type(encoder_posterior)}' not yet implemented") + return self.scale_factor * z + + def get_learned_conditioning(self, c): + if self.cond_stage_forward is None: + if hasattr(self.cond_stage_model, 'encode') and callable(self.cond_stage_model.encode): + c = self.cond_stage_model.encode(c) + if isinstance(c, DiagonalGaussianDistribution): + c = c.mode() + else: + c = self.cond_stage_model(c) + else: + assert hasattr(self.cond_stage_model, self.cond_stage_forward) + c = getattr(self.cond_stage_model, self.cond_stage_forward)(c) + return c + + def meshgrid(self, h, w): + y = torch.arange(0, h).view(h, 1, 1).repeat(1, w, 1) + x = torch.arange(0, w).view(1, w, 1).repeat(h, 1, 1) + + arr = torch.cat([y, x], dim=-1) + return arr + + def delta_border(self, h, w): + """ + :param h: height + :param w: width + :return: normalized distance to image border, + wtith min distance = 0 at border and max dist = 0.5 at image center + """ + lower_right_corner = torch.tensor([h - 1, w - 1]).view(1, 1, 2) + arr = self.meshgrid(h, w) / lower_right_corner + dist_left_up = torch.min(arr, dim=-1, keepdims=True)[0] + dist_right_down = torch.min(1 - arr, dim=-1, keepdims=True)[0] + edge_dist = torch.min(torch.cat([dist_left_up, dist_right_down], dim=-1), dim=-1)[0] + return edge_dist + + def get_weighting(self, h, w, Ly, Lx, device): + weighting = self.delta_border(h, w) + weighting = torch.clip(weighting, self.split_input_params["clip_min_weight"], + self.split_input_params["clip_max_weight"], ) + weighting = weighting.view(1, h * w, 1).repeat(1, 1, Ly * Lx).to(device) + + if self.split_input_params["tie_braker"]: + L_weighting = self.delta_border(Ly, Lx) + L_weighting = torch.clip(L_weighting, + self.split_input_params["clip_min_tie_weight"], + self.split_input_params["clip_max_tie_weight"]) + + L_weighting = L_weighting.view(1, 1, Ly * Lx).to(device) + weighting = weighting * L_weighting + return weighting + + def get_fold_unfold(self, x, kernel_size, stride, uf=1, df=1): # todo load once not every time, shorten code + """ + :param x: img of size (bs, c, h, w) + :return: n img crops of size (n, bs, c, kernel_size[0], kernel_size[1]) + """ + bs, nc, h, w = x.shape + + # number of crops in image + Ly = (h - kernel_size[0]) // stride[0] + 1 + Lx = (w - kernel_size[1]) // stride[1] + 1 + + if uf == 1 and df == 1: + fold_params = dict(kernel_size=kernel_size, dilation=1, padding=0, stride=stride) + unfold = torch.nn.Unfold(**fold_params) + + fold = torch.nn.Fold(output_size=x.shape[2:], **fold_params) + + weighting = self.get_weighting(kernel_size[0], kernel_size[1], Ly, Lx, x.device).to(x.dtype) + normalization = fold(weighting).view(1, 1, h, w) # normalizes the overlap + weighting = weighting.view((1, 1, kernel_size[0], kernel_size[1], Ly * Lx)) + + elif uf > 1 and df == 1: + fold_params = dict(kernel_size=kernel_size, dilation=1, padding=0, stride=stride) + unfold = torch.nn.Unfold(**fold_params) + + fold_params2 = dict(kernel_size=(kernel_size[0] * uf, kernel_size[0] * uf), + dilation=1, padding=0, + stride=(stride[0] * uf, stride[1] * uf)) + fold = torch.nn.Fold(output_size=(x.shape[2] * uf, x.shape[3] * uf), **fold_params2) + + weighting = self.get_weighting(kernel_size[0] * uf, kernel_size[1] * uf, Ly, Lx, x.device).to(x.dtype) + normalization = fold(weighting).view(1, 1, h * uf, w * uf) # normalizes the overlap + weighting = weighting.view((1, 1, kernel_size[0] * uf, kernel_size[1] * uf, Ly * Lx)) + + elif df > 1 and uf == 1: + fold_params = dict(kernel_size=kernel_size, dilation=1, padding=0, stride=stride) + unfold = torch.nn.Unfold(**fold_params) + + fold_params2 = dict(kernel_size=(kernel_size[0] // df, kernel_size[0] // df), + dilation=1, padding=0, + stride=(stride[0] // df, stride[1] // df)) + fold = torch.nn.Fold(output_size=(x.shape[2] // df, x.shape[3] // df), **fold_params2) + + weighting = self.get_weighting(kernel_size[0] // df, kernel_size[1] // df, Ly, Lx, x.device).to(x.dtype) + normalization = fold(weighting).view(1, 1, h // df, w // df) # normalizes the overlap + weighting = weighting.view((1, 1, kernel_size[0] // df, kernel_size[1] // df, Ly * Lx)) + + else: + raise NotImplementedError + + return fold, unfold, normalization, weighting + + @torch.no_grad() + def get_input(self, batch, k, return_first_stage_outputs=False, force_c_encode=False, + cond_key=None, return_original_cond=False, bs=None): + x = super().get_input(batch, k) + if bs is not None: + x = x[:bs] + x = x.to(self.device) + encoder_posterior = self.encode_first_stage(x) + z = self.get_first_stage_encoding(encoder_posterior).detach() + + if self.model.conditioning_key is not None: + if cond_key is None: + cond_key = self.cond_stage_key + if cond_key != self.first_stage_key: + if cond_key in ['caption', 'coordinates_bbox']: + xc = batch[cond_key] + elif cond_key == 'class_label': + xc = batch + else: + xc = super().get_input(batch, cond_key).to(self.device) + else: + xc = x + if not self.cond_stage_trainable or force_c_encode: + if isinstance(xc, dict) or isinstance(xc, list): + # import pudb; pudb.set_trace() + c = self.get_learned_conditioning(xc) + else: + c = self.get_learned_conditioning(xc.to(self.device)) + else: + c = xc + if bs is not None: + c = c[:bs] + + if self.use_positional_encodings: + pos_x, pos_y = self.compute_latent_shifts(batch) + ckey = __conditioning_keys__[self.model.conditioning_key] + c = {ckey: c, 'pos_x': pos_x, 'pos_y': pos_y} + + else: + c = None + xc = None + if self.use_positional_encodings: + pos_x, pos_y = self.compute_latent_shifts(batch) + c = {'pos_x': pos_x, 'pos_y': pos_y} + out = [z, c] + if return_first_stage_outputs: + xrec = self.decode_first_stage(z) + out.extend([x, xrec]) + if return_original_cond: + out.append(xc) + return out + + @torch.no_grad() + def decode_first_stage(self, z, predict_cids=False, force_not_quantize=False): + if predict_cids: + if z.dim() == 4: + z = torch.argmax(z.exp(), dim=1).long() + z = self.first_stage_model.quantize.get_codebook_entry(z, shape=None) + z = rearrange(z, 'b h w c -> b c h w').contiguous() + + z = 1. / self.scale_factor * z + + if hasattr(self, "split_input_params"): + if self.split_input_params["patch_distributed_vq"]: + ks = self.split_input_params["ks"] # eg. (128, 128) + stride = self.split_input_params["stride"] # eg. (64, 64) + uf = self.split_input_params["vqf"] + bs, nc, h, w = z.shape + if ks[0] > h or ks[1] > w: + ks = (min(ks[0], h), min(ks[1], w)) + print("reducing Kernel") + + if stride[0] > h or stride[1] > w: + stride = (min(stride[0], h), min(stride[1], w)) + print("reducing stride") + + fold, unfold, normalization, weighting = self.get_fold_unfold(z, ks, stride, uf=uf) + + z = unfold(z) # (bn, nc * prod(**ks), L) + # 1. Reshape to img shape + z = z.view((z.shape[0], -1, ks[0], ks[1], z.shape[-1])) # (bn, nc, ks[0], ks[1], L ) + + # 2. apply model loop over last dim + if isinstance(self.first_stage_model, VQModelInterface): + output_list = [self.first_stage_model.decode(z[:, :, :, :, i], + force_not_quantize=predict_cids or force_not_quantize) + for i in range(z.shape[-1])] + else: + + output_list = [self.first_stage_model.decode(z[:, :, :, :, i]) + for i in range(z.shape[-1])] + + o = torch.stack(output_list, axis=-1) # # (bn, nc, ks[0], ks[1], L) + o = o * weighting + # Reverse 1. reshape to img shape + o = o.view((o.shape[0], -1, o.shape[-1])) # (bn, nc * ks[0] * ks[1], L) + # stitch crops together + decoded = fold(o) + decoded = decoded / normalization # norm is shape (1, 1, h, w) + return decoded + else: + if isinstance(self.first_stage_model, VQModelInterface): + return self.first_stage_model.decode(z, force_not_quantize=predict_cids or force_not_quantize) + else: + return self.first_stage_model.decode(z) + + else: + if isinstance(self.first_stage_model, VQModelInterface): + return self.first_stage_model.decode(z, force_not_quantize=predict_cids or force_not_quantize) + else: + return self.first_stage_model.decode(z) + + # same as above but without decorator + def differentiable_decode_first_stage(self, z, predict_cids=False, force_not_quantize=False): + if predict_cids: + if z.dim() == 4: + z = torch.argmax(z.exp(), dim=1).long() + z = self.first_stage_model.quantize.get_codebook_entry(z, shape=None) + z = rearrange(z, 'b h w c -> b c h w').contiguous() + + z = 1. / self.scale_factor * z + + if hasattr(self, "split_input_params"): + if self.split_input_params["patch_distributed_vq"]: + ks = self.split_input_params["ks"] # eg. (128, 128) + stride = self.split_input_params["stride"] # eg. (64, 64) + uf = self.split_input_params["vqf"] + bs, nc, h, w = z.shape + if ks[0] > h or ks[1] > w: + ks = (min(ks[0], h), min(ks[1], w)) + print("reducing Kernel") + + if stride[0] > h or stride[1] > w: + stride = (min(stride[0], h), min(stride[1], w)) + print("reducing stride") + + fold, unfold, normalization, weighting = self.get_fold_unfold(z, ks, stride, uf=uf) + + z = unfold(z) # (bn, nc * prod(**ks), L) + # 1. Reshape to img shape + z = z.view((z.shape[0], -1, ks[0], ks[1], z.shape[-1])) # (bn, nc, ks[0], ks[1], L ) + + # 2. apply model loop over last dim + if isinstance(self.first_stage_model, VQModelInterface): + output_list = [self.first_stage_model.decode(z[:, :, :, :, i], + force_not_quantize=predict_cids or force_not_quantize) + for i in range(z.shape[-1])] + else: + + output_list = [self.first_stage_model.decode(z[:, :, :, :, i]) + for i in range(z.shape[-1])] + + o = torch.stack(output_list, axis=-1) # # (bn, nc, ks[0], ks[1], L) + o = o * weighting + # Reverse 1. reshape to img shape + o = o.view((o.shape[0], -1, o.shape[-1])) # (bn, nc * ks[0] * ks[1], L) + # stitch crops together + decoded = fold(o) + decoded = decoded / normalization # norm is shape (1, 1, h, w) + return decoded + else: + if isinstance(self.first_stage_model, VQModelInterface): + return self.first_stage_model.decode(z, force_not_quantize=predict_cids or force_not_quantize) + else: + return self.first_stage_model.decode(z) + + else: + if isinstance(self.first_stage_model, VQModelInterface): + return self.first_stage_model.decode(z, force_not_quantize=predict_cids or force_not_quantize) + else: + return self.first_stage_model.decode(z) + + @torch.no_grad() + def encode_first_stage(self, x): + if hasattr(self, "split_input_params"): + if self.split_input_params["patch_distributed_vq"]: + ks = self.split_input_params["ks"] # eg. (128, 128) + stride = self.split_input_params["stride"] # eg. (64, 64) + df = self.split_input_params["vqf"] + self.split_input_params['original_image_size'] = x.shape[-2:] + bs, nc, h, w = x.shape + if ks[0] > h or ks[1] > w: + ks = (min(ks[0], h), min(ks[1], w)) + print("reducing Kernel") + + if stride[0] > h or stride[1] > w: + stride = (min(stride[0], h), min(stride[1], w)) + print("reducing stride") + + fold, unfold, normalization, weighting = self.get_fold_unfold(x, ks, stride, df=df) + z = unfold(x) # (bn, nc * prod(**ks), L) + # Reshape to img shape + z = z.view((z.shape[0], -1, ks[0], ks[1], z.shape[-1])) # (bn, nc, ks[0], ks[1], L ) + + output_list = [self.first_stage_model.encode(z[:, :, :, :, i]) + for i in range(z.shape[-1])] + + o = torch.stack(output_list, axis=-1) + o = o * weighting + + # Reverse reshape to img shape + o = o.view((o.shape[0], -1, o.shape[-1])) # (bn, nc * ks[0] * ks[1], L) + # stitch crops together + decoded = fold(o) + decoded = decoded / normalization + return decoded + + else: + return self.first_stage_model.encode(x) + else: + return self.first_stage_model.encode(x) + + def shared_step(self, batch, **kwargs): + x, c = self.get_input(batch, self.first_stage_key) + loss = self(x, c) + return loss + + def forward(self, x, c, *args, **kwargs): + t = torch.randint(0, self.num_timesteps, (x.shape[0],), device=self.device).long() + if self.model.conditioning_key is not None: + assert c is not None + if self.cond_stage_trainable: + c = self.get_learned_conditioning(c) + if self.shorten_cond_schedule: # TODO: drop this option + tc = self.cond_ids[t].to(self.device) + c = self.q_sample(x_start=c, t=tc, noise=torch.randn_like(c.float())) + return self.p_losses(x, c, t, *args, **kwargs) + + def _rescale_annotations(self, bboxes, crop_coordinates): # TODO: move to dataset + def rescale_bbox(bbox): + x0 = clamp((bbox[0] - crop_coordinates[0]) / crop_coordinates[2]) + y0 = clamp((bbox[1] - crop_coordinates[1]) / crop_coordinates[3]) + w = min(bbox[2] / crop_coordinates[2], 1 - x0) + h = min(bbox[3] / crop_coordinates[3], 1 - y0) + return x0, y0, w, h + + return [rescale_bbox(b) for b in bboxes] + + def apply_model(self, x_noisy, t, cond, return_ids=False): + + if isinstance(cond, dict): + # hybrid case, cond is exptected to be a dict + pass + else: + if not isinstance(cond, list): + cond = [cond] + key = 'c_concat' if self.model.conditioning_key == 'concat' else 'c_crossattn' + cond = {key: cond} + + if hasattr(self, "split_input_params"): + assert len(cond) == 1 # todo can only deal with one conditioning atm + assert not return_ids + ks = self.split_input_params["ks"] # eg. (128, 128) + stride = self.split_input_params["stride"] # eg. (64, 64) + + h, w = x_noisy.shape[-2:] + + fold, unfold, normalization, weighting = self.get_fold_unfold(x_noisy, ks, stride) + + z = unfold(x_noisy) # (bn, nc * prod(**ks), L) + # Reshape to img shape + z = z.view((z.shape[0], -1, ks[0], ks[1], z.shape[-1])) # (bn, nc, ks[0], ks[1], L ) + z_list = [z[:, :, :, :, i] for i in range(z.shape[-1])] + + if self.cond_stage_key in ["image", "LR_image", "segmentation", + 'bbox_img'] and self.model.conditioning_key: # todo check for completeness + c_key = next(iter(cond.keys())) # get key + c = next(iter(cond.values())) # get value + assert (len(c) == 1) # todo extend to list with more than one elem + c = c[0] # get element + + c = unfold(c) + c = c.view((c.shape[0], -1, ks[0], ks[1], c.shape[-1])) # (bn, nc, ks[0], ks[1], L ) + + cond_list = [{c_key: [c[:, :, :, :, i]]} for i in range(c.shape[-1])] + + elif self.cond_stage_key == 'coordinates_bbox': + assert 'original_image_size' in self.split_input_params, 'BoudingBoxRescaling is missing original_image_size' + + # assuming padding of unfold is always 0 and its dilation is always 1 + n_patches_per_row = int((w - ks[0]) / stride[0] + 1) + full_img_h, full_img_w = self.split_input_params['original_image_size'] + # as we are operating on latents, we need the factor from the original image size to the + # spatial latent size to properly rescale the crops for regenerating the bbox annotations + num_downs = self.first_stage_model.encoder.num_resolutions - 1 + rescale_latent = 2 ** (num_downs) + + # get top left postions of patches as conforming for the bbbox tokenizer, therefore we + # need to rescale the tl patch coordinates to be in between (0,1) + tl_patch_coordinates = [(rescale_latent * stride[0] * (patch_nr % n_patches_per_row) / full_img_w, + rescale_latent * stride[1] * (patch_nr // n_patches_per_row) / full_img_h) + for patch_nr in range(z.shape[-1])] + + # patch_limits are tl_coord, width and height coordinates as (x_tl, y_tl, h, w) + patch_limits = [(x_tl, y_tl, + rescale_latent * ks[0] / full_img_w, + rescale_latent * ks[1] / full_img_h) for x_tl, y_tl in tl_patch_coordinates] + # patch_values = [(np.arange(x_tl,min(x_tl+ks, 1.)),np.arange(y_tl,min(y_tl+ks, 1.))) for x_tl, y_tl in tl_patch_coordinates] + + # tokenize crop coordinates for the bounding boxes of the respective patches + patch_limits_tknzd = [torch.LongTensor(self.bbox_tokenizer._crop_encoder(bbox))[None].to(self.device) + for bbox in patch_limits] # list of length l with tensors of shape (1, 2) + print(patch_limits_tknzd[0].shape) + # cut tknzd crop position from conditioning + assert isinstance(cond, dict), 'cond must be dict to be fed into model' + cut_cond = cond['c_crossattn'][0][..., :-2].to(self.device) + print(cut_cond.shape) + + adapted_cond = torch.stack([torch.cat([cut_cond, p], dim=1) for p in patch_limits_tknzd]) + adapted_cond = rearrange(adapted_cond, 'l b n -> (l b) n') + print(adapted_cond.shape) + adapted_cond = self.get_learned_conditioning(adapted_cond) + print(adapted_cond.shape) + adapted_cond = rearrange(adapted_cond, '(l b) n d -> l b n d', l=z.shape[-1]) + print(adapted_cond.shape) + + cond_list = [{'c_crossattn': [e]} for e in adapted_cond] + + else: + cond_list = [cond for i in range(z.shape[-1])] # Todo make this more efficient + + # apply model by loop over crops + output_list = [self.model(z_list[i], t, **cond_list[i]) for i in range(z.shape[-1])] + assert not isinstance(output_list[0], + tuple) # todo cant deal with multiple model outputs check this never happens + + o = torch.stack(output_list, axis=-1) + o = o * weighting + # Reverse reshape to img shape + o = o.view((o.shape[0], -1, o.shape[-1])) # (bn, nc * ks[0] * ks[1], L) + # stitch crops together + x_recon = fold(o) / normalization + + else: + x_recon = self.model(x_noisy, t, **cond) + + if isinstance(x_recon, tuple) and not return_ids: + return x_recon[0] + else: + return x_recon + + def _predict_eps_from_xstart(self, x_t, t, pred_xstart): + return (extract_into_tensor(self.sqrt_recip_alphas_cumprod, t, x_t.shape) * x_t - pred_xstart) / \ + extract_into_tensor(self.sqrt_recipm1_alphas_cumprod, t, x_t.shape) + + def _prior_bpd(self, x_start): + """ + Get the prior KL term for the variational lower-bound, measured in + bits-per-dim. + This term can't be optimized, as it only depends on the encoder. + :param x_start: the [N x C x ...] tensor of inputs. + :return: a batch of [N] KL values (in bits), one per batch element. + """ + batch_size = x_start.shape[0] + t = torch.tensor([self.num_timesteps - 1] * batch_size, device=x_start.device) + qt_mean, _, qt_log_variance = self.q_mean_variance(x_start, t) + kl_prior = normal_kl(mean1=qt_mean, logvar1=qt_log_variance, mean2=0.0, logvar2=0.0) + return mean_flat(kl_prior) / np.log(2.0) + + def p_losses(self, x_start, cond, t, noise=None): + noise = default(noise, lambda: torch.randn_like(x_start)) + x_noisy = self.q_sample(x_start=x_start, t=t, noise=noise) + model_output = self.apply_model(x_noisy, t, cond) + + loss_dict = {} + prefix = 'train' if self.training else 'val' + + if self.parameterization == "x0": + target = x_start + elif self.parameterization == "eps": + target = noise + else: + raise NotImplementedError() + + loss_simple = self.get_loss(model_output, target, mean=False).mean([1, 2, 3]) + loss_dict.update({f'{prefix}/loss_simple': loss_simple.mean()}) + + logvar_t = self.logvar[t].to(self.device) + loss = loss_simple / torch.exp(logvar_t) + logvar_t + # loss = loss_simple / torch.exp(self.logvar) + self.logvar + if self.learn_logvar: + loss_dict.update({f'{prefix}/loss_gamma': loss.mean()}) + loss_dict.update({'logvar': self.logvar.data.mean()}) + + loss = self.l_simple_weight * loss.mean() + + loss_vlb = self.get_loss(model_output, target, mean=False).mean(dim=(1, 2, 3)) + loss_vlb = (self.lvlb_weights[t] * loss_vlb).mean() + loss_dict.update({f'{prefix}/loss_vlb': loss_vlb}) + loss += (self.original_elbo_weight * loss_vlb) + loss_dict.update({f'{prefix}/loss': loss}) + + return loss, loss_dict + + def p_mean_variance(self, x, c, t, clip_denoised: bool, return_codebook_ids=False, quantize_denoised=False, + return_x0=False, score_corrector=None, corrector_kwargs=None): + t_in = t + model_out = self.apply_model(x, t_in, c, return_ids=return_codebook_ids) + + if score_corrector is not None: + assert self.parameterization == "eps" + model_out = score_corrector.modify_score(self, model_out, x, t, c, **corrector_kwargs) + + if return_codebook_ids: + model_out, logits = model_out + + if self.parameterization == "eps": + x_recon = self.predict_start_from_noise(x, t=t, noise=model_out) + elif self.parameterization == "x0": + x_recon = model_out + else: + raise NotImplementedError() + + if clip_denoised: + x_recon.clamp_(-1., 1.) + if quantize_denoised: + x_recon, _, [_, _, indices] = self.first_stage_model.quantize(x_recon) + model_mean, posterior_variance, posterior_log_variance = self.q_posterior(x_start=x_recon, x_t=x, t=t) + if return_codebook_ids: + return model_mean, posterior_variance, posterior_log_variance, logits + elif return_x0: + return model_mean, posterior_variance, posterior_log_variance, x_recon + else: + return model_mean, posterior_variance, posterior_log_variance + + @torch.no_grad() + def p_sample(self, x, c, t, clip_denoised=False, repeat_noise=False, + return_codebook_ids=False, quantize_denoised=False, return_x0=False, + temperature=1., noise_dropout=0., score_corrector=None, corrector_kwargs=None): + b, *_, device = *x.shape, x.device + outputs = self.p_mean_variance(x=x, c=c, t=t, clip_denoised=clip_denoised, + return_codebook_ids=return_codebook_ids, + quantize_denoised=quantize_denoised, + return_x0=return_x0, + score_corrector=score_corrector, corrector_kwargs=corrector_kwargs) + if return_codebook_ids: + raise DeprecationWarning("Support dropped.") + model_mean, _, model_log_variance, logits = outputs + elif return_x0: + model_mean, _, model_log_variance, x0 = outputs + else: + model_mean, _, model_log_variance = outputs + + noise = noise_like(x.shape, device, repeat_noise) * temperature + if noise_dropout > 0.: + noise = torch.nn.functional.dropout(noise, p=noise_dropout) + # no noise when t == 0 + nonzero_mask = (1 - (t == 0).float()).reshape(b, *((1,) * (len(x.shape) - 1))) + + if return_codebook_ids: + return model_mean + nonzero_mask * (0.5 * model_log_variance).exp() * noise, logits.argmax(dim=1) + if return_x0: + return model_mean + nonzero_mask * (0.5 * model_log_variance).exp() * noise, x0 + else: + return model_mean + nonzero_mask * (0.5 * model_log_variance).exp() * noise + + @torch.no_grad() + def progressive_denoising(self, cond, shape, verbose=True, callback=None, quantize_denoised=False, + img_callback=None, mask=None, x0=None, temperature=1., noise_dropout=0., + score_corrector=None, corrector_kwargs=None, batch_size=None, x_T=None, start_T=None, + log_every_t=None): + if not log_every_t: + log_every_t = self.log_every_t + timesteps = self.num_timesteps + if batch_size is not None: + b = batch_size if batch_size is not None else shape[0] + shape = [batch_size] + list(shape) + else: + b = batch_size = shape[0] + if x_T is None: + img = torch.randn(shape, device=self.device) + else: + img = x_T + intermediates = [] + if cond is not None: + if isinstance(cond, dict): + cond = {key: cond[key][:batch_size] if not isinstance(cond[key], list) else + list(map(lambda x: x[:batch_size], cond[key])) for key in cond} + else: + cond = [c[:batch_size] for c in cond] if isinstance(cond, list) else cond[:batch_size] + + if start_T is not None: + timesteps = min(timesteps, start_T) + iterator = tqdm(reversed(range(0, timesteps)), desc='Progressive Generation', + total=timesteps) if verbose else reversed( + range(0, timesteps)) + if type(temperature) == float: + temperature = [temperature] * timesteps + + for i in iterator: + ts = torch.full((b,), i, device=self.device, dtype=torch.long) + if self.shorten_cond_schedule: + assert self.model.conditioning_key != 'hybrid' + tc = self.cond_ids[ts].to(cond.device) + cond = self.q_sample(x_start=cond, t=tc, noise=torch.randn_like(cond)) + + img, x0_partial = self.p_sample(img, cond, ts, + clip_denoised=self.clip_denoised, + quantize_denoised=quantize_denoised, return_x0=True, + temperature=temperature[i], noise_dropout=noise_dropout, + score_corrector=score_corrector, corrector_kwargs=corrector_kwargs) + if mask is not None: + assert x0 is not None + img_orig = self.q_sample(x0, ts) + img = img_orig * mask + (1. - mask) * img + + if i % log_every_t == 0 or i == timesteps - 1: + intermediates.append(x0_partial) + if callback: callback(i) + if img_callback: img_callback(img, i) + return img, intermediates + + @torch.no_grad() + def p_sample_loop(self, cond, shape, return_intermediates=False, + x_T=None, verbose=True, callback=None, timesteps=None, quantize_denoised=False, + mask=None, x0=None, img_callback=None, start_T=None, + log_every_t=None): + + if not log_every_t: + log_every_t = self.log_every_t + device = self.betas.device + b = shape[0] + if x_T is None: + img = torch.randn(shape, device=device) + else: + img = x_T + + intermediates = [img] + if timesteps is None: + timesteps = self.num_timesteps + + if start_T is not None: + timesteps = min(timesteps, start_T) + iterator = tqdm(reversed(range(0, timesteps)), desc='Sampling t', total=timesteps) if verbose else reversed( + range(0, timesteps)) + + if mask is not None: + assert x0 is not None + assert x0.shape[2:3] == mask.shape[2:3] # spatial size has to match + + for i in iterator: + ts = torch.full((b,), i, device=device, dtype=torch.long) + if self.shorten_cond_schedule: + assert self.model.conditioning_key != 'hybrid' + tc = self.cond_ids[ts].to(cond.device) + cond = self.q_sample(x_start=cond, t=tc, noise=torch.randn_like(cond)) + + img = self.p_sample(img, cond, ts, + clip_denoised=self.clip_denoised, + quantize_denoised=quantize_denoised) + if mask is not None: + img_orig = self.q_sample(x0, ts) + img = img_orig * mask + (1. - mask) * img + + if i % log_every_t == 0 or i == timesteps - 1: + intermediates.append(img) + if callback: callback(i) + if img_callback: img_callback(img, i) + + if return_intermediates: + return img, intermediates + return img + + @torch.no_grad() + def sample(self, cond, batch_size=16, return_intermediates=False, x_T=None, + verbose=True, timesteps=None, quantize_denoised=False, + mask=None, x0=None, shape=None,**kwargs): + if shape is None: + shape = (batch_size, self.channels, self.image_size, self.image_size) + if cond is not None: + if isinstance(cond, dict): + cond = {key: cond[key][:batch_size] if not isinstance(cond[key], list) else + list(map(lambda x: x[:batch_size], cond[key])) for key in cond} + else: + cond = [c[:batch_size] for c in cond] if isinstance(cond, list) else cond[:batch_size] + return self.p_sample_loop(cond, + shape, + return_intermediates=return_intermediates, x_T=x_T, + verbose=verbose, timesteps=timesteps, quantize_denoised=quantize_denoised, + mask=mask, x0=x0) + + @torch.no_grad() + def sample_log(self,cond,batch_size,ddim, ddim_steps,**kwargs): + + if ddim: + ddim_sampler = DDIMSampler(self) + shape = (self.channels, self.image_size, self.image_size) + samples, intermediates =ddim_sampler.sample(ddim_steps,batch_size, + shape,cond,verbose=False,**kwargs) + + else: + samples, intermediates = self.sample(cond=cond, batch_size=batch_size, + return_intermediates=True,**kwargs) + + return samples, intermediates + + + @torch.no_grad() + def log_images(self, batch, N=8, n_row=4, sample=True, ddim_steps=200, ddim_eta=1., return_keys=None, + quantize_denoised=True, inpaint=True, plot_denoise_rows=False, plot_progressive_rows=True, + plot_diffusion_rows=True, **kwargs): + + use_ddim = ddim_steps is not None + + log = dict() + z, c, x, xrec, xc = self.get_input(batch, self.first_stage_key, + return_first_stage_outputs=True, + force_c_encode=True, + return_original_cond=True, + bs=N) + N = min(x.shape[0], N) + n_row = min(x.shape[0], n_row) + log["inputs"] = x + log["reconstruction"] = xrec + if self.model.conditioning_key is not None: + if hasattr(self.cond_stage_model, "decode"): + xc = self.cond_stage_model.decode(c) + log["conditioning"] = xc + elif self.cond_stage_key in ["caption"]: + xc = log_txt_as_img((x.shape[2], x.shape[3]), batch["caption"]) + log["conditioning"] = xc + elif self.cond_stage_key == 'class_label': + xc = log_txt_as_img((x.shape[2], x.shape[3]), batch["human_label"]) + log['conditioning'] = xc + elif isimage(xc): + log["conditioning"] = xc + if ismap(xc): + log["original_conditioning"] = self.to_rgb(xc) + + 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 + with self.ema_scope("Plotting"): + samples, z_denoise_row = self.sample_log(cond=c,batch_size=N,ddim=use_ddim, + ddim_steps=ddim_steps,eta=ddim_eta) + # samples, z_denoise_row = self.sample(cond=c, batch_size=N, return_intermediates=True) + 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 quantize_denoised and not isinstance(self.first_stage_model, AutoencoderKL) and not isinstance( + self.first_stage_model, IdentityFirstStage): + # also display when quantizing x0 while sampling + with self.ema_scope("Plotting Quantized Denoised"): + samples, z_denoise_row = self.sample_log(cond=c,batch_size=N,ddim=use_ddim, + ddim_steps=ddim_steps,eta=ddim_eta, + quantize_denoised=True) + # samples, z_denoise_row = self.sample(cond=c, batch_size=N, return_intermediates=True, + # quantize_denoised=True) + x_samples = self.decode_first_stage(samples.to(self.device)) + log["samples_x0_quantized"] = x_samples + + if inpaint: + # make a simple center square + b, h, w = z.shape[0], z.shape[2], z.shape[3] + mask = torch.ones(N, h, w).to(self.device) + # zeros will be filled in + mask[:, h // 4:3 * h // 4, w // 4:3 * w // 4] = 0. + mask = mask[:, None, ...] + with self.ema_scope("Plotting Inpaint"): + + samples, _ = self.sample_log(cond=c,batch_size=N,ddim=use_ddim, eta=ddim_eta, + ddim_steps=ddim_steps, x0=z[:N], mask=mask) + x_samples = self.decode_first_stage(samples.to(self.device)) + log["samples_inpainting"] = x_samples + log["mask"] = mask + + # outpaint + with self.ema_scope("Plotting Outpaint"): + samples, _ = self.sample_log(cond=c, batch_size=N, ddim=use_ddim,eta=ddim_eta, + ddim_steps=ddim_steps, x0=z[:N], mask=mask) + x_samples = self.decode_first_stage(samples.to(self.device)) + log["samples_outpainting"] = x_samples + + if plot_progressive_rows: + with self.ema_scope("Plotting Progressives"): + img, progressives = self.progressive_denoising(c, + shape=(self.channels, self.image_size, self.image_size), + batch_size=N) + prog_row = self._get_denoise_row_from_list(progressives, desc="Progressive Generation") + log["progressive_row"] = prog_row + + if return_keys: + if np.intersect1d(list(log.keys()), return_keys).shape[0] == 0: + return log + else: + return {key: log[key] for key in return_keys} + return log + + def configure_optimizers(self): + lr = self.learning_rate + params = list(self.model.parameters()) + if self.cond_stage_trainable: + print(f"{self.__class__.__name__}: Also optimizing conditioner params!") + params = params + list(self.cond_stage_model.parameters()) + if self.learn_logvar: + print('Diffusion model optimizing logvar') + params.append(self.logvar) + opt = torch.optim.AdamW(params, lr=lr) + if self.use_scheduler: + assert 'target' in self.scheduler_config + scheduler = instantiate_from_config(self.scheduler_config) + + print("Setting up LambdaLR scheduler...") + scheduler = [ + { + 'scheduler': LambdaLR(opt, lr_lambda=scheduler.schedule), + 'interval': 'step', + 'frequency': 1 + }] + return [opt], scheduler + return opt + + @torch.no_grad() + def to_rgb(self, x): + x = x.float() + if not hasattr(self, "colorize"): + self.colorize = torch.randn(3, x.shape[1], 1, 1).to(x) + x = nn.functional.conv2d(x, weight=self.colorize) + x = 2. * (x - x.min()) / (x.max() - x.min()) - 1. + return x + + +class DiffusionWrapperV1(pl.LightningModule): + def __init__(self, diff_model_config, conditioning_key): + super().__init__() + self.diffusion_model = instantiate_from_config(diff_model_config) + self.conditioning_key = conditioning_key + assert self.conditioning_key in [None, 'concat', 'crossattn', 'hybrid', 'adm'] + + def forward(self, x, t, c_concat: list = None, c_crossattn: list = None): + if self.conditioning_key is None: + out = self.diffusion_model(x, t) + elif self.conditioning_key == 'concat': + xc = torch.cat([x] + c_concat, dim=1) + out = self.diffusion_model(xc, t) + elif self.conditioning_key == 'crossattn': + cc = torch.cat(c_crossattn, 1) + out = self.diffusion_model(x, t, context=cc) + elif self.conditioning_key == 'hybrid': + xc = torch.cat([x] + c_concat, dim=1) + cc = torch.cat(c_crossattn, 1) + out = self.diffusion_model(xc, t, context=cc) + elif self.conditioning_key == 'adm': + cc = c_crossattn[0] + out = self.diffusion_model(x, t, y=cc) + else: + raise NotImplementedError() + + return out + + +class Layout2ImgDiffusionV1(LatentDiffusionV1): + # TODO: move all layout-specific hacks to this class + def __init__(self, cond_stage_key, *args, **kwargs): + assert cond_stage_key == 'coordinates_bbox', 'Layout2ImgDiffusion only for cond_stage_key="coordinates_bbox"' + super().__init__(cond_stage_key=cond_stage_key, *args, **kwargs) + + def log_images(self, batch, N=8, *args, **kwargs): + logs = super().log_images(batch=batch, N=N, *args, **kwargs) + + key = 'train' if self.training else 'validation' + dset = self.trainer.datamodule.datasets[key] + mapper = dset.conditional_builders[self.cond_stage_key] + + bbox_imgs = [] + map_fn = lambda catno: dset.get_textual_label(dset.get_category_id(catno)) + for tknzd_bbox in batch[self.cond_stage_key][:N]: + bboximg = mapper.plot(tknzd_bbox.detach().cpu(), map_fn, (256, 256)) + bbox_imgs.append(bboximg) + + cond_img = torch.stack(bbox_imgs, dim=0) + logs['bbox_image'] = cond_img + return logs + +setattr(ldm.models.diffusion.ddpm, "DDPMV1", DDPMV1) +setattr(ldm.models.diffusion.ddpm, "LatentDiffusionV1", LatentDiffusionV1) +setattr(ldm.models.diffusion.ddpm, "DiffusionWrapperV1", DiffusionWrapperV1) +setattr(ldm.models.diffusion.ddpm, "Layout2ImgDiffusionV1", Layout2ImgDiffusionV1) diff --git a/extensions-builtin/ScuNET/preload.py b/extensions-builtin/ScuNET/preload.py new file mode 100644 index 00000000..f12c5b90 --- /dev/null +++ b/extensions-builtin/ScuNET/preload.py @@ -0,0 +1,6 @@ +import os
+from modules import paths
+
+
+def preload(parser):
+ parser.add_argument("--scunet-models-path", type=str, help="Path to directory with ScuNET model file(s).", default=os.path.join(paths.models_path, 'ScuNET'))
diff --git a/modules/scunet_model.py b/extensions-builtin/ScuNET/scripts/scunet_model.py index 52360241..e0fbf3a3 100644 --- a/modules/scunet_model.py +++ b/extensions-builtin/ScuNET/scripts/scunet_model.py @@ -9,7 +9,7 @@ from basicsr.utils.download_util import load_file_from_url import modules.upscaler from modules import devices, modelloader -from modules.scunet_model_arch import SCUNet as net +from scunet_model_arch import SCUNet as net class UpscalerScuNET(modules.upscaler.Upscaler): @@ -49,7 +49,7 @@ class UpscalerScuNET(modules.upscaler.Upscaler): if model is None: return img - device = devices.device_scunet + device = devices.get_device_for('scunet') img = np.array(img) img = img[:, :, ::-1] img = np.moveaxis(img, 2, 0) / 255 @@ -66,7 +66,7 @@ class UpscalerScuNET(modules.upscaler.Upscaler): return PIL.Image.fromarray(output, 'RGB') def load_model(self, path: str): - device = devices.device_scunet + device = devices.get_device_for('scunet') if "http" in path: filename = load_file_from_url(url=self.model_url, model_dir=self.model_path, file_name="%s.pth" % self.name, progress=True) diff --git a/modules/scunet_model_arch.py b/extensions-builtin/ScuNET/scunet_model_arch.py index 43ca8d36..43ca8d36 100644 --- a/modules/scunet_model_arch.py +++ b/extensions-builtin/ScuNET/scunet_model_arch.py diff --git a/extensions-builtin/SwinIR/preload.py b/extensions-builtin/SwinIR/preload.py new file mode 100644 index 00000000..567e44bc --- /dev/null +++ b/extensions-builtin/SwinIR/preload.py @@ -0,0 +1,6 @@ +import os
+from modules import paths
+
+
+def preload(parser):
+ parser.add_argument("--swinir-models-path", type=str, help="Path to directory with SwinIR model file(s).", default=os.path.join(paths.models_path, 'SwinIR'))
diff --git a/modules/swinir_model.py b/extensions-builtin/SwinIR/scripts/swinir_model.py index 483eabd4..9a74b253 100644 --- a/modules/swinir_model.py +++ b/extensions-builtin/SwinIR/scripts/swinir_model.py @@ -7,13 +7,16 @@ from PIL import Image from basicsr.utils.download_util import load_file_from_url from tqdm import tqdm -from modules import modelloader, devices +from modules import modelloader, devices, script_callbacks, shared from modules.shared import cmd_opts, opts -from modules.swinir_model_arch import SwinIR as net -from modules.swinir_model_arch_v2 import Swin2SR as net2 +from swinir_model_arch import SwinIR as net +from swinir_model_arch_v2 import Swin2SR as net2 from modules.upscaler import Upscaler, UpscalerData +device_swinir = devices.get_device_for('swinir') + + class UpscalerSwinIR(Upscaler): def __init__(self, dirname): self.name = "SwinIR" @@ -38,7 +41,7 @@ class UpscalerSwinIR(Upscaler): model = self.load_model(model_file) if model is None: return img - model = model.to(devices.device_swinir) + model = model.to(device_swinir, dtype=devices.dtype) img = upscale(img, model) try: torch.cuda.empty_cache() @@ -90,24 +93,26 @@ class UpscalerSwinIR(Upscaler): model.load_state_dict(pretrained_model[params], strict=True) else: model.load_state_dict(pretrained_model, strict=True) - if not cmd_opts.no_half: - model = model.half() return model def upscale( img, model, - tile=opts.SWIN_tile, - tile_overlap=opts.SWIN_tile_overlap, + tile=None, + tile_overlap=None, window_size=8, scale=4, ): + tile = tile or opts.SWIN_tile + tile_overlap = tile_overlap or opts.SWIN_tile_overlap + + img = np.array(img) img = img[:, :, ::-1] img = np.moveaxis(img, 2, 0) / 255 img = torch.from_numpy(img).float() - img = img.unsqueeze(0).to(devices.device_swinir) + img = img.unsqueeze(0).to(device_swinir, dtype=devices.dtype) with torch.no_grad(), devices.autocast(): _, _, h_old, w_old = img.size() h_pad = (h_old // window_size + 1) * window_size - h_old @@ -135,8 +140,8 @@ def inference(img, model, tile, tile_overlap, window_size, scale): stride = tile - tile_overlap h_idx_list = list(range(0, h - tile, stride)) + [h - tile] w_idx_list = list(range(0, w - tile, stride)) + [w - tile] - E = torch.zeros(b, c, h * sf, w * sf, dtype=torch.half, device=devices.device_swinir).type_as(img) - W = torch.zeros_like(E, dtype=torch.half, device=devices.device_swinir) + E = torch.zeros(b, c, h * sf, w * sf, dtype=devices.dtype, device=device_swinir).type_as(img) + W = torch.zeros_like(E, dtype=devices.dtype, device=device_swinir) with tqdm(total=len(h_idx_list) * len(w_idx_list), desc="SwinIR tiles") as pbar: for h_idx in h_idx_list: @@ -155,3 +160,13 @@ def inference(img, model, tile, tile_overlap, window_size, scale): output = E.div_(W) return output + + +def on_ui_settings(): + import gradio as gr + + shared.opts.add_option("SWIN_tile", shared.OptionInfo(192, "Tile size for all SwinIR.", gr.Slider, {"minimum": 16, "maximum": 512, "step": 16}, section=('upscaling', "Upscaling"))) + shared.opts.add_option("SWIN_tile_overlap", shared.OptionInfo(8, "Tile overlap, in pixels for SwinIR. Low values = visible seam.", gr.Slider, {"minimum": 0, "maximum": 48, "step": 1}, section=('upscaling', "Upscaling"))) + + +script_callbacks.on_ui_settings(on_ui_settings) diff --git a/modules/swinir_model_arch.py b/extensions-builtin/SwinIR/swinir_model_arch.py index 863f42db..863f42db 100644 --- a/modules/swinir_model_arch.py +++ b/extensions-builtin/SwinIR/swinir_model_arch.py diff --git a/modules/swinir_model_arch_v2.py b/extensions-builtin/SwinIR/swinir_model_arch_v2.py index 0e28ae6e..0e28ae6e 100644 --- a/modules/swinir_model_arch_v2.py +++ b/extensions-builtin/SwinIR/swinir_model_arch_v2.py diff --git a/extensions-builtin/prompt-bracket-checker/javascript/prompt-bracket-checker.js b/extensions-builtin/prompt-bracket-checker/javascript/prompt-bracket-checker.js new file mode 100644 index 00000000..41d384c7 --- /dev/null +++ b/extensions-builtin/prompt-bracket-checker/javascript/prompt-bracket-checker.js @@ -0,0 +1,107 @@ +// Stable Diffusion WebUI - Bracket checker +// Version 1.0 +// By Hingashi no Florin/Bwin4L +// Counts open and closed brackets (round, square, curly) in the prompt and negative prompt text boxes in the txt2img and img2img tabs. +// If there's a mismatch, the keyword counter turns red and if you hover on it, a tooltip tells you what's wrong. + +function checkBrackets(evt) { + textArea = evt.target; + tabName = evt.target.parentElement.parentElement.id.split("_")[0]; + counterElt = document.querySelector('gradio-app').shadowRoot.querySelector('#' + tabName + '_token_counter'); + + promptName = evt.target.parentElement.parentElement.id.includes('neg') ? ' negative' : ''; + + errorStringParen = '(' + tabName + promptName + ' prompt) - Different number of opening and closing parentheses detected.\n'; + errorStringSquare = '[' + tabName + promptName + ' prompt] - Different number of opening and closing square brackets detected.\n'; + errorStringCurly = '{' + tabName + promptName + ' prompt} - Different number of opening and closing curly brackets detected.\n'; + + openBracketRegExp = /\(/g; + closeBracketRegExp = /\)/g; + + openSquareBracketRegExp = /\[/g; + closeSquareBracketRegExp = /\]/g; + + openCurlyBracketRegExp = /\{/g; + closeCurlyBracketRegExp = /\}/g; + + totalOpenBracketMatches = 0; + totalCloseBracketMatches = 0; + totalOpenSquareBracketMatches = 0; + totalCloseSquareBracketMatches = 0; + totalOpenCurlyBracketMatches = 0; + totalCloseCurlyBracketMatches = 0; + + openBracketMatches = textArea.value.match(openBracketRegExp); + if(openBracketMatches) { + totalOpenBracketMatches = openBracketMatches.length; + } + + closeBracketMatches = textArea.value.match(closeBracketRegExp); + if(closeBracketMatches) { + totalCloseBracketMatches = closeBracketMatches.length; + } + + openSquareBracketMatches = textArea.value.match(openSquareBracketRegExp); + if(openSquareBracketMatches) { + totalOpenSquareBracketMatches = openSquareBracketMatches.length; + } + + closeSquareBracketMatches = textArea.value.match(closeSquareBracketRegExp); + if(closeSquareBracketMatches) { + totalCloseSquareBracketMatches = closeSquareBracketMatches.length; + } + + openCurlyBracketMatches = textArea.value.match(openCurlyBracketRegExp); + if(openCurlyBracketMatches) { + totalOpenCurlyBracketMatches = openCurlyBracketMatches.length; + } + + closeCurlyBracketMatches = textArea.value.match(closeCurlyBracketRegExp); + if(closeCurlyBracketMatches) { + totalCloseCurlyBracketMatches = closeCurlyBracketMatches.length; + } + + if(totalOpenBracketMatches != totalCloseBracketMatches) { + if(!counterElt.title.includes(errorStringParen)) { + counterElt.title += errorStringParen; + } + } else { + counterElt.title = counterElt.title.replace(errorStringParen, ''); + } + + if(totalOpenSquareBracketMatches != totalCloseSquareBracketMatches) { + if(!counterElt.title.includes(errorStringSquare)) { + counterElt.title += errorStringSquare; + } + } else { + counterElt.title = counterElt.title.replace(errorStringSquare, ''); + } + + if(totalOpenCurlyBracketMatches != totalCloseCurlyBracketMatches) { + if(!counterElt.title.includes(errorStringCurly)) { + counterElt.title += errorStringCurly; + } + } else { + counterElt.title = counterElt.title.replace(errorStringCurly, ''); + } + + if(counterElt.title != '') { + counterElt.style = 'color: #FF5555;'; + } else { + counterElt.style = 'color: #55FF55;'; + } +} + +var shadowRootLoaded = setInterval(function() { + var shadowTextArea = document.querySelector('gradio-app').shadowRoot.querySelectorAll('#txt2img_prompt > label > textarea'); + if(shadowTextArea.length < 1) { + return false; + } + + clearInterval(shadowRootLoaded); + + document.querySelector('gradio-app').shadowRoot.querySelector('#txt2img_prompt').onkeyup = checkBrackets; + document.querySelector('gradio-app').shadowRoot.querySelector('#txt2img_neg_prompt').onkeyup = checkBrackets; + document.querySelector('gradio-app').shadowRoot.querySelector('#img2img_prompt').onkeyup = checkBrackets; + document.querySelector('gradio-app').shadowRoot.querySelector('#img2img_neg_prompt').onkeyup = checkBrackets; +}, 1000); @@ -186,6 +186,7 @@ def prepare_enviroment(): parser.add_argument("--ui-settings-file", type=str, help="filename to use for ui settings", default='config.json')
args, _ = parser.parse_known_args(sys.argv)
+ sys.argv, _ = extract_arg(sys.argv, '-f')
sys.argv, skip_torch_cuda_test = extract_arg(sys.argv, '--skip-torch-cuda-test')
sys.argv, reinstall_xformers = extract_arg(sys.argv, '--reinstall-xformers')
sys.argv, update_check = extract_arg(sys.argv, '--update-check')
diff --git a/modules/api/api.py b/modules/api/api.py index 54ee7cb0..89935a70 100644 --- a/modules/api/api.py +++ b/modules/api/api.py @@ -157,12 +157,7 @@ class Api: args.pop('include_init_images', None) # this is meant to be done by "exclude": True in model, but it's for a reason that I cannot determine. p = StableDiffusionProcessingImg2Img(**args) - imgs = [] - for img in init_images: - img = decode_base64_to_image(img) - imgs = [img] * p.batch_size - - p.init_images = imgs + p.init_images = [decode_base64_to_image(x) for x in init_images] shared.state.begin() diff --git a/modules/deepbooru.py b/modules/deepbooru.py index 31ec7e17..dfc83357 100644 --- a/modules/deepbooru.py +++ b/modules/deepbooru.py @@ -21,7 +21,7 @@ class DeepDanbooru: files = modelloader.load_models( model_path=os.path.join(paths.models_path, "torch_deepdanbooru"), model_url='https://github.com/AUTOMATIC1111/TorchDeepDanbooru/releases/download/v1/model-resnet_custom_v3.pt', - ext_filter=".pt", + ext_filter=[".pt"], download_name='model-resnet_custom_v3.pt', ) diff --git a/modules/devices.py b/modules/devices.py index d6a76844..f8cffae1 100644 --- a/modules/devices.py +++ b/modules/devices.py @@ -44,6 +44,15 @@ def get_optimal_device(): return cpu +def get_device_for(task): + from modules import shared + + if task in shared.cmd_opts.use_cpu: + return cpu + + return get_optimal_device() + + def torch_gc(): if torch.cuda.is_available(): with torch.cuda.device(get_cuda_device_string()): @@ -67,7 +76,7 @@ def enable_tf32(): errors.run(enable_tf32, "Enabling TF32") cpu = torch.device("cpu") -device = device_interrogate = device_gfpgan = device_swinir = device_esrgan = device_scunet = device_codeformer = None +device = device_interrogate = device_gfpgan = device_esrgan = device_codeformer = None dtype = torch.float16 dtype_vae = torch.float16 diff --git a/modules/extensions.py b/modules/extensions.py index db9c4200..b522125c 100644 --- a/modules/extensions.py +++ b/modules/extensions.py @@ -8,6 +8,7 @@ from modules import paths, shared extensions = []
extensions_dir = os.path.join(paths.script_path, "extensions")
+extensions_builtin_dir = os.path.join(paths.script_path, "extensions-builtin")
def active():
@@ -15,12 +16,13 @@ def active(): class Extension:
- def __init__(self, name, path, enabled=True):
+ def __init__(self, name, path, enabled=True, is_builtin=False):
self.name = name
self.path = path
self.enabled = enabled
self.status = ''
self.can_update = False
+ self.is_builtin = is_builtin
repo = None
try:
@@ -79,11 +81,19 @@ def list_extensions(): if not os.path.isdir(extensions_dir):
return
- for dirname in sorted(os.listdir(extensions_dir)):
- path = os.path.join(extensions_dir, dirname)
- if not os.path.isdir(path):
- continue
+ paths = []
+ for dirname in [extensions_dir, extensions_builtin_dir]:
+ if not os.path.isdir(dirname):
+ return
- extension = Extension(name=dirname, path=path, enabled=dirname not in shared.opts.disabled_extensions)
+ for extension_dirname in sorted(os.listdir(dirname)):
+ path = os.path.join(dirname, extension_dirname)
+ if not os.path.isdir(path):
+ continue
+
+ paths.append((extension_dirname, path, dirname == extensions_builtin_dir))
+
+ for dirname, path, is_builtin in paths:
+ extension = Extension(name=dirname, path=path, enabled=dirname not in shared.opts.disabled_extensions, is_builtin=is_builtin)
extensions.append(extension)
diff --git a/modules/extras.py b/modules/extras.py index 6021a024..0ad8deec 100644 --- a/modules/extras.py +++ b/modules/extras.py @@ -62,7 +62,7 @@ def run_extras(extras_mode, resize_mode, image, image_folder, input_dir, output_ # Also keep track of original file names
imageNameArr = []
outputs = []
-
+
if extras_mode == 1:
#convert file to pillow image
for img in image_folder:
@@ -234,7 +234,7 @@ def run_pnginfo(image): return '', geninfo, info
-def run_modelmerger(primary_model_name, secondary_model_name, teritary_model_name, interp_method, multiplier, save_as_half, custom_name, checkpoint_format):
+def run_modelmerger(primary_model_name, secondary_model_name, tertiary_model_name, interp_method, multiplier, save_as_half, custom_name, checkpoint_format):
def weighted_sum(theta0, theta1, alpha):
return ((1 - alpha) * theta0) + (alpha * theta1)
@@ -246,19 +246,8 @@ def run_modelmerger(primary_model_name, secondary_model_name, teritary_model_nam primary_model_info = sd_models.checkpoints_list[primary_model_name]
secondary_model_info = sd_models.checkpoints_list[secondary_model_name]
- teritary_model_info = sd_models.checkpoints_list.get(teritary_model_name, None)
-
- print(f"Loading {primary_model_info.filename}...")
- theta_0 = sd_models.read_state_dict(primary_model_info.filename, map_location='cpu')
-
- print(f"Loading {secondary_model_info.filename}...")
- theta_1 = sd_models.read_state_dict(secondary_model_info.filename, map_location='cpu')
-
- if teritary_model_info is not None:
- print(f"Loading {teritary_model_info.filename}...")
- theta_2 = sd_models.read_state_dict(teritary_model_info.filename, map_location='cpu')
- else:
- theta_2 = None
+ tertiary_model_info = sd_models.checkpoints_list.get(tertiary_model_name, None)
+ result_is_inpainting_model = False
theta_funcs = {
"Weighted sum": (None, weighted_sum),
@@ -266,9 +255,16 @@ def run_modelmerger(primary_model_name, secondary_model_name, teritary_model_nam }
theta_func1, theta_func2 = theta_funcs[interp_method]
- print(f"Merging...")
+ if theta_func1 and not tertiary_model_info:
+ return ["Failed: Interpolation method requires a tertiary model."] + [gr.Dropdown.update(choices=sd_models.checkpoint_tiles()) for _ in range(4)]
+
+ print(f"Loading {secondary_model_info.filename}...")
+ theta_1 = sd_models.read_state_dict(secondary_model_info.filename, map_location='cpu')
if theta_func1:
+ print(f"Loading {tertiary_model_info.filename}...")
+ theta_2 = sd_models.read_state_dict(tertiary_model_info.filename, map_location='cpu')
+
for key in tqdm.tqdm(theta_1.keys()):
if 'model' in key:
if key in theta_2:
@@ -276,12 +272,31 @@ def run_modelmerger(primary_model_name, secondary_model_name, teritary_model_nam theta_1[key] = theta_func1(theta_1[key], t2)
else:
theta_1[key] = torch.zeros_like(theta_1[key])
- del theta_2
+ del theta_2
+
+ print(f"Loading {primary_model_info.filename}...")
+ theta_0 = sd_models.read_state_dict(primary_model_info.filename, map_location='cpu')
+
+ print("Merging...")
for key in tqdm.tqdm(theta_0.keys()):
if 'model' in key and key in theta_1:
+ a = theta_0[key]
+ b = theta_1[key]
- theta_0[key] = theta_func2(theta_0[key], theta_1[key], multiplier)
+ # this enables merging an inpainting model (A) with another one (B);
+ # where normal model would have 4 channels, for latenst space, inpainting model would
+ # have another 4 channels for unmasked picture's latent space, plus one channel for mask, for a total of 9
+ if a.shape != b.shape and a.shape[0:1] + a.shape[2:] == b.shape[0:1] + b.shape[2:]:
+ if a.shape[1] == 4 and b.shape[1] == 9:
+ raise RuntimeError("When merging inpainting model with a normal one, A must be the inpainting model.")
+
+ assert a.shape[1] == 9 and b.shape[1] == 4, f"Bad dimensions for merged layer {key}: A={a.shape}, B={b.shape}"
+
+ theta_0[key][:, 0:4, :, :] = theta_func2(a[:, 0:4, :, :], b, multiplier)
+ result_is_inpainting_model = True
+ else:
+ theta_0[key] = theta_func2(a, b, multiplier)
if save_as_half:
theta_0[key] = theta_0[key].half()
@@ -292,11 +307,20 @@ def run_modelmerger(primary_model_name, secondary_model_name, teritary_model_nam theta_0[key] = theta_1[key]
if save_as_half:
theta_0[key] = theta_0[key].half()
+ del theta_1
ckpt_dir = shared.cmd_opts.ckpt_dir or sd_models.model_path
- filename = primary_model_info.model_name + '_' + str(round(1-multiplier, 2)) + '-' + secondary_model_info.model_name + '_' + str(round(multiplier, 2)) + '-' + interp_method.replace(" ", "_") + '-merged.' + checkpoint_format
+ filename = \
+ primary_model_info.model_name + '_' + str(round(1-multiplier, 2)) + '-' + \
+ secondary_model_info.model_name + '_' + str(round(multiplier, 2)) + '-' + \
+ interp_method.replace(" ", "_") + \
+ '-merged.' + \
+ ("inpainting." if result_is_inpainting_model else "") + \
+ checkpoint_format
+
filename = filename if custom_name == '' else (custom_name + '.' + checkpoint_format)
+
output_modelname = os.path.join(ckpt_dir, filename)
print(f"Saving to {output_modelname}...")
@@ -309,5 +333,5 @@ def run_modelmerger(primary_model_name, secondary_model_name, teritary_model_nam sd_models.list_models()
- print(f"Checkpoint saved.")
+ print("Checkpoint saved.")
return ["Checkpoint saved to " + output_modelname] + [gr.Dropdown.update(choices=sd_models.checkpoint_tiles()) for _ in range(4)]
diff --git a/modules/img2img.py b/modules/img2img.py index 830cfa15..81da4b13 100644 --- a/modules/img2img.py +++ b/modules/img2img.py @@ -4,7 +4,7 @@ import sys import traceback
import numpy as np
-from PIL import Image, ImageOps, ImageFilter, ImageEnhance
+from PIL import Image, ImageOps, ImageFilter, ImageEnhance, ImageChops
from modules import devices, sd_samplers
from modules.processing import Processed, StableDiffusionProcessingImg2Img, process_images
@@ -66,22 +66,23 @@ def img2img(mode: int, prompt: str, negative_prompt: str, prompt_style: str, pro if is_inpaint:
# Drawn mask
if mask_mode == 0:
- image = init_img_with_mask
- is_mask_sketch = isinstance(image, dict)
+ is_mask_sketch = isinstance(init_img_with_mask, dict)
is_mask_paint = not is_mask_sketch
if is_mask_sketch:
# Sketch: mask iff. not transparent
- image, mask = image["image"], image["mask"]
- pred = np.array(mask)[..., -1] > 0
+ image, mask = init_img_with_mask["image"], init_img_with_mask["mask"]
+ alpha_mask = ImageOps.invert(image.split()[-1]).convert('L').point(lambda x: 255 if x > 0 else 0, mode='1')
+ mask = ImageChops.lighter(alpha_mask, mask.convert('L')).convert('L')
else:
# Color-sketch: mask iff. painted over
- orig = init_img_with_mask_orig or image
+ image = init_img_with_mask
+ orig = init_img_with_mask_orig or init_img_with_mask
pred = np.any(np.array(image) != np.array(orig), axis=-1)
- mask = Image.fromarray(pred.astype(np.uint8) * 255, "L")
- if is_mask_paint:
+ mask = Image.fromarray(pred.astype(np.uint8) * 255, "L")
mask = ImageEnhance.Brightness(mask).enhance(1 - mask_alpha / 100)
blur = ImageFilter.GaussianBlur(mask_blur)
image = Image.composite(image.filter(blur), orig, mask.filter(blur))
+
image = image.convert("RGB")
# Uploaded mask
else:
diff --git a/modules/interrogate.py b/modules/interrogate.py index 40c6b082..0068b81c 100644 --- a/modules/interrogate.py +++ b/modules/interrogate.py @@ -1,4 +1,3 @@ -import contextlib
import os
import sys
import traceback
@@ -11,10 +10,9 @@ from torchvision import transforms from torchvision.transforms.functional import InterpolationMode
import modules.shared as shared
-from modules import devices, paths, lowvram
+from modules import devices, paths, lowvram, modelloader
blip_image_eval_size = 384
-blip_model_url = 'https://storage.googleapis.com/sfr-vision-language-research/BLIP/models/model_base_caption_capfilt_large.pth'
clip_model_name = 'ViT-L/14'
Category = namedtuple("Category", ["name", "topn", "items"])
@@ -47,7 +45,14 @@ class InterrogateModels: def load_blip_model(self):
import models.blip
- blip_model = models.blip.blip_decoder(pretrained=blip_model_url, image_size=blip_image_eval_size, vit='base', med_config=os.path.join(paths.paths["BLIP"], "configs", "med_config.json"))
+ files = modelloader.load_models(
+ model_path=os.path.join(paths.models_path, "BLIP"),
+ model_url='https://storage.googleapis.com/sfr-vision-language-research/BLIP/models/model_base_caption_capfilt_large.pth',
+ ext_filter=[".pth"],
+ download_name='model_base_caption_capfilt_large.pth',
+ )
+
+ blip_model = models.blip.blip_decoder(pretrained=files[0], image_size=blip_image_eval_size, vit='base', med_config=os.path.join(paths.paths["BLIP"], "configs", "med_config.json"))
blip_model.eval()
return blip_model
diff --git a/modules/modelloader.py b/modules/modelloader.py index 7d2f0ade..e647f6fa 100644 --- a/modules/modelloader.py +++ b/modules/modelloader.py @@ -124,10 +124,9 @@ def move_files(src_path: str, dest_path: str, ext_filter: str = None): def load_upscalers(): - sd = shared.script_path # We can only do this 'magic' method to dynamically load upscalers if they are referenced, # so we'll try to import any _model.py files before looking in __subclasses__ - modules_dir = os.path.join(sd, "modules") + modules_dir = os.path.join(shared.script_path, "modules") for file in os.listdir(modules_dir): if "_model.py" in file: model_name = file.replace("_model.py", "") @@ -136,22 +135,13 @@ def load_upscalers(): importlib.import_module(full_model) except: pass + datas = [] - c_o = vars(shared.cmd_opts) + commandline_options = vars(shared.cmd_opts) for cls in Upscaler.__subclasses__(): name = cls.__name__ - module_name = cls.__module__ - module = importlib.import_module(module_name) - class_ = getattr(module, name) cmd_name = f"{name.lower().replace('upscaler', '')}_models_path" - opt_string = None - try: - if cmd_name in c_o: - opt_string = c_o[cmd_name] - except: - pass - scaler = class_(opt_string) - for child in scaler.scalers: - datas.append(child) + scaler = cls(commandline_options.get(cmd_name, None)) + datas += scaler.scalers shared.sd_upscalers = datas diff --git a/modules/processing.py b/modules/processing.py index b9cb6d32..dd22a2fa 100644 --- a/modules/processing.py +++ b/modules/processing.py @@ -21,7 +21,10 @@ import modules.face_restoration import modules.images as images
import modules.styles
import logging
+from ldm.data.util import AddMiDaS
+from ldm.models.diffusion.ddpm import LatentDepth2ImageDiffusion
+from einops import repeat, rearrange
# some of those options should not be changed at all because they would break the model, so I removed them from options.
opt_C = 4
@@ -150,11 +153,26 @@ class StableDiffusionProcessing(): return image_conditioning
- def img2img_image_conditioning(self, source_image, latent_image, image_mask = None):
- if self.sampler.conditioning_key not in {'hybrid', 'concat'}:
- # Dummy zero conditioning if we're not using inpainting model.
- return latent_image.new_zeros(latent_image.shape[0], 5, 1, 1)
+ def depth2img_image_conditioning(self, source_image):
+ # Use the AddMiDaS helper to Format our source image to suit the MiDaS model
+ transformer = AddMiDaS(model_type="dpt_hybrid")
+ transformed = transformer({"jpg": rearrange(source_image[0], "c h w -> h w c")})
+ midas_in = torch.from_numpy(transformed["midas_in"][None, ...]).to(device=shared.device)
+ midas_in = repeat(midas_in, "1 ... -> n ...", n=self.batch_size)
+
+ conditioning_image = self.sd_model.get_first_stage_encoding(self.sd_model.encode_first_stage(source_image))
+ conditioning = torch.nn.functional.interpolate(
+ self.sd_model.depth_model(midas_in),
+ size=conditioning_image.shape[2:],
+ mode="bicubic",
+ align_corners=False,
+ )
+
+ (depth_min, depth_max) = torch.aminmax(conditioning)
+ conditioning = 2. * (conditioning - depth_min) / (depth_max - depth_min) - 1.
+ return conditioning
+ def inpainting_image_conditioning(self, source_image, latent_image, image_mask = None):
self.is_using_inpainting_conditioning = True
# Handle the different mask inputs
@@ -191,6 +209,18 @@ class StableDiffusionProcessing(): return image_conditioning
+ def img2img_image_conditioning(self, source_image, latent_image, image_mask=None):
+ # HACK: Using introspection as the Depth2Image model doesn't appear to uniquely
+ # identify itself with a field common to all models. The conditioning_key is also hybrid.
+ if isinstance(self.sd_model, LatentDepth2ImageDiffusion):
+ return self.depth2img_image_conditioning(source_image)
+
+ if self.sampler.conditioning_key in {'hybrid', 'concat'}:
+ return self.inpainting_image_conditioning(source_image, latent_image, image_mask=image_mask)
+
+ # Dummy zero conditioning if we're not using inpainting or depth model.
+ return latent_image.new_zeros(latent_image.shape[0], 5, 1, 1)
+
def init(self, all_prompts, all_seeds, all_subseeds):
pass
diff --git a/modules/sd_hijack.py b/modules/sd_hijack.py index 95a17093..690a9ec2 100644 --- a/modules/sd_hijack.py +++ b/modules/sd_hijack.py @@ -1,17 +1,11 @@ -import math
-import os
-import sys
-import traceback
import torch
-import numpy as np
-from torch import einsum
from torch.nn.functional import silu
import modules.textual_inversion.textual_inversion
-from modules import prompt_parser, devices, sd_hijack_optimizations, shared, sd_hijack_checkpoint
+from modules import devices, sd_hijack_optimizations, shared, sd_hijack_checkpoint
from modules.hypernetworks import hypernetwork
-from modules.shared import opts, device, cmd_opts
-from modules import sd_hijack_clip, sd_hijack_open_clip
+from modules.shared import cmd_opts
+from modules import sd_hijack_clip, sd_hijack_open_clip, sd_hijack_unet
from modules.sd_hijack_optimizations import invokeAI_mps_available
@@ -35,10 +29,12 @@ ldm.modules.attention.BasicTransformerBlock.ATTENTION_MODES["softmax-xformers"] ldm.modules.attention.print = lambda *args: None
ldm.modules.diffusionmodules.model.print = lambda *args: None
+
def apply_optimizations():
undo_optimizations()
ldm.modules.diffusionmodules.model.nonlinearity = silu
+ ldm.modules.diffusionmodules.openaimodel.th = sd_hijack_unet.th
if cmd_opts.force_enable_xformers or (cmd_opts.xformers and shared.xformers_available and torch.version.cuda and (6, 0) <= torch.cuda.get_device_capability(shared.device) <= (9, 0)):
print("Applying xformers cross attention optimization.")
diff --git a/modules/sd_hijack_unet.py b/modules/sd_hijack_unet.py new file mode 100644 index 00000000..1b9d7757 --- /dev/null +++ b/modules/sd_hijack_unet.py @@ -0,0 +1,30 @@ +import torch
+
+
+class TorchHijackForUnet:
+ """
+ This is torch, but with cat that resizes tensors to appropriate dimensions if they do not match;
+ this makes it possible to create pictures with dimensions that are muliples of 8 rather than 64
+ """
+
+ def __getattr__(self, item):
+ if item == 'cat':
+ return self.cat
+
+ if hasattr(torch, item):
+ return getattr(torch, item)
+
+ raise AttributeError("'{}' object has no attribute '{}'".format(type(self).__name__, item))
+
+ def cat(self, tensors, *args, **kwargs):
+ if len(tensors) == 2:
+ a, b = tensors
+ if a.shape[-2:] != b.shape[-2:]:
+ a = torch.nn.functional.interpolate(a, b.shape[-2:], mode="nearest")
+
+ tensors = (a, b)
+
+ return torch.cat(tensors, *args, **kwargs)
+
+
+th = TorchHijackForUnet()
diff --git a/modules/sd_models.py b/modules/sd_models.py index 283cf1cd..139952ba 100644 --- a/modules/sd_models.py +++ b/modules/sd_models.py @@ -7,6 +7,9 @@ import torch import re
import safetensors.torch
from omegaconf import OmegaConf
+from os import mkdir
+from urllib import request
+import ldm.modules.midas as midas
from ldm.util import instantiate_from_config
@@ -36,6 +39,7 @@ def setup_model(): os.makedirs(model_path)
list_models()
+ enable_midas_autodownload()
def checkpoint_tiles():
@@ -227,6 +231,48 @@ def load_model_weights(model, checkpoint_info, vae_file="auto"): sd_vae.load_vae(model, vae_file)
+def enable_midas_autodownload():
+ """
+ Gives the ldm.modules.midas.api.load_model function automatic downloading.
+
+ When the 512-depth-ema model, and other future models like it, is loaded,
+ it calls midas.api.load_model to load the associated midas depth model.
+ This function applies a wrapper to download the model to the correct
+ location automatically.
+ """
+
+ midas_path = os.path.join(models_path, 'midas')
+
+ # stable-diffusion-stability-ai hard-codes the midas model path to
+ # a location that differs from where other scripts using this model look.
+ # HACK: Overriding the path here.
+ for k, v in midas.api.ISL_PATHS.items():
+ file_name = os.path.basename(v)
+ midas.api.ISL_PATHS[k] = os.path.join(midas_path, file_name)
+
+ midas_urls = {
+ "dpt_large": "https://github.com/intel-isl/DPT/releases/download/1_0/dpt_large-midas-2f21e586.pt",
+ "dpt_hybrid": "https://github.com/intel-isl/DPT/releases/download/1_0/dpt_hybrid-midas-501f0c75.pt",
+ "midas_v21": "https://github.com/AlexeyAB/MiDaS/releases/download/midas_dpt/midas_v21-f6b98070.pt",
+ "midas_v21_small": "https://github.com/AlexeyAB/MiDaS/releases/download/midas_dpt/midas_v21_small-70d6b9c8.pt",
+ }
+
+ midas.api.load_model_inner = midas.api.load_model
+
+ def load_model_wrapper(model_type):
+ path = midas.api.ISL_PATHS[model_type]
+ if not os.path.exists(path):
+ if not os.path.exists(midas_path):
+ mkdir(midas_path)
+
+ print(f"Downloading midas model weights for {model_type} to {path}")
+ request.urlretrieve(midas_urls[model_type], path)
+ print(f"{model_type} downloaded")
+
+ return midas.api.load_model_inner(model_type)
+
+ midas.api.load_model = load_model_wrapper
+
def load_model(checkpoint_info=None):
from modules import lowvram, sd_hijack
checkpoint_info = checkpoint_info or select_checkpoint()
diff --git a/modules/shared.py b/modules/shared.py index 4182e2ac..67f8f77b 100644 --- a/modules/shared.py +++ b/modules/shared.py @@ -50,9 +50,6 @@ parser.add_argument("--gfpgan-models-path", type=str, help="Path to directory wi parser.add_argument("--esrgan-models-path", type=str, help="Path to directory with ESRGAN model file(s).", default=os.path.join(models_path, 'ESRGAN'))
parser.add_argument("--bsrgan-models-path", type=str, help="Path to directory with BSRGAN model file(s).", default=os.path.join(models_path, 'BSRGAN'))
parser.add_argument("--realesrgan-models-path", type=str, help="Path to directory with RealESRGAN model file(s).", default=os.path.join(models_path, 'RealESRGAN'))
-parser.add_argument("--scunet-models-path", type=str, help="Path to directory with ScuNET model file(s).", default=os.path.join(models_path, 'ScuNET'))
-parser.add_argument("--swinir-models-path", type=str, help="Path to directory with SwinIR model file(s).", default=os.path.join(models_path, 'SwinIR'))
-parser.add_argument("--ldsr-models-path", type=str, help="Path to directory with LDSR model file(s).", default=os.path.join(models_path, 'LDSR'))
parser.add_argument("--clip-models-path", type=str, help="Path to directory with CLIP model file(s).", default=None)
parser.add_argument("--xformers", action='store_true', help="enable xformers for cross attention layers")
parser.add_argument("--force-enable-xformers", action='store_true', help="enable xformers for cross attention layers regardless of whether the checking code thinks you can run it; do not make bug reports if this fails to work")
@@ -61,7 +58,7 @@ parser.add_argument("--opt-split-attention", action='store_true', help="force-en parser.add_argument("--opt-split-attention-invokeai", action='store_true', help="force-enables InvokeAI's cross-attention layer optimization. By default, it's on when cuda is unavailable.")
parser.add_argument("--opt-split-attention-v1", action='store_true', help="enable older version of split attention optimization that does not consume all the VRAM it can find")
parser.add_argument("--disable-opt-split-attention", action='store_true', help="force-disables cross-attention layer optimization")
-parser.add_argument("--use-cpu", nargs='+',choices=['all', 'sd', 'interrogate', 'gfpgan', 'swinir', 'esrgan', 'scunet', 'codeformer'], help="use CPU as torch device for specified modules", default=[], type=str.lower)
+parser.add_argument("--use-cpu", nargs='+', help="use CPU as torch device for specified modules", default=[], type=str.lower)
parser.add_argument("--listen", action='store_true', help="launch gradio with 0.0.0.0 as server name, allowing to respond to network requests")
parser.add_argument("--port", type=int, help="launch gradio with given server port, you need root/admin rights for ports < 1024, defaults to 7860 if available", default=None)
parser.add_argument("--show-negative-prompt", action='store_true', help="does not do anything", default=False)
@@ -95,6 +92,7 @@ parser.add_argument("--tls-certfile", type=str, help="Partially enables TLS, req parser.add_argument("--server-name", type=str, help="Sets hostname of server", default=None)
script_loading.preload_extensions(extensions.extensions_dir, parser)
+script_loading.preload_extensions(extensions.extensions_builtin_dir, parser)
cmd_opts = parser.parse_args()
@@ -112,8 +110,8 @@ restricted_opts = { cmd_opts.disable_extension_access = (cmd_opts.share or cmd_opts.listen or cmd_opts.server_name) and not cmd_opts.enable_insecure_extension_access
-devices.device, devices.device_interrogate, devices.device_gfpgan, devices.device_swinir, devices.device_esrgan, devices.device_scunet, devices.device_codeformer = \
-(devices.cpu if any(y in cmd_opts.use_cpu for y in [x, 'all']) else devices.get_optimal_device() for x in ['sd', 'interrogate', 'gfpgan', 'swinir', 'esrgan', 'scunet', 'codeformer'])
+devices.device, devices.device_interrogate, devices.device_gfpgan, devices.device_esrgan, devices.device_codeformer = \
+ (devices.cpu if any(y in cmd_opts.use_cpu for y in [x, 'all']) else devices.get_optimal_device() for x in ['sd', 'interrogate', 'gfpgan', 'esrgan', 'codeformer'])
device = devices.device
weight_load_location = None if cmd_opts.lowram else "cpu"
@@ -326,9 +324,6 @@ options_templates.update(options_section(('upscaling', "Upscaling"), { "ESRGAN_tile": OptionInfo(192, "Tile size for ESRGAN upscalers. 0 = no tiling.", gr.Slider, {"minimum": 0, "maximum": 512, "step": 16}),
"ESRGAN_tile_overlap": OptionInfo(8, "Tile overlap, in pixels for ESRGAN upscalers. Low values = visible seam.", gr.Slider, {"minimum": 0, "maximum": 48, "step": 1}),
"realesrgan_enabled_models": OptionInfo(["R-ESRGAN 4x+", "R-ESRGAN 4x+ Anime6B"], "Select which Real-ESRGAN models to show in the web UI. (Requires restart)", gr.CheckboxGroup, lambda: {"choices": realesrgan_models_names()}),
- "SWIN_tile": OptionInfo(192, "Tile size for all SwinIR.", gr.Slider, {"minimum": 16, "maximum": 512, "step": 16}),
- "SWIN_tile_overlap": OptionInfo(8, "Tile overlap, in pixels for SwinIR. Low values = visible seam.", gr.Slider, {"minimum": 0, "maximum": 48, "step": 1}),
- "ldsr_steps": OptionInfo(100, "LDSR processing steps. Lower = faster", gr.Slider, {"minimum": 1, "maximum": 200, "step": 1}),
"upscaler_for_img2img": OptionInfo(None, "Upscaler for img2img", gr.Dropdown, lambda: {"choices": [x.name for x in sd_upscalers]}),
"use_scale_latent_for_hires_fix": OptionInfo(False, "Upscale latent space image when doing hires. fix"),
}))
diff --git a/modules/ui.py b/modules/ui.py index 2eb0b684..38096aad 100644 --- a/modules/ui.py +++ b/modules/ui.py @@ -28,7 +28,6 @@ import modules.codeformer_model import modules.generation_parameters_copypaste as parameters_copypaste
import modules.gfpgan_model
import modules.hypernetworks.ui
-import modules.ldsr_model
import modules.scripts
import modules.shared as shared
import modules.styles
@@ -303,8 +302,8 @@ def create_seed_inputs(): with gr.Row(visible=False) as seed_extra_row_2:
seed_extras.append(seed_extra_row_2)
- seed_resize_from_w = gr.Slider(minimum=0, maximum=2048, step=64, label="Resize seed from width", value=0)
- seed_resize_from_h = gr.Slider(minimum=0, maximum=2048, step=64, label="Resize seed from height", value=0)
+ seed_resize_from_w = gr.Slider(minimum=0, maximum=2048, step=8, label="Resize seed from width", value=0)
+ seed_resize_from_h = gr.Slider(minimum=0, maximum=2048, step=8, label="Resize seed from height", value=0)
random_seed.click(fn=lambda: -1, show_progress=False, inputs=[], outputs=[seed])
random_subseed.click(fn=lambda: -1, show_progress=False, inputs=[], outputs=[subseed])
@@ -636,8 +635,8 @@ def create_ui(): sampler_index = gr.Radio(label='Sampling method', elem_id="txt2img_sampling", choices=[x.name for x in samplers], value=samplers[0].name, type="index")
with gr.Group():
- width = gr.Slider(minimum=64, maximum=2048, step=64, label="Width", value=512)
- height = gr.Slider(minimum=64, maximum=2048, step=64, label="Height", value=512)
+ width = gr.Slider(minimum=64, maximum=2048, step=8, label="Width", value=512)
+ height = gr.Slider(minimum=64, maximum=2048, step=8, label="Height", value=512)
with gr.Row():
restore_faces = gr.Checkbox(label='Restore faces', value=False, visible=len(shared.face_restorers) > 1)
@@ -645,8 +644,8 @@ def create_ui(): enable_hr = gr.Checkbox(label='Highres. fix', value=False)
with gr.Row(visible=False) as hr_options:
- firstphase_width = gr.Slider(minimum=0, maximum=1024, step=64, label="Firstpass width", value=0)
- firstphase_height = gr.Slider(minimum=0, maximum=1024, step=64, label="Firstpass height", value=0)
+ firstphase_width = gr.Slider(minimum=0, maximum=1024, step=8, label="Firstpass width", value=0)
+ firstphase_height = gr.Slider(minimum=0, maximum=1024, step=8, label="Firstpass height", value=0)
denoising_strength = gr.Slider(minimum=0.0, maximum=1.0, step=0.01, label='Denoising strength', value=0.7)
with gr.Row(equal_height=True):
@@ -792,23 +791,26 @@ def create_ui(): init_img = gr.Image(label="Image for img2img", elem_id="img2img_image", show_label=False, source="upload", interactive=True, type="pil", tool=cmd_opts.gradio_img2img_tool).style(height=480)
with gr.TabItem('Inpaint', id='inpaint'):
- init_img_with_mask_orig = gr.State(None)
init_img_with_mask = gr.Image(label="Image for inpainting with mask", show_label=False, elem_id="img2maskimg", source="upload", interactive=True, type="pil", tool=cmd_opts.gradio_inpaint_tool, image_mode="RGBA").style(height=480)
+ init_img_with_mask_orig = gr.State(None)
- def update_orig(image, state):
- if image is not None:
- same_size = state is not None and state.size == image.size
- has_exact_match = np.any(np.all(np.array(image) == np.array(state), axis=-1))
- edited = same_size and has_exact_match
- return image if not edited or state is None else state
+ use_color_sketch = cmd_opts.gradio_inpaint_tool == "color-sketch"
+ if use_color_sketch:
+ def update_orig(image, state):
+ if image is not None:
+ same_size = state is not None and state.size == image.size
+ has_exact_match = np.any(np.all(np.array(image) == np.array(state), axis=-1))
+ edited = same_size and has_exact_match
+ return image if not edited or state is None else state
+
+ init_img_with_mask.change(update_orig, [init_img_with_mask, init_img_with_mask_orig], init_img_with_mask_orig)
- init_img_with_mask.change(update_orig, [init_img_with_mask, init_img_with_mask_orig], init_img_with_mask_orig)
init_img_inpaint = gr.Image(label="Image for img2img", show_label=False, source="upload", interactive=True, type="pil", visible=False, elem_id="img_inpaint_base")
init_mask_inpaint = gr.Image(label="Mask", source="upload", interactive=True, type="pil", visible=False, elem_id="img_inpaint_mask")
- show_mask_alpha = cmd_opts.gradio_inpaint_tool == "color-sketch"
- mask_alpha = gr.Slider(label="Mask transparency", interactive=show_mask_alpha, visible=show_mask_alpha)
- mask_blur = gr.Slider(label='Mask blur', minimum=0, maximum=64, step=1, value=4)
+ with gr.Row():
+ mask_blur = gr.Slider(label='Mask blur', minimum=0, maximum=64, step=1, value=4)
+ mask_alpha = gr.Slider(label="Mask transparency", interactive=use_color_sketch, visible=use_color_sketch)
with gr.Row():
mask_mode = gr.Radio(label="Mask mode", show_label=False, choices=["Draw mask", "Upload mask"], type="index", value="Draw mask", elem_id="mask_mode")
@@ -833,8 +835,8 @@ def create_ui(): sampler_index = gr.Radio(label='Sampling method', choices=[x.name for x in samplers_for_img2img], value=samplers_for_img2img[0].name, type="index")
with gr.Group():
- width = gr.Slider(minimum=64, maximum=2048, step=64, label="Width", value=512, elem_id="img2img_width")
- height = gr.Slider(minimum=64, maximum=2048, step=64, label="Height", value=512, elem_id="img2img_height")
+ width = gr.Slider(minimum=64, maximum=2048, step=8, label="Width", value=512, elem_id="img2img_width")
+ height = gr.Slider(minimum=64, maximum=2048, step=8, label="Height", value=512, elem_id="img2img_height")
with gr.Row():
restore_faces = gr.Checkbox(label='Restore faces', value=False, visible=len(shared.face_restorers) > 1)
@@ -1169,8 +1171,8 @@ def create_ui(): with gr.Tab(label="Preprocess images"):
process_src = gr.Textbox(label='Source directory')
process_dst = gr.Textbox(label='Destination directory')
- process_width = gr.Slider(minimum=64, maximum=2048, step=64, label="Width", value=512)
- process_height = gr.Slider(minimum=64, maximum=2048, step=64, label="Height", value=512)
+ process_width = gr.Slider(minimum=64, maximum=2048, step=8, label="Width", value=512)
+ process_height = gr.Slider(minimum=64, maximum=2048, step=8, label="Height", value=512)
preprocess_txt_action = gr.Dropdown(label='Existing Caption txt Action', value="ignore", choices=["ignore", "copy", "prepend", "append"])
with gr.Row():
@@ -1228,8 +1230,8 @@ def create_ui(): dataset_directory = gr.Textbox(label='Dataset directory', placeholder="Path to directory with input images")
log_directory = gr.Textbox(label='Log directory', placeholder="Path to directory where to write outputs", value="textual_inversion")
template_file = gr.Textbox(label='Prompt template file', value=os.path.join(script_path, "textual_inversion_templates", "style_filewords.txt"))
- training_width = gr.Slider(minimum=64, maximum=2048, step=64, label="Width", value=512)
- training_height = gr.Slider(minimum=64, maximum=2048, step=64, label="Height", value=512)
+ training_width = gr.Slider(minimum=64, maximum=2048, step=8, label="Width", value=512)
+ training_height = gr.Slider(minimum=64, maximum=2048, step=8, label="Height", value=512)
steps = gr.Number(label='Max steps', value=100000, precision=0)
create_image_every = gr.Number(label='Save an image to log directory every N steps, 0 to disable', value=500, precision=0)
save_embedding_every = gr.Number(label='Save a copy of embedding to log directory every N steps, 0 to disable', value=500, precision=0)
@@ -1448,7 +1450,7 @@ def create_ui(): opts.save(shared.config_filename)
except RuntimeError:
return opts.dumpjson(), f'{len(changed)} settings changed without save: {", ".join(changed)}.'
- return opts.dumpjson(), f'{len(changed)} settings changed: {", ".join(changed)}.'
+ return opts.dumpjson(), f'{len(changed)} settings changed{": " if len(changed) > 0 else ""}{", ".join(changed)}.'
def run_settings_single(value, key):
if not opts.same_type(value, opts.data_labels[key].default):
diff --git a/modules/ui_extensions.py b/modules/ui_extensions.py index 42667941..b487ac25 100644 --- a/modules/ui_extensions.py +++ b/modules/ui_extensions.py @@ -78,6 +78,12 @@ def extension_table(): """
for ext in extensions.extensions:
+ remote = ""
+ if ext.is_builtin:
+ remote = "built-in"
+ elif ext.remote:
+ remote = f"""<a href="{html.escape(ext.remote or '')}" target="_blank">{html.escape("built-in" if ext.is_builtin else ext.remote or '')}</a>"""
+
if ext.can_update:
ext_status = f"""<label><input class="gr-check-radio gr-checkbox" name="update_{html.escape(ext.name)}" checked="checked" type="checkbox">{html.escape(ext.status)}</label>"""
else:
@@ -86,7 +92,7 @@ def extension_table(): code += f"""
<tr>
<td><label><input class="gr-check-radio gr-checkbox" name="enable_{html.escape(ext.name)}" type="checkbox" {'checked="checked"' if ext.enabled else ''}>{html.escape(ext.name)}</label></td>
- <td><a href="{html.escape(ext.remote or '')}" target="_blank">{html.escape(ext.remote or '')}</a></td>
+ <td>{remote}</td>
<td{' class="extension_status"' if ext.remote is not None else ''}>{ext_status}</td>
</tr>
"""
diff --git a/requirements.txt b/requirements.txt index 5f3d9623..05818aa6 100644 --- a/requirements.txt +++ b/requirements.txt @@ -30,3 +30,4 @@ inflection GitPython
torchsde
safetensors
+psutil; sys_platform == 'darwin'
diff --git a/webui-macos-env.sh b/webui-macos-env.sh new file mode 100644 index 00000000..95ca9c55 --- /dev/null +++ b/webui-macos-env.sh @@ -0,0 +1,19 @@ +#!/bin/bash +#################################################################### +# macOS defaults # +# Please modify webui-user.sh to change these instead of this file # +#################################################################### + +if [[ -x "$(command -v python3.10)" ]] +then + python_cmd="python3.10" +fi + +export install_dir="$HOME" +export COMMANDLINE_ARGS="--skip-torch-cuda-test --no-half --use-cpu interrogate" +export TORCH_COMMAND="pip install torch==1.12.1 torchvision==0.13.1" +export K_DIFFUSION_REPO="https://github.com/brkirch/k-diffusion.git" +export K_DIFFUSION_COMMIT_HASH="51c9778f269cedb55a4d88c79c0246d35bdadb71" +export PYTORCH_ENABLE_MPS_FALLBACK=1 + +#################################################################### diff --git a/webui-user.sh b/webui-user.sh index 16e42759..bfa53cb7 100644 --- a/webui-user.sh +++ b/webui-user.sh @@ -10,7 +10,7 @@ #clone_dir="stable-diffusion-webui" # Commandline arguments for webui.py, for example: export COMMANDLINE_ARGS="--medvram --opt-split-attention" -export COMMANDLINE_ARGS="" +#export COMMANDLINE_ARGS="" # python3 executable #python_cmd="python3" @@ -53,10 +53,11 @@ def initialize(): codeformer.setup_model(cmd_opts.codeformer_models_path)
gfpgan.setup_model(cmd_opts.gfpgan_models_path)
shared.face_restorers.append(modules.face_restoration.FaceRestoration())
- modelloader.load_upscalers()
modules.scripts.load_scripts()
+ modelloader.load_upscalers()
+
modules.sd_vae.refresh_vae_list()
modules.sd_models.load_model()
shared.opts.onchange("sd_model_checkpoint", wrap_queued_call(lambda: modules.sd_models.reload_model_weights()))
@@ -177,6 +178,8 @@ def webui(): print('Reloading custom scripts')
modules.scripts.reload_scripts()
+ modelloader.load_upscalers()
+
print('Reloading modules: modules.ui')
importlib.reload(modules.ui)
print('Refreshing Model List')
@@ -4,6 +4,14 @@ # change the variables in webui-user.sh instead # ################################################# +# If run from macOS, load defaults from webui-macos-env.sh +if [[ "$OSTYPE" == "darwin"* ]]; then + if [[ -f webui-macos-env.sh ]] + then + source ./webui-macos-env.sh + fi +fi + # Read variables from webui-user.sh # shellcheck source=/dev/null if [[ -f webui-user.sh ]] @@ -51,10 +59,11 @@ fi can_run_as_root=0 # read any command line flags to the webui.sh script -while getopts "f" flag +while getopts "f" flag > /dev/null 2>&1 do case ${flag} in f) can_run_as_root=1;; + *) break;; esac done |