aboutsummaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
-rw-r--r--README.md5
-rw-r--r--extensions-builtin/LDSR/ldsr_model_arch.py48
-rw-r--r--extensions-builtin/LDSR/scripts/ldsr_model.py3
-rw-r--r--extensions-builtin/LDSR/sd_hijack_ddpm_v1.py1449
-rw-r--r--extensions-builtin/prompt-bracket-checker/javascript/prompt-bracket-checker.js107
-rw-r--r--javascript/contextMenus.js24
-rw-r--r--javascript/hints.js1
-rw-r--r--javascript/progressbar.js12
-rw-r--r--javascript/ui.js11
-rw-r--r--launch.py4
-rw-r--r--modules/api/api.py25
-rw-r--r--modules/api/models.py2
-rw-r--r--modules/devices.py9
-rw-r--r--modules/extras.py47
-rw-r--r--modules/generation_parameters_copypaste.py4
-rw-r--r--modules/images.py89
-rw-r--r--modules/import_hook.py5
-rw-r--r--modules/lowvram.py12
-rw-r--r--modules/ngrok.py4
-rw-r--r--modules/processing.py100
-rw-r--r--modules/safe.py16
-rw-r--r--modules/safety.py42
-rw-r--r--modules/scripts.py24
-rw-r--r--modules/sd_hijack.py14
-rw-r--r--modules/sd_hijack_inpainting.py9
-rw-r--r--modules/sd_hijack_optimizations.py10
-rw-r--r--modules/sd_hijack_unet.py30
-rw-r--r--modules/sd_models.py53
-rw-r--r--modules/sd_samplers.py29
-rw-r--r--modules/sd_vae.py37
-rw-r--r--modules/shared.py6
-rw-r--r--modules/textual_inversion/dataset.py10
-rw-r--r--modules/textual_inversion/textual_inversion.py16
-rw-r--r--modules/ui.py66
-rw-r--r--modules/ui_extensions.py28
-rw-r--r--requirements.txt3
-rw-r--r--requirements_versions.txt3
-rw-r--r--script.js7
-rw-r--r--scripts/prompt_matrix.py17
-rw-r--r--scripts/sd_upscale.py15
-rw-r--r--scripts/xy_grid.py44
-rw-r--r--style.css2
-rw-r--r--webui-macos-env.sh6
-rw-r--r--webui.py7
44 files changed, 2147 insertions, 308 deletions
diff --git a/README.md b/README.md
index 8a4ffade..556000fb 100644
--- a/README.md
+++ b/README.md
@@ -82,8 +82,8 @@ Check the [custom scripts](https://github.com/AUTOMATIC1111/stable-diffusion-web
- Use VAEs
- Estimated completion time in progress bar
- API
-- Support for dedicated [inpainting model](https://github.com/runwayml/stable-diffusion#inpainting-with-stable-diffusion) by RunwayML.
-- via extension: [Aesthetic Gradients](https://github.com/AUTOMATIC1111/stable-diffusion-webui-aesthetic-gradients), a way to generate images with a specific aesthetic by using clip images embds (implementation of [https://github.com/vicgalle/stable-diffusion-aesthetic-gradients](https://github.com/vicgalle/stable-diffusion-aesthetic-gradients))
+- Support for dedicated [inpainting model](https://github.com/runwayml/stable-diffusion#inpainting-with-stable-diffusion) by RunwayML.
+- via extension: [Aesthetic Gradients](https://github.com/AUTOMATIC1111/stable-diffusion-webui-aesthetic-gradients), a way to generate images with a specific aesthetic by using clip images embeds (implementation of [https://github.com/vicgalle/stable-diffusion-aesthetic-gradients](https://github.com/vicgalle/stable-diffusion-aesthetic-gradients))
- [Stable Diffusion 2.0](https://github.com/Stability-AI/stablediffusion) support - see [wiki](https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Features#stable-diffusion-20) for instructions
## Installation and Running
@@ -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/extensions-builtin/LDSR/ldsr_model_arch.py b/extensions-builtin/LDSR/ldsr_model_arch.py
index 90e0a2f0..8b048ae0 100644
--- a/extensions-builtin/LDSR/ldsr_model_arch.py
+++ b/extensions-builtin/LDSR/ldsr_model_arch.py
@@ -11,24 +11,41 @@ from omegaconf import OmegaConf
from ldm.models.diffusion.ddim import DDIMSampler
from ldm.util import instantiate_from_config, ismap
+from modules import shared, sd_hijack
warnings.filterwarnings("ignore", category=UserWarning)
+cached_ldsr_model: torch.nn.Module = None
+
# Create LDSR Class
class LDSR:
def load_model_from_config(self, half_attention):
- print(f"Loading model from {self.modelPath}")
- pl_sd = torch.load(self.modelPath, map_location="cpu")
- sd = pl_sd["state_dict"]
- config = OmegaConf.load(self.yamlPath)
- model = instantiate_from_config(config.model)
- model.load_state_dict(sd, strict=False)
- model.cuda()
- if half_attention:
- model = model.half()
-
- model.eval()
+ global cached_ldsr_model
+
+ if shared.opts.ldsr_cached and cached_ldsr_model is not None:
+ print(f"Loading model from cache")
+ model: torch.nn.Module = cached_ldsr_model
+ else:
+ print(f"Loading model from {self.modelPath}")
+ 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: torch.nn.Module = instantiate_from_config(config.model)
+ model.load_state_dict(sd, strict=False)
+ model = model.to(shared.device)
+ if half_attention:
+ model = model.half()
+ if shared.cmd_opts.opt_channelslast:
+ model = model.to(memory_format=torch.channels_last)
+
+ sd_hijack.model_hijack.hijack(model) # apply optimization
+ model.eval()
+
+ if shared.opts.ldsr_cached:
+ cached_ldsr_model = model
+
return {"model": model}
def __init__(self, model_path, yaml_path):
@@ -93,7 +110,8 @@ class LDSR:
down_sample_method = 'Lanczos'
gc.collect()
- torch.cuda.empty_cache()
+ if torch.cuda.is_available:
+ torch.cuda.empty_cache()
im_og = image
width_og, height_og = im_og.size
@@ -130,7 +148,9 @@ class LDSR:
del model
gc.collect()
- torch.cuda.empty_cache()
+ if torch.cuda.is_available:
+ torch.cuda.empty_cache()
+
return a
@@ -145,7 +165,7 @@ def get_cond(selected_path):
c = rearrange(c, '1 c h w -> 1 h w c')
c = 2. * c - 1.
- c = c.to(torch.device("cuda"))
+ c = c.to(shared.device)
example["LR_image"] = c
example["image"] = c_up
diff --git a/extensions-builtin/LDSR/scripts/ldsr_model.py b/extensions-builtin/LDSR/scripts/ldsr_model.py
index 1cef29a4..29d5f94e 100644
--- a/extensions-builtin/LDSR/scripts/ldsr_model.py
+++ b/extensions-builtin/LDSR/scripts/ldsr_model.py
@@ -7,7 +7,7 @@ from basicsr.utils.download_util import load_file_from_url
from modules.upscaler import Upscaler, UpscalerData
from ldsr_model_arch import LDSR
from modules import shared, script_callbacks
-import sd_hijack_autoencoder
+import sd_hijack_autoencoder, sd_hijack_ddpm_v1
class UpscalerLDSR(Upscaler):
@@ -59,6 +59,7 @@ 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")))
+ shared.opts.add_option("ldsr_cached", shared.OptionInfo(False, "Cache LDSR model in memory", gr.Checkbox, {"interactive": True}, section=('upscaling', "Upscaling")))
script_callbacks.on_ui_settings(on_ui_settings)
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/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..eccfb0f9
--- /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 = '';
+ }
+}
+
+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);
diff --git a/javascript/contextMenus.js b/javascript/contextMenus.js
index fe67c42e..11bcce1b 100644
--- a/javascript/contextMenus.js
+++ b/javascript/contextMenus.js
@@ -9,7 +9,7 @@ contextMenuInit = function(){
function showContextMenu(event,element,menuEntries){
let posx = event.clientX + document.body.scrollLeft + document.documentElement.scrollLeft;
- let posy = event.clientY + document.body.scrollTop + document.documentElement.scrollTop;
+ let posy = event.clientY + document.body.scrollTop + document.documentElement.scrollTop;
let oldMenu = gradioApp().querySelector('#context-menu')
if(oldMenu){
@@ -61,15 +61,15 @@ contextMenuInit = function(){
}
- function appendContextMenuOption(targetEmementSelector,entryName,entryFunction){
-
- currentItems = menuSpecs.get(targetEmementSelector)
-
+ function appendContextMenuOption(targetElementSelector,entryName,entryFunction){
+
+ currentItems = menuSpecs.get(targetElementSelector)
+
if(!currentItems){
currentItems = []
- menuSpecs.set(targetEmementSelector,currentItems);
+ menuSpecs.set(targetElementSelector,currentItems);
}
- let newItem = {'id':targetEmementSelector+'_'+uid(),
+ let newItem = {'id':targetElementSelector+'_'+uid(),
'name':entryName,
'func':entryFunction,
'isNew':true}
@@ -97,7 +97,7 @@ contextMenuInit = function(){
if(source.id && source.id.indexOf('check_progress')>-1){
return
}
-
+
let oldMenu = gradioApp().querySelector('#context-menu')
if(oldMenu){
oldMenu.remove()
@@ -117,7 +117,7 @@ contextMenuInit = function(){
})
});
eventListenerApplied=true
-
+
}
return [appendContextMenuOption, removeContextMenuOption, addContextMenuEventListener]
@@ -152,8 +152,8 @@ addContextMenuEventListener = initResponse[2];
generateOnRepeat('#img2img_generate','#img2img_interrupt');
})
- let cancelGenerateForever = function(){
- clearInterval(window.generateOnRepeatInterval)
+ let cancelGenerateForever = function(){
+ clearInterval(window.generateOnRepeatInterval)
}
appendContextMenuOption('#txt2img_interrupt','Cancel generate forever',cancelGenerateForever)
@@ -162,7 +162,7 @@ addContextMenuEventListener = initResponse[2];
appendContextMenuOption('#img2img_generate', 'Cancel generate forever',cancelGenerateForever)
appendContextMenuOption('#roll','Roll three',
- function(){
+ function(){
let rollbutton = get_uiCurrentTabContent().querySelector('#roll');
setTimeout(function(){rollbutton.click()},100)
setTimeout(function(){rollbutton.click()},200)
diff --git a/javascript/hints.js b/javascript/hints.js
index 57db35be..28e1651b 100644
--- a/javascript/hints.js
+++ b/javascript/hints.js
@@ -17,6 +17,7 @@ titles = {
"\u2199\ufe0f": "Read generation parameters from prompt or last generation if prompt is empty into user interface.",
"\u{1f4c2}": "Open images output directory",
"\u{1f4be}": "Save style",
+ "\U0001F5D1": "Clear prompt",
"\u{1f4cb}": "Apply selected styles to current prompt",
"Inpaint a part of image": "Draw a mask over an image, and the script will regenerate the masked area with content according to prompt",
diff --git a/javascript/progressbar.js b/javascript/progressbar.js
index d58737c4..d6323ed9 100644
--- a/javascript/progressbar.js
+++ b/javascript/progressbar.js
@@ -3,7 +3,7 @@ global_progressbars = {}
galleries = {}
galleryObservers = {}
-// this tracks laumnches of window.setTimeout for progressbar to prevent starting a new timeout when the previous is still running
+// this tracks launches of window.setTimeout for progressbar to prevent starting a new timeout when the previous is still running
timeoutIds = {}
function check_progressbar(id_part, id_progressbar, id_progressbar_span, id_skip, id_interrupt, id_preview, id_gallery){
@@ -20,21 +20,21 @@ function check_progressbar(id_part, id_progressbar, id_progressbar_span, id_skip
var skip = id_skip ? gradioApp().getElementById(id_skip) : null
var interrupt = gradioApp().getElementById(id_interrupt)
-
+
if(opts.show_progress_in_title && progressbar && progressbar.offsetParent){
if(progressbar.innerText){
let newtitle = '[' + progressbar.innerText.trim() + '] Stable Diffusion';
if(document.title != newtitle){
- document.title = newtitle;
+ document.title = newtitle;
}
}else{
let newtitle = 'Stable Diffusion'
if(document.title != newtitle){
- document.title = newtitle;
+ document.title = newtitle;
}
}
}
-
+
if(progressbar!= null && progressbar != global_progressbars[id_progressbar]){
global_progressbars[id_progressbar] = progressbar
@@ -63,7 +63,7 @@ function check_progressbar(id_part, id_progressbar, id_progressbar_span, id_skip
skip.style.display = "none"
}
interrupt.style.display = "none"
-
+
//disconnect observer once generation finished, so user can close selected image if they want
if (galleryObservers[id_gallery]) {
galleryObservers[id_gallery].disconnect();
diff --git a/javascript/ui.js b/javascript/ui.js
index 2ca66d79..587dd782 100644
--- a/javascript/ui.js
+++ b/javascript/ui.js
@@ -100,7 +100,7 @@ function create_submit_args(args){
// As it is currently, txt2img and img2img send back the previous output args (txt2img_gallery, generation_info, html_info) whenever you generate a new image.
// This can lead to uploading a huge gallery of previously generated images, which leads to an unnecessary delay between submitting and beginning to generate.
- // I don't know why gradio is seding outputs along with inputs, but we can prevent sending the image gallery here, which seems to be an issue for some.
+ // I don't know why gradio is sending outputs along with inputs, but we can prevent sending the image gallery here, which seems to be an issue for some.
// If gradio at some point stops sending outputs, this may break something
if(Array.isArray(res[res.length - 3])){
res[res.length - 3] = null
@@ -131,6 +131,15 @@ function ask_for_style_name(_, prompt_text, negative_prompt_text) {
return [name_, prompt_text, negative_prompt_text]
}
+function confirm_clear_prompt(prompt, negative_prompt) {
+ if(confirm("Delete prompt?")) {
+ prompt = ""
+ negative_prompt = ""
+ }
+
+ return [prompt, negative_prompt]
+}
+
opts = {}
diff --git a/launch.py b/launch.py
index 0e1bbaf2..af0d418b 100644
--- a/launch.py
+++ b/launch.py
@@ -157,7 +157,7 @@ def run_extensions_installers(settings_file):
run_extension_installer(os.path.join(dir_extensions, dirname_extension))
-def prepare_enviroment():
+def prepare_environment():
torch_command = os.environ.get('TORCH_COMMAND', "pip install torch==1.12.1+cu113 torchvision==0.13.1+cu113 --extra-index-url https://download.pytorch.org/whl/cu113")
requirements_file = os.environ.get('REQS_FILE', "requirements_versions.txt")
commandline_args = os.environ.get('COMMANDLINE_ARGS', "")
@@ -291,5 +291,5 @@ def start():
if __name__ == "__main__":
- prepare_enviroment()
+ prepare_environment()
start()
diff --git a/modules/api/api.py b/modules/api/api.py
index 54ee7cb0..33845045 100644
--- a/modules/api/api.py
+++ b/modules/api/api.py
@@ -67,10 +67,10 @@ def encode_pil_to_base64(image):
class Api:
def __init__(self, app: FastAPI, queue_lock: Lock):
if shared.cmd_opts.api_auth:
- self.credenticals = dict()
+ self.credentials = dict()
for auth in shared.cmd_opts.api_auth.split(","):
user, password = auth.split(":")
- self.credenticals[user] = password
+ self.credentials[user] = password
self.router = APIRouter()
self.app = app
@@ -93,7 +93,7 @@ class Api:
self.add_api_route("/sdapi/v1/hypernetworks", self.get_hypernetworks, methods=["GET"], response_model=List[HypernetworkItem])
self.add_api_route("/sdapi/v1/face-restorers", self.get_face_restorers, methods=["GET"], response_model=List[FaceRestorerItem])
self.add_api_route("/sdapi/v1/realesrgan-models", self.get_realesrgan_models, methods=["GET"], response_model=List[RealesrganItem])
- self.add_api_route("/sdapi/v1/prompt-styles", self.get_promp_styles, methods=["GET"], response_model=List[PromptStyleItem])
+ self.add_api_route("/sdapi/v1/prompt-styles", self.get_prompt_styles, methods=["GET"], response_model=List[PromptStyleItem])
self.add_api_route("/sdapi/v1/artist-categories", self.get_artists_categories, methods=["GET"], response_model=List[str])
self.add_api_route("/sdapi/v1/artists", self.get_artists, methods=["GET"], response_model=List[ArtistItem])
@@ -102,9 +102,9 @@ class Api:
return self.app.add_api_route(path, endpoint, dependencies=[Depends(self.auth)], **kwargs)
return self.app.add_api_route(path, endpoint, **kwargs)
- def auth(self, credenticals: HTTPBasicCredentials = Depends(HTTPBasic())):
- if credenticals.username in self.credenticals:
- if compare_digest(credenticals.password, self.credenticals[credenticals.username]):
+ def auth(self, credentials: HTTPBasicCredentials = Depends(HTTPBasic())):
+ if credentials.username in self.credentials:
+ if compare_digest(credentials.password, self.credentials[credentials.username]):
return True
raise HTTPException(status_code=401, detail="Incorrect username or password", headers={"WWW-Authenticate": "Basic"})
@@ -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()
@@ -244,7 +239,7 @@ class Api:
def interrogateapi(self, interrogatereq: InterrogateRequest):
image_b64 = interrogatereq.image
if image_b64 is None:
- raise HTTPException(status_code=404, detail="Image not found")
+ raise HTTPException(status_code=404, detail="Image not found")
img = decode_base64_to_image(image_b64)
img = img.convert('RGB')
@@ -257,7 +252,7 @@ class Api:
processed = deepbooru.model.tag(img)
else:
raise HTTPException(status_code=404, detail="Model not found")
-
+
return InterrogateResponse(caption=processed)
def interruptapi(self):
@@ -313,7 +308,7 @@ class Api:
def get_realesrgan_models(self):
return [{"name":x.name,"path":x.data_path, "scale":x.scale} for x in get_realesrgan_models(None)]
- def get_promp_styles(self):
+ def get_prompt_styles(self):
styleList = []
for k in shared.prompt_styles.styles:
style = shared.prompt_styles.styles[k]
diff --git a/modules/api/models.py b/modules/api/models.py
index f77951fc..a22bc6b3 100644
--- a/modules/api/models.py
+++ b/modules/api/models.py
@@ -128,7 +128,7 @@ class ExtrasBaseRequest(BaseModel):
upscaling_resize: float = Field(default=2, title="Upscaling Factor", ge=1, le=4, description="By how much to upscale the image, only used when resize_mode=0.")
upscaling_resize_w: int = Field(default=512, title="Target Width", ge=1, description="Target width for the upscaler to hit. Only used when resize_mode=1.")
upscaling_resize_h: int = Field(default=512, title="Target Height", ge=1, description="Target height for the upscaler to hit. Only used when resize_mode=1.")
- upscaling_crop: bool = Field(default=True, title="Crop to fit", description="Should the upscaler crop the image to fit in the choosen size?")
+ upscaling_crop: bool = Field(default=True, title="Crop to fit", description="Should the upscaler crop the image to fit in the chosen size?")
upscaler_1: str = Field(default="None", title="Main upscaler", description=f"The name of the main upscaler to use, it has to be one of this list: {' , '.join([x.name for x in sd_upscalers])}")
upscaler_2: str = Field(default="None", title="Secondary upscaler", description=f"The name of the secondary upscaler to use, it has to be one of this list: {' , '.join([x.name for x in sd_upscalers])}")
extras_upscaler_2_visibility: float = Field(default=0, title="Secondary upscaler visibility", ge=0, le=1, allow_inf_nan=False, description="Sets the visibility of secondary upscaler, values should be between 0 and 1.")
diff --git a/modules/devices.py b/modules/devices.py
index f8cffae1..800510b7 100644
--- a/modules/devices.py
+++ b/modules/devices.py
@@ -125,7 +125,16 @@ def layer_norm_fix(*args, **kwargs):
return orig_layer_norm(*args, **kwargs)
+# MPS workaround for https://github.com/pytorch/pytorch/issues/90532
+orig_tensor_numpy = torch.Tensor.numpy
+def numpy_fix(self, *args, **kwargs):
+ if self.requires_grad:
+ self = self.detach()
+ return orig_tensor_numpy(self, *args, **kwargs)
+
+
# PyTorch 1.13 doesn't need these fixes but unfortunately is slower and has regressions that prevent training from working
if has_mps() and version.parse(torch.__version__) < version.parse("1.13"):
torch.Tensor.to = tensor_to_fix
torch.nn.functional.layer_norm = layer_norm_fix
+ torch.Tensor.numpy = numpy_fix
diff --git a/modules/extras.py b/modules/extras.py
index bc349d5e..6fa7d856 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:
@@ -188,13 +188,19 @@ def run_extras(extras_mode, resize_mode, image, image_folder, input_dir, output_
for op in extras_ops:
image, info = op(image, info)
- if opts.use_original_name_batch and image_name != None:
+ if opts.use_original_name_batch and image_name is not None:
basename = os.path.splitext(os.path.basename(image_name))[0]
else:
basename = ''
+ # Add upscaler name as a suffix.
+ suffix = f"-{shared.sd_upscalers[extras_upscaler_1].name}" if shared.opts.use_upscaler_name_as_suffix else ""
+ # Add second upscaler if applicable.
+ if suffix and extras_upscaler_2 and extras_upscaler_2_visibility:
+ suffix += f"-{shared.sd_upscalers[extras_upscaler_2].name}"
+
images.save_image(image, path=outpath, basename=basename, seed=None, prompt=None, extension=opts.samples_format, info=info, short_filename=True,
- no_prompt=True, grid=False, pnginfo_section_name="extras", existing_info=existing_pnginfo, forced_filename=None)
+ no_prompt=True, grid=False, pnginfo_section_name="extras", existing_info=existing_pnginfo, forced_filename=None, suffix=suffix)
if opts.enable_pnginfo:
image.info = existing_pnginfo
@@ -234,7 +240,7 @@ 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)
@@ -246,30 +252,25 @@ 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)
+ tertiary_model_info = sd_models.checkpoints_list.get(tertiary_model_name, None)
result_is_inpainting_model = False
- 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
-
theta_funcs = {
"Weighted sum": (None, weighted_sum),
"Add difference": (get_difference, add_difference),
}
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:
@@ -277,7 +278,12 @@ 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:
@@ -307,6 +313,7 @@ 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
@@ -332,5 +339,5 @@ 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/generation_parameters_copypaste.py b/modules/generation_parameters_copypaste.py
index 44fe1a6c..565e342d 100644
--- a/modules/generation_parameters_copypaste.py
+++ b/modules/generation_parameters_copypaste.py
@@ -77,6 +77,7 @@ def integrate_settings_paste_fields(component_dict):
'inpainting_mask_weight': 'Conditional mask weight',
'sd_model_checkpoint': 'Model hash',
'eta_noise_seed_delta': 'ENSD',
+ 'initial_noise_multiplier': 'Noise multiplier',
}
settings_paste_fields = [
(component_dict[k], lambda d, k=k, v=v: ui.apply_setting(k, d.get(v, None)))
@@ -121,8 +122,7 @@ def run_bind():
if send_generate_info and paste_fields[tab]["fields"] is not None:
if send_generate_info in paste_fields:
- paste_field_names = ['Prompt', 'Negative prompt', 'Steps', 'Face restoration', 'Size-1', 'Size-2'] + (["Seed"] if shared.opts.send_seed else [])
-
+ paste_field_names = ['Prompt', 'Negative prompt', 'Steps', 'Face restoration'] + (['Size-1', 'Size-2'] if shared.opts.send_size else []) + (["Seed"] if shared.opts.send_seed else [])
button.click(
fn=lambda *x: x,
inputs=[field for field, name in paste_fields[send_generate_info]["fields"] if name in paste_field_names],
diff --git a/modules/images.py b/modules/images.py
index 08a72e67..809ad9f7 100644
--- a/modules/images.py
+++ b/modules/images.py
@@ -136,8 +136,19 @@ def draw_grid_annotations(im, width, height, hor_texts, ver_texts):
lines.append(word)
return lines
- def draw_texts(drawing, draw_x, draw_y, lines):
+ def get_font(fontsize):
+ try:
+ return ImageFont.truetype(opts.font or Roboto, fontsize)
+ except Exception:
+ return ImageFont.truetype(Roboto, fontsize)
+
+ def draw_texts(drawing, draw_x, draw_y, lines, initial_fnt, initial_fontsize):
for i, line in enumerate(lines):
+ fnt = initial_fnt
+ fontsize = initial_fontsize
+ while drawing.multiline_textsize(line.text, font=fnt)[0] > line.allowed_width and fontsize > 0:
+ fontsize -= 1
+ fnt = get_font(fontsize)
drawing.multiline_text((draw_x, draw_y + line.size[1] / 2), line.text, font=fnt, fill=color_active if line.is_active else color_inactive, anchor="mm", align="center")
if not line.is_active:
@@ -148,10 +159,7 @@ def draw_grid_annotations(im, width, height, hor_texts, ver_texts):
fontsize = (width + height) // 25
line_spacing = fontsize // 2
- try:
- fnt = ImageFont.truetype(opts.font or Roboto, fontsize)
- except Exception:
- fnt = ImageFont.truetype(Roboto, fontsize)
+ fnt = get_font(fontsize)
color_active = (0, 0, 0)
color_inactive = (153, 153, 153)
@@ -178,6 +186,7 @@ def draw_grid_annotations(im, width, height, hor_texts, ver_texts):
for line in texts:
bbox = calc_d.multiline_textbbox((0, 0), line.text, font=fnt)
line.size = (bbox[2] - bbox[0], bbox[3] - bbox[1])
+ line.allowed_width = allowed_width
hor_text_heights = [sum([line.size[1] + line_spacing for line in lines]) - line_spacing for lines in hor_texts]
ver_text_heights = [sum([line.size[1] + line_spacing for line in lines]) - line_spacing * len(lines) for lines in
@@ -194,13 +203,13 @@ def draw_grid_annotations(im, width, height, hor_texts, ver_texts):
x = pad_left + width * col + width / 2
y = pad_top / 2 - hor_text_heights[col] / 2
- draw_texts(d, x, y, hor_texts[col])
+ draw_texts(d, x, y, hor_texts[col], fnt, fontsize)
for row in range(rows):
x = pad_left / 2
y = pad_top + height * row + height / 2 - ver_text_heights[row] / 2
- draw_texts(d, x, y, ver_texts[row])
+ draw_texts(d, x, y, ver_texts[row], fnt, fontsize)
return result
@@ -429,7 +438,7 @@ def save_image(image, path, basename, seed=None, prompt=None, extension='png', i
The directory to save the image. Note, the option `save_to_dirs` will make the image to be saved into a sub directory.
basename (`str`):
The base filename which will be applied to `filename pattern`.
- seed, prompt, short_filename,
+ seed, prompt, short_filename,
extension (`str`):
Image file extension, default is `png`.
pngsectionname (`str`):
@@ -501,30 +510,39 @@ def save_image(image, path, basename, seed=None, prompt=None, extension='png', i
image = params.image
fullfn = params.filename
info = params.pnginfo.get(pnginfo_section_name, None)
- fullfn_without_extension, extension = os.path.splitext(params.filename)
- def exif_bytes():
- return piexif.dump({
- "Exif": {
- piexif.ExifIFD.UserComment: piexif.helper.UserComment.dump(info or "", encoding="unicode")
- },
- })
+ def _atomically_save_image(image_to_save, filename_without_extension, extension):
+ # save image with .tmp extension to avoid race condition when another process detects new image in the directory
+ temp_file_path = filename_without_extension + ".tmp"
+ image_format = Image.registered_extensions()[extension]
- if extension.lower() == '.png':
- pnginfo_data = PngImagePlugin.PngInfo()
- if opts.enable_pnginfo:
- for k, v in params.pnginfo.items():
- pnginfo_data.add_text(k, str(v))
+ if extension.lower() == '.png':
+ pnginfo_data = PngImagePlugin.PngInfo()
+ if opts.enable_pnginfo:
+ for k, v in params.pnginfo.items():
+ pnginfo_data.add_text(k, str(v))
- image.save(fullfn, quality=opts.jpeg_quality, pnginfo=pnginfo_data)
+ image_to_save.save(temp_file_path, format=image_format, quality=opts.jpeg_quality, pnginfo=pnginfo_data)
- elif extension.lower() in (".jpg", ".jpeg", ".webp"):
- image.save(fullfn, quality=opts.jpeg_quality)
+ elif extension.lower() in (".jpg", ".jpeg", ".webp"):
+ image_to_save.save(temp_file_path, format=image_format, quality=opts.jpeg_quality)
- if opts.enable_pnginfo and info is not None:
- piexif.insert(exif_bytes(), fullfn)
- else:
- image.save(fullfn, quality=opts.jpeg_quality)
+ if opts.enable_pnginfo and info is not None:
+ exif_bytes = piexif.dump({
+ "Exif": {
+ piexif.ExifIFD.UserComment: piexif.helper.UserComment.dump(info or "", encoding="unicode")
+ },
+ })
+
+ piexif.insert(exif_bytes, temp_file_path)
+ else:
+ image_to_save.save(temp_file_path, format=image_format, quality=opts.jpeg_quality)
+
+ # atomically rename the file with correct extension
+ os.replace(temp_file_path, filename_without_extension + extension)
+
+ fullfn_without_extension, extension = os.path.splitext(params.filename)
+ _atomically_save_image(image, fullfn_without_extension, extension)
image.already_saved_as = fullfn
@@ -538,9 +556,7 @@ def save_image(image, path, basename, seed=None, prompt=None, extension='png', i
elif oversize:
image = image.resize((image.width * target_side_length // image.height, target_side_length), LANCZOS)
- image.save(fullfn_without_extension + ".jpg", quality=opts.jpeg_quality)
- if opts.enable_pnginfo and info is not None:
- piexif.insert(exif_bytes(), fullfn_without_extension + ".jpg")
+ _atomically_save_image(image, fullfn_without_extension, ".jpg")
if opts.save_txt and info is not None:
txt_fullfn = f"{fullfn_without_extension}.txt"
@@ -583,7 +599,7 @@ def read_info_from_image(image):
Negative prompt: {json_info["uc"]}
Steps: {json_info["steps"]}, Sampler: {sampler}, CFG scale: {json_info["scale"]}, Seed: {json_info["seed"]}, Size: {image.width}x{image.height}, Clip skip: 2, ENSD: 31337"""
except Exception:
- print(f"Error parsing NovelAI iamge generation parameters:", file=sys.stderr)
+ print(f"Error parsing NovelAI image generation parameters:", file=sys.stderr)
print(traceback.format_exc(), file=sys.stderr)
return geninfo, items
@@ -606,3 +622,14 @@ def image_data(data):
pass
return '', None
+
+
+def flatten(img, bgcolor):
+ """replaces transparency with bgcolor (example: "#ffffff"), returning an RGB mode image with no transparency"""
+
+ if img.mode == "RGBA":
+ background = Image.new('RGBA', img.size, bgcolor)
+ background.paste(img, mask=img)
+ img = background
+
+ return img.convert('RGB')
diff --git a/modules/import_hook.py b/modules/import_hook.py
new file mode 100644
index 00000000..28c67dfa
--- /dev/null
+++ b/modules/import_hook.py
@@ -0,0 +1,5 @@
+import sys
+
+# this will break any attempt to import xformers which will prevent stability diffusion repo from trying to use it
+if "--xformers" not in "".join(sys.argv):
+ sys.modules["xformers"] = None
diff --git a/modules/lowvram.py b/modules/lowvram.py
index aa464a95..042a0254 100644
--- a/modules/lowvram.py
+++ b/modules/lowvram.py
@@ -55,18 +55,20 @@ def setup_for_low_vram(sd_model, use_medvram):
if hasattr(sd_model.cond_stage_model, 'model'):
sd_model.cond_stage_model.transformer = sd_model.cond_stage_model.model
- # remove three big modules, cond, first_stage, and unet from the model and then
+ # remove four big modules, cond, first_stage, depth (if applicable), and unet from the model and then
# send the model to GPU. Then put modules back. the modules will be in CPU.
- stored = sd_model.cond_stage_model.transformer, sd_model.first_stage_model, sd_model.model
- sd_model.cond_stage_model.transformer, sd_model.first_stage_model, sd_model.model = None, None, None
+ stored = sd_model.cond_stage_model.transformer, sd_model.first_stage_model, getattr(sd_model, 'depth_model', None), sd_model.model
+ sd_model.cond_stage_model.transformer, sd_model.first_stage_model, sd_model.depth_model, sd_model.model = None, None, None, None
sd_model.to(devices.device)
- sd_model.cond_stage_model.transformer, sd_model.first_stage_model, sd_model.model = stored
+ sd_model.cond_stage_model.transformer, sd_model.first_stage_model, sd_model.depth_model, sd_model.model = stored
- # register hooks for those the first two models
+ # register hooks for those the first three models
sd_model.cond_stage_model.transformer.register_forward_pre_hook(send_me_to_gpu)
sd_model.first_stage_model.register_forward_pre_hook(send_me_to_gpu)
sd_model.first_stage_model.encode = first_stage_model_encode_wrap
sd_model.first_stage_model.decode = first_stage_model_decode_wrap
+ if sd_model.depth_model:
+ sd_model.depth_model.register_forward_pre_hook(send_me_to_gpu)
parents[sd_model.cond_stage_model.transformer] = sd_model.cond_stage_model
if hasattr(sd_model.cond_stage_model, 'model'):
diff --git a/modules/ngrok.py b/modules/ngrok.py
index 64c9a3c2..3df2c06b 100644
--- a/modules/ngrok.py
+++ b/modules/ngrok.py
@@ -2,7 +2,7 @@ from pyngrok import ngrok, conf, exception
def connect(token, port, region):
account = None
- if token == None:
+ if token is None:
token = 'None'
else:
if ':' in token:
@@ -14,7 +14,7 @@ def connect(token, port, region):
auth_token=token, region=region
)
try:
- if account == None:
+ if account is None:
public_url = ngrok.connect(port, pyngrok_config=config, bind_tls=True).public_url
else:
public_url = ngrok.connect(port, pyngrok_config=config, bind_tls=True, auth=account).public_url
diff --git a/modules/processing.py b/modules/processing.py
index ab5a34d0..75b0067c 100644
--- a/modules/processing.py
+++ b/modules/processing.py
@@ -13,15 +13,21 @@ from skimage import exposure
from typing import Any, Dict, List, Optional
import modules.sd_hijack
-from modules import devices, prompt_parser, masking, sd_samplers, lowvram, generation_parameters_copypaste
+from modules import devices, prompt_parser, masking, sd_samplers, lowvram, generation_parameters_copypaste, script_callbacks
from modules.sd_hijack import model_hijack
from modules.shared import opts, cmd_opts, state
import modules.shared as shared
import modules.face_restoration
import modules.images as images
import modules.styles
+import modules.sd_models as sd_models
+import modules.sd_vae as sd_vae
import logging
+from ldm.data.util import AddMiDaS
+from ldm.models.diffusion.ddpm import LatentDepth2ImageDiffusion
+from einops import repeat, rearrange
+from blendmodes.blend import blendLayers, BlendType
# 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
@@ -34,17 +40,19 @@ def setup_color_correction(image):
return correction_target
-def apply_color_correction(correction, image):
+def apply_color_correction(correction, original_image):
logging.info("Applying color correction.")
image = Image.fromarray(cv2.cvtColor(exposure.match_histograms(
cv2.cvtColor(
- np.asarray(image),
+ np.asarray(original_image),
cv2.COLOR_RGB2LAB
),
correction,
channel_axis=2
), cv2.COLOR_LAB2RGB).astype("uint8"))
-
+
+ image = blendLayers(image, original_image, BlendType.LUMINOSITY)
+
return image
@@ -72,7 +80,7 @@ class StableDiffusionProcessing():
"""
The first set of paramaters: sd_models -> do_not_reload_embeddings represent the minimum required to create a StableDiffusionProcessing
"""
- def __init__(self, sd_model=None, outpath_samples=None, outpath_grids=None, prompt: str = "", styles: List[str] = None, seed: int = -1, subseed: int = -1, subseed_strength: float = 0, seed_resize_from_h: int = -1, seed_resize_from_w: int = -1, seed_enable_extras: bool = True, sampler_name: str = None, batch_size: int = 1, n_iter: int = 1, steps: int = 50, cfg_scale: float = 7.0, width: int = 512, height: int = 512, restore_faces: bool = False, tiling: bool = False, do_not_save_samples: bool = False, do_not_save_grid: bool = False, extra_generation_params: Dict[Any, Any] = None, overlay_images: Any = None, negative_prompt: str = None, eta: float = None, do_not_reload_embeddings: bool = False, denoising_strength: float = 0, ddim_discretize: str = None, s_churn: float = 0.0, s_tmax: float = None, s_tmin: float = 0.0, s_noise: float = 1.0, override_settings: Dict[str, Any] = None, sampler_index: int = None):
+ def __init__(self, sd_model=None, outpath_samples=None, outpath_grids=None, prompt: str = "", styles: List[str] = None, seed: int = -1, subseed: int = -1, subseed_strength: float = 0, seed_resize_from_h: int = -1, seed_resize_from_w: int = -1, seed_enable_extras: bool = True, sampler_name: str = None, batch_size: int = 1, n_iter: int = 1, steps: int = 50, cfg_scale: float = 7.0, width: int = 512, height: int = 512, restore_faces: bool = False, tiling: bool = False, do_not_save_samples: bool = False, do_not_save_grid: bool = False, extra_generation_params: Dict[Any, Any] = None, overlay_images: Any = None, negative_prompt: str = None, eta: float = None, do_not_reload_embeddings: bool = False, denoising_strength: float = 0, ddim_discretize: str = None, s_churn: float = 0.0, s_tmax: float = None, s_tmin: float = 0.0, s_noise: float = 1.0, override_settings: Dict[str, Any] = None, override_settings_restore_afterwards: bool = True, sampler_index: int = None):
if sampler_index is not None:
print("sampler_index argument for StableDiffusionProcessing does not do anything; use sampler_name", file=sys.stderr)
@@ -113,6 +121,7 @@ class StableDiffusionProcessing():
self.s_tmax = s_tmax or float('inf') # not representable as a standard ui option
self.s_noise = s_noise or opts.s_noise
self.override_settings = {k: v for k, v in (override_settings or {}).items() if k not in shared.restricted_opts}
+ self.override_settings_restore_afterwards = override_settings_restore_afterwards
self.is_using_inpainting_conditioning = False
if not seed_enable_extras:
@@ -142,19 +151,34 @@ class StableDiffusionProcessing():
# The "masked-image" in this case will just be all zeros since the entire image is masked.
image_conditioning = torch.zeros(x.shape[0], 3, height, width, device=x.device)
- image_conditioning = self.sd_model.get_first_stage_encoding(self.sd_model.encode_first_stage(image_conditioning))
+ image_conditioning = self.sd_model.get_first_stage_encoding(self.sd_model.encode_first_stage(image_conditioning))
# Add the fake full 1s mask to the first dimension.
image_conditioning = torch.nn.functional.pad(image_conditioning, (0, 0, 0, 0, 1, 0), value=1.0)
- image_conditioning = image_conditioning.to(x.dtype)
+ image_conditioning = image_conditioning.to(x.dtype)
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
@@ -179,7 +203,7 @@ class StableDiffusionProcessing():
source_image * (1.0 - conditioning_mask),
getattr(self, "inpainting_mask_weight", shared.opts.inpainting_mask_weight)
)
-
+
# Encode the new masked image using first stage of network.
conditioning_image = self.sd_model.get_first_stage_encoding(self.sd_model.encode_first_stage(conditioning_image))
@@ -191,6 +215,18 @@ 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
@@ -424,15 +460,21 @@ def process_images(p: StableDiffusionProcessing) -> Processed:
try:
for k, v in p.override_settings.items():
- setattr(opts, k, v) # we don't call onchange for simplicity which makes changing model impossible
- if k == 'sd_hypernetwork': shared.reload_hypernetworks() # make onchange call for changing hypernet since it is relatively fast to load on-change, while SD models are not
+ setattr(opts, k, v)
+ if k == 'sd_hypernetwork': shared.reload_hypernetworks() # make onchange call for changing hypernet
+ if k == 'sd_model_checkpoint': sd_models.reload_model_weights() # make onchange call for changing SD model
+ if k == 'sd_vae': sd_vae.reload_vae_weights() # make onchange call for changing VAE
res = process_images_inner(p)
- finally: # restore opts to original state
- for k, v in stored_opts.items():
- setattr(opts, k, v)
- if k == 'sd_hypernetwork': shared.reload_hypernetworks()
+ finally:
+ # restore opts to original state
+ if p.override_settings_restore_afterwards:
+ for k, v in stored_opts.items():
+ setattr(opts, k, v)
+ if k == 'sd_hypernetwork': shared.reload_hypernetworks()
+ if k == 'sd_model_checkpoint': sd_models.reload_model_weights()
+ if k == 'sd_vae': sd_vae.reload_vae_weights()
return res
@@ -501,7 +543,7 @@ def process_images_inner(p: StableDiffusionProcessing) -> Processed:
for n in range(p.n_iter):
if state.skipped:
state.skipped = False
-
+
if state.interrupted:
break
@@ -541,9 +583,8 @@ def process_images_inner(p: StableDiffusionProcessing) -> Processed:
devices.torch_gc()
- if opts.filter_nsfw:
- import modules.safety as safety
- x_samples_ddim = modules.safety.censor_batch(x_samples_ddim)
+ if p.scripts is not None:
+ p.scripts.postprocess_batch(p, x_samples_ddim, batch_number=n)
for i, x_sample in enumerate(x_samples_ddim):
x_sample = 255. * np.moveaxis(x_sample.cpu().numpy(), 0, 2)
@@ -577,7 +618,7 @@ def process_images_inner(p: StableDiffusionProcessing) -> Processed:
image.info["parameters"] = text
output_images.append(image)
- del x_samples_ddim
+ del x_samples_ddim
devices.torch_gc()
@@ -669,7 +710,7 @@ class StableDiffusionProcessingTxt2Img(StableDiffusionProcessing):
samples = samples[:, :, self.truncate_y//2:samples.shape[2]-self.truncate_y//2, self.truncate_x//2:samples.shape[3]-self.truncate_x//2]
- """saves image before applying hires fix, if enabled in options; takes as an arguyment either an image or batch with latent space images"""
+ """saves image before applying hires fix, if enabled in options; takes as an argument either an image or batch with latent space images"""
def save_intermediate(image, index):
if not opts.save or self.do_not_save_samples or not opts.save_images_before_highres_fix:
return
@@ -685,7 +726,7 @@ class StableDiffusionProcessingTxt2Img(StableDiffusionProcessing):
samples = torch.nn.functional.interpolate(samples, size=(self.height // opt_f, self.width // opt_f), mode="bilinear")
- # Avoid making the inpainting conditioning unless necessary as
+ # Avoid making the inpainting conditioning unless necessary as
# this does need some extra compute to decode / encode the image again.
if getattr(self, "inpainting_mask_weight", shared.opts.inpainting_mask_weight) < 1.0:
image_conditioning = self.img2img_image_conditioning(decode_first_stage(self.sd_model, samples), samples)
@@ -734,7 +775,7 @@ class StableDiffusionProcessingTxt2Img(StableDiffusionProcessing):
class StableDiffusionProcessingImg2Img(StableDiffusionProcessing):
sampler = None
- def __init__(self, init_images: list=None, resize_mode: int=0, denoising_strength: float=0.75, mask: Any=None, mask_blur: int=4, inpainting_fill: int=0, inpaint_full_res: bool=True, inpaint_full_res_padding: int=0, inpainting_mask_invert: int=0, **kwargs):
+ def __init__(self, init_images: list = None, resize_mode: int = 0, denoising_strength: float = 0.75, mask: Any = None, mask_blur: int = 4, inpainting_fill: int = 0, inpaint_full_res: bool = True, inpaint_full_res_padding: int = 0, inpainting_mask_invert: int = 0, initial_noise_multiplier: float = None, **kwargs):
super().__init__(**kwargs)
self.init_images = init_images
@@ -749,6 +790,7 @@ class StableDiffusionProcessingImg2Img(StableDiffusionProcessing):
self.inpaint_full_res = inpaint_full_res
self.inpaint_full_res_padding = inpaint_full_res_padding
self.inpainting_mask_invert = inpainting_mask_invert
+ self.initial_noise_multiplier = opts.initial_noise_multiplier if initial_noise_multiplier is None else initial_noise_multiplier
self.mask = None
self.nmask = None
self.image_conditioning = None
@@ -793,7 +835,7 @@ class StableDiffusionProcessingImg2Img(StableDiffusionProcessing):
self.color_corrections = []
imgs = []
for img in self.init_images:
- image = img.convert("RGB")
+ image = images.flatten(img, opts.img2img_background_color)
if crop_region is None and self.resize_mode != 3:
image = images.resize_image(self.resize_mode, image, self.width, self.height)
@@ -866,6 +908,10 @@ class StableDiffusionProcessingImg2Img(StableDiffusionProcessing):
def sample(self, conditioning, unconditional_conditioning, seeds, subseeds, subseed_strength, prompts):
x = create_random_tensors([opt_C, self.height // opt_f, self.width // opt_f], seeds=seeds, subseeds=subseeds, subseed_strength=self.subseed_strength, seed_resize_from_h=self.seed_resize_from_h, seed_resize_from_w=self.seed_resize_from_w, p=self)
+ if self.initial_noise_multiplier != 1.0:
+ self.extra_generation_params["Noise multiplier"] = self.initial_noise_multiplier
+ x *= self.initial_noise_multiplier
+
samples = self.sampler.sample_img2img(self, self.init_latent, x, conditioning, unconditional_conditioning, image_conditioning=self.image_conditioning)
if self.mask is not None:
diff --git a/modules/safe.py b/modules/safe.py
index 10460ad0..479c8b86 100644
--- a/modules/safe.py
+++ b/modules/safe.py
@@ -37,16 +37,16 @@ class RestrictedUnpickler(pickle.Unpickler):
if module == 'collections' and name == 'OrderedDict':
return getattr(collections, name)
- if module == 'torch._utils' and name in ['_rebuild_tensor_v2', '_rebuild_parameter']:
+ if module == 'torch._utils' and name in ['_rebuild_tensor_v2', '_rebuild_parameter', '_rebuild_device_tensor_from_numpy']:
return getattr(torch._utils, name)
- if module == 'torch' and name in ['FloatStorage', 'HalfStorage', 'IntStorage', 'LongStorage', 'DoubleStorage', 'ByteStorage']:
+ if module == 'torch' and name in ['FloatStorage', 'HalfStorage', 'IntStorage', 'LongStorage', 'DoubleStorage', 'ByteStorage', 'float32']:
return getattr(torch, name)
if module == 'torch.nn.modules.container' and name in ['ParameterDict']:
return getattr(torch.nn.modules.container, name)
- if module == 'numpy.core.multiarray' and name == 'scalar':
- return numpy.core.multiarray.scalar
- if module == 'numpy' and name == 'dtype':
- return numpy.dtype
+ if module == 'numpy.core.multiarray' and name in ['scalar', '_reconstruct']:
+ return getattr(numpy.core.multiarray, name)
+ if module == 'numpy' and name in ['dtype', 'ndarray']:
+ return getattr(numpy, name)
if module == '_codecs' and name == 'encode':
return encode
if module == "pytorch_lightning.callbacks" and name == 'model_checkpoint':
@@ -80,7 +80,7 @@ def check_pt(filename, extra_handler):
# new pytorch format is a zip file
with zipfile.ZipFile(filename) as z:
check_zip_filenames(filename, z.namelist())
-
+
# find filename of data.pkl in zip file: '<directory name>/data.pkl'
data_pkl_filenames = [f for f in z.namelist() if data_pkl_re.match(f)]
if len(data_pkl_filenames) == 0:
@@ -108,7 +108,7 @@ def load(filename, *args, **kwargs):
def load_with_extra(filename, extra_handler=None, *args, **kwargs):
"""
- this functon is intended to be used by extensions that want to load models with
+ this function is intended to be used by extensions that want to load models with
some extra classes in them that the usual unpickler would find suspicious.
Use the extra_handler argument to specify a function that takes module and field name as text,
diff --git a/modules/safety.py b/modules/safety.py
deleted file mode 100644
index cff4b278..00000000
--- a/modules/safety.py
+++ /dev/null
@@ -1,42 +0,0 @@
-import torch
-from diffusers.pipelines.stable_diffusion.safety_checker import StableDiffusionSafetyChecker
-from transformers import AutoFeatureExtractor
-from PIL import Image
-
-import modules.shared as shared
-
-safety_model_id = "CompVis/stable-diffusion-safety-checker"
-safety_feature_extractor = None
-safety_checker = None
-
-def numpy_to_pil(images):
- """
- Convert a numpy image or a batch of images to a PIL image.
- """
- if images.ndim == 3:
- images = images[None, ...]
- images = (images * 255).round().astype("uint8")
- pil_images = [Image.fromarray(image) for image in images]
-
- return pil_images
-
-# check and replace nsfw content
-def check_safety(x_image):
- global safety_feature_extractor, safety_checker
-
- if safety_feature_extractor is None:
- safety_feature_extractor = AutoFeatureExtractor.from_pretrained(safety_model_id)
- safety_checker = StableDiffusionSafetyChecker.from_pretrained(safety_model_id)
-
- safety_checker_input = safety_feature_extractor(numpy_to_pil(x_image), return_tensors="pt")
- x_checked_image, has_nsfw_concept = safety_checker(images=x_image, clip_input=safety_checker_input.pixel_values)
-
- return x_checked_image, has_nsfw_concept
-
-
-def censor_batch(x):
- x_samples_ddim_numpy = x.cpu().permute(0, 2, 3, 1).numpy()
- x_checked_image, has_nsfw_concept = check_safety(x_samples_ddim_numpy)
- x = torch.from_numpy(x_checked_image).permute(0, 3, 1, 2)
-
- return x
diff --git a/modules/scripts.py b/modules/scripts.py
index b934d881..722f8685 100644
--- a/modules/scripts.py
+++ b/modules/scripts.py
@@ -36,7 +36,7 @@ class Script:
def ui(self, is_img2img):
"""this function should create gradio UI elements. See https://gradio.app/docs/#components
The return value should be an array of all components that are used in processing.
- Values of those returned componenbts will be passed to run() and process() functions.
+ Values of those returned components will be passed to run() and process() functions.
"""
pass
@@ -47,7 +47,7 @@ class Script:
This function should return:
- False if the script should not be shown in UI at all
- - True if the script should be shown in UI if it's scelected in the scripts drowpdown
+ - True if the script should be shown in UI if it's selected in the scripts dropdown
- script.AlwaysVisible if the script should be shown in UI at all times
"""
@@ -88,6 +88,17 @@ class Script:
pass
+ def postprocess_batch(self, p, *args, **kwargs):
+ """
+ Same as process_batch(), but called for every batch after it has been generated.
+
+ **kwargs will have same items as process_batch, and also:
+ - batch_number - index of current batch, from 0 to number of batches-1
+ - images - torch tensor with all generated images, with values ranging from 0 to 1;
+ """
+
+ pass
+
def postprocess(self, p, processed, *args):
"""
This function is called after processing ends for AlwaysVisible scripts.
@@ -347,6 +358,15 @@ class ScriptRunner:
print(f"Error running postprocess: {script.filename}", file=sys.stderr)
print(traceback.format_exc(), file=sys.stderr)
+ def postprocess_batch(self, p, images, **kwargs):
+ for script in self.alwayson_scripts:
+ try:
+ script_args = p.script_args[script.args_from:script.args_to]
+ script.postprocess_batch(p, *script_args, images=images, **kwargs)
+ except Exception:
+ print(f"Error running postprocess_batch: {script.filename}", file=sys.stderr)
+ print(traceback.format_exc(), file=sys.stderr)
+
def before_component(self, component, **kwargs):
for script in self.scripts:
try:
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_inpainting.py b/modules/sd_hijack_inpainting.py
index 938f9a58..85e7281f 100644
--- a/modules/sd_hijack_inpainting.py
+++ b/modules/sd_hijack_inpainting.py
@@ -209,7 +209,7 @@ def p_sample_plms(self, x, c, t, index, repeat_noise=False, use_original_steps=F
else:
x_in = torch.cat([x] * 2)
t_in = torch.cat([t] * 2)
-
+
if isinstance(c, dict):
assert isinstance(unconditional_conditioning, dict)
c_in = dict()
@@ -278,7 +278,7 @@ def p_sample_plms(self, x, c, t, index, repeat_noise=False, use_original_steps=F
x_prev, pred_x0 = get_x_prev_and_pred_x0(e_t_prime, index)
return x_prev, pred_x0, e_t
-
+
# =================================================================================================
# Monkey patch LatentInpaintDiffusion to load the checkpoint with a proper config.
# Adapted from:
@@ -324,12 +324,11 @@ def should_hijack_inpainting(checkpoint_info):
def do_inpainting_hijack():
# most of this stuff seems to no longer be needed because it is already included into SD2.0
- # LatentInpaintDiffusion remains because SD2.0's LatentInpaintDiffusion can't be loaded without specifying a checkpoint
# p_sample_plms is needed because PLMS can't work with dicts as conditionings
- # this file should be cleaned up later if weverything tuens out to work fine
+ # this file should be cleaned up later if everything turns out to work fine
# ldm.models.diffusion.ddpm.get_unconditional_conditioning = get_unconditional_conditioning
- ldm.models.diffusion.ddpm.LatentInpaintDiffusion = LatentInpaintDiffusion
+ # ldm.models.diffusion.ddpm.LatentInpaintDiffusion = LatentInpaintDiffusion
# ldm.models.diffusion.ddim.DDIMSampler.p_sample_ddim = p_sample_ddim
# ldm.models.diffusion.ddim.DDIMSampler.sample = sample_ddim
diff --git a/modules/sd_hijack_optimizations.py b/modules/sd_hijack_optimizations.py
index 98123fbf..02c87f40 100644
--- a/modules/sd_hijack_optimizations.py
+++ b/modules/sd_hijack_optimizations.py
@@ -127,7 +127,7 @@ def check_for_psutil():
invokeAI_mps_available = check_for_psutil()
-# -- Taken from https://github.com/invoke-ai/InvokeAI --
+# -- Taken from https://github.com/invoke-ai/InvokeAI and modified --
if invokeAI_mps_available:
import psutil
mem_total_gb = psutil.virtual_memory().total // (1 << 30)
@@ -152,14 +152,16 @@ def einsum_op_slice_1(q, k, v, slice_size):
return r
def einsum_op_mps_v1(q, k, v):
- if q.shape[1] <= 4096: # (512x512) max q.shape[1]: 4096
+ if q.shape[0] * q.shape[1] <= 2**16: # (512x512) max q.shape[1]: 4096
return einsum_op_compvis(q, k, v)
else:
slice_size = math.floor(2**30 / (q.shape[0] * q.shape[1]))
+ if slice_size % 4096 == 0:
+ slice_size -= 1
return einsum_op_slice_1(q, k, v, slice_size)
def einsum_op_mps_v2(q, k, v):
- if mem_total_gb > 8 and q.shape[1] <= 4096:
+ if mem_total_gb > 8 and q.shape[0] * q.shape[1] <= 2**16:
return einsum_op_compvis(q, k, v)
else:
return einsum_op_slice_0(q, k, v, 1)
@@ -188,7 +190,7 @@ def einsum_op(q, k, v):
return einsum_op_cuda(q, k, v)
if q.device.type == 'mps':
- if mem_total_gb >= 32:
+ if mem_total_gb >= 32 and q.shape[0] % 32 != 0 and q.shape[0] * q.shape[1] < 2**18:
return einsum_op_mps_v1(q, k, v)
return einsum_op_mps_v2(q, k, v)
diff --git a/modules/sd_hijack_unet.py b/modules/sd_hijack_unet.py
new file mode 100644
index 00000000..18daf8c1
--- /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 multiples 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..f36b299f 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():
@@ -223,10 +227,54 @@ def load_model_weights(model, checkpoint_info, vae_file="auto"):
model.sd_model_checkpoint = checkpoint_file
model.sd_checkpoint_info = checkpoint_info
+ sd_vae.delete_base_vae()
+ sd_vae.clear_loaded_vae()
vae_file = sd_vae.resolve_vae(checkpoint_file, vae_file=vae_file)
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()
@@ -245,13 +293,16 @@ def load_model(checkpoint_info=None):
if should_hijack_inpainting(checkpoint_info):
# Hardcoded config for now...
sd_config.model.target = "ldm.models.diffusion.ddpm.LatentInpaintDiffusion"
- sd_config.model.params.use_ema = False
sd_config.model.params.conditioning_key = "hybrid"
sd_config.model.params.unet_config.params.in_channels = 9
+ sd_config.model.params.finetune_keys = None
# Create a "fake" config with a different name so that we know to unload it when switching models.
checkpoint_info = checkpoint_info._replace(config=checkpoint_info.config.replace(".yaml", "-inpainting.yaml"))
+ if not hasattr(sd_config.model.params, "use_ema"):
+ sd_config.model.params.use_ema = False
+
do_inpainting_hijack()
if shared.cmd_opts.no_half:
diff --git a/modules/sd_samplers.py b/modules/sd_samplers.py
index 4c123d3b..d26e48dc 100644
--- a/modules/sd_samplers.py
+++ b/modules/sd_samplers.py
@@ -23,16 +23,16 @@ samplers_k_diffusion = [
('Euler', 'sample_euler', ['k_euler'], {}),
('LMS', 'sample_lms', ['k_lms'], {}),
('Heun', 'sample_heun', ['k_heun'], {}),
- ('DPM2', 'sample_dpm_2', ['k_dpm_2'], {}),
- ('DPM2 a', 'sample_dpm_2_ancestral', ['k_dpm_2_a'], {}),
+ ('DPM2', 'sample_dpm_2', ['k_dpm_2'], {'discard_next_to_last_sigma': True}),
+ ('DPM2 a', 'sample_dpm_2_ancestral', ['k_dpm_2_a'], {'discard_next_to_last_sigma': True}),
('DPM++ 2S a', 'sample_dpmpp_2s_ancestral', ['k_dpmpp_2s_a'], {}),
('DPM++ 2M', 'sample_dpmpp_2m', ['k_dpmpp_2m'], {}),
('DPM++ SDE', 'sample_dpmpp_sde', ['k_dpmpp_sde'], {}),
('DPM fast', 'sample_dpm_fast', ['k_dpm_fast'], {}),
('DPM adaptive', 'sample_dpm_adaptive', ['k_dpm_ad'], {}),
('LMS Karras', 'sample_lms', ['k_lms_ka'], {'scheduler': 'karras'}),
- ('DPM2 Karras', 'sample_dpm_2', ['k_dpm_2_ka'], {'scheduler': 'karras'}),
- ('DPM2 a Karras', 'sample_dpm_2_ancestral', ['k_dpm_2_a_ka'], {'scheduler': 'karras'}),
+ ('DPM2 Karras', 'sample_dpm_2', ['k_dpm_2_ka'], {'scheduler': 'karras', 'discard_next_to_last_sigma': True}),
+ ('DPM2 a Karras', 'sample_dpm_2_ancestral', ['k_dpm_2_a_ka'], {'scheduler': 'karras', 'discard_next_to_last_sigma': True}),
('DPM++ 2S a Karras', 'sample_dpmpp_2s_ancestral', ['k_dpmpp_2s_a_ka'], {'scheduler': 'karras'}),
('DPM++ 2M Karras', 'sample_dpmpp_2m', ['k_dpmpp_2m_ka'], {'scheduler': 'karras'}),
('DPM++ SDE Karras', 'sample_dpmpp_sde', ['k_dpmpp_sde_ka'], {'scheduler': 'karras'}),
@@ -444,9 +444,7 @@ class KDiffusionSampler:
return extra_params_kwargs
- def sample_img2img(self, p, x, noise, conditioning, unconditional_conditioning, steps=None, image_conditioning=None):
- steps, t_enc = setup_img2img_steps(p, steps)
-
+ def get_sigmas(self, p, steps):
if p.sampler_noise_scheduler_override:
sigmas = p.sampler_noise_scheduler_override(steps)
elif self.config is not None and self.config.options.get('scheduler', None) == 'karras':
@@ -454,6 +452,16 @@ class KDiffusionSampler:
else:
sigmas = self.model_wrap.get_sigmas(steps)
+ if self.config is not None and self.config.options.get('discard_next_to_last_sigma', False):
+ sigmas = torch.cat([sigmas[:-2], sigmas[-1:]])
+
+ return sigmas
+
+ def sample_img2img(self, p, x, noise, conditioning, unconditional_conditioning, steps=None, image_conditioning=None):
+ steps, t_enc = setup_img2img_steps(p, steps)
+
+ sigmas = self.get_sigmas(p, steps)
+
sigma_sched = sigmas[steps - t_enc - 1:]
xi = x + noise * sigma_sched[0]
@@ -485,12 +493,7 @@ class KDiffusionSampler:
def sample(self, p, x, conditioning, unconditional_conditioning, steps=None, image_conditioning = None):
steps = steps or p.steps
- if p.sampler_noise_scheduler_override:
- sigmas = p.sampler_noise_scheduler_override(steps)
- elif self.config is not None and self.config.options.get('scheduler', None) == 'karras':
- sigmas = k_diffusion.sampling.get_sigmas_karras(n=steps, sigma_min=0.1, sigma_max=10, device=shared.device)
- else:
- sigmas = self.model_wrap.get_sigmas(steps)
+ sigmas = self.get_sigmas(p, steps)
x = x * sigmas[0]
diff --git a/modules/sd_vae.py b/modules/sd_vae.py
index 9c120975..25638a83 100644
--- a/modules/sd_vae.py
+++ b/modules/sd_vae.py
@@ -4,6 +4,7 @@ from collections import namedtuple
from modules import shared, devices, script_callbacks
from modules.paths import models_path
import glob
+from copy import deepcopy
model_dir = "Stable-diffusion"
@@ -15,7 +16,7 @@ vae_path = os.path.abspath(os.path.join(models_path, vae_dir))
vae_ignore_keys = {"model_ema.decay", "model_ema.num_updates"}
-default_vae_dict = {"auto": "auto", "None": "None"}
+default_vae_dict = {"auto": "auto", "None": None, None: None}
default_vae_list = ["auto", "None"]
@@ -39,7 +40,8 @@ def get_base_vae(model):
def store_base_vae(model):
global base_vae, checkpoint_info
if checkpoint_info != model.sd_checkpoint_info:
- base_vae = model.first_stage_model.state_dict().copy()
+ assert not loaded_vae_file, "Trying to store non-base VAE!"
+ base_vae = deepcopy(model.first_stage_model.state_dict())
checkpoint_info = model.sd_checkpoint_info
@@ -50,9 +52,11 @@ def delete_base_vae():
def restore_base_vae(model):
- global base_vae, checkpoint_info
+ global loaded_vae_file
if base_vae is not None and checkpoint_info == model.sd_checkpoint_info:
- load_vae_dict(model, base_vae)
+ print("Restoring base VAE")
+ _load_vae_dict(model, base_vae)
+ loaded_vae_file = None
delete_base_vae()
@@ -148,9 +152,10 @@ def load_vae(model, vae_file=None):
if vae_file:
assert os.path.isfile(vae_file), f"VAE file doesn't exist: {vae_file}"
print(f"Loading VAE weights from: {vae_file}")
+ store_base_vae(model)
vae_ckpt = torch.load(vae_file, map_location=shared.weight_load_location)
vae_dict_1 = {k: v for k, v in vae_ckpt["state_dict"].items() if k[0:4] != "loss" and k not in vae_ignore_keys}
- load_vae_dict(model, vae_dict_1)
+ _load_vae_dict(model, vae_dict_1)
# If vae used is not in dict, update it
# It will be removed on refresh though
@@ -158,30 +163,22 @@ def load_vae(model, vae_file=None):
if vae_opt not in vae_dict:
vae_dict[vae_opt] = vae_file
vae_list.append(vae_opt)
+ elif loaded_vae_file:
+ restore_base_vae(model)
loaded_vae_file = vae_file
- """
- # Save current VAE to VAE settings, maybe? will it work?
- if save_settings:
- if vae_file is None:
- vae_opt = "None"
-
- # shared.opts.sd_vae = vae_opt
- """
-
first_load = False
# don't call this from outside
-def load_vae_dict(model, vae_dict_1=None):
- if vae_dict_1:
- store_base_vae(model)
- model.first_stage_model.load_state_dict(vae_dict_1)
- else:
- restore_base_vae()
+def _load_vae_dict(model, vae_dict_1):
+ model.first_stage_model.load_state_dict(vae_dict_1)
model.first_stage_model.to(devices.dtype_vae)
+def clear_loaded_vae():
+ global loaded_vae_file
+ loaded_vae_file = None
def reload_vae_weights(sd_model=None, vae_file="auto"):
from modules import lowvram, devices, sd_hijack
diff --git a/modules/shared.py b/modules/shared.py
index dc45fcaa..8ea3b441 100644
--- a/modules/shared.py
+++ b/modules/shared.py
@@ -5,6 +5,7 @@ import os
import sys
import time
+from PIL import Image
import gradio as gr
import tqdm
@@ -293,6 +294,7 @@ options_templates.update(options_section(('saving-images', "Saving images/grids"
"export_for_4chan": OptionInfo(True, "If PNG image is larger than 4MB or any dimension is larger than 4000, downscale and save copy as JPG"),
"use_original_name_batch": OptionInfo(False, "Use original name for output filename during batch process in extras tab"),
+ "use_upscaler_name_as_suffix": OptionInfo(False, "Use upscaler name as filename suffix in the extras tab"),
"save_selected_only": OptionInfo(True, "When using 'Save' button, only save a single selected image"),
"do_not_add_watermark": OptionInfo(False, "Do not add watermark to images"),
@@ -359,14 +361,15 @@ options_templates.update(options_section(('sd', "Stable Diffusion"), {
"sd_hypernetwork": OptionInfo("None", "Hypernetwork", gr.Dropdown, lambda: {"choices": ["None"] + [x for x in hypernetworks.keys()]}, refresh=reload_hypernetworks),
"sd_hypernetwork_strength": OptionInfo(1.0, "Hypernetwork strength", gr.Slider, {"minimum": 0.0, "maximum": 1.0, "step": 0.001}),
"inpainting_mask_weight": OptionInfo(1.0, "Inpainting conditioning mask strength", gr.Slider, {"minimum": 0.0, "maximum": 1.0, "step": 0.01}),
+ "initial_noise_multiplier": OptionInfo(1.0, "Noise multiplier for img2img", gr.Slider, {"minimum": 0.5, "maximum": 1.5, "step": 0.01 }),
"img2img_color_correction": OptionInfo(False, "Apply color correction to img2img results to match original colors."),
"img2img_fix_steps": OptionInfo(False, "With img2img, do exactly the amount of steps the slider specifies (normally you'd do less with less denoising)."),
+ "img2img_background_color": OptionInfo("#ffffff", "With img2img, fill image's transparent parts with this color.", gr.ColorPicker, {}),
"enable_quantization": OptionInfo(False, "Enable quantization in K samplers for sharper and cleaner results. This may change existing seeds. Requires restart to apply."),
"enable_emphasis": OptionInfo(True, "Emphasis: use (text) to make model pay more attention to text and [text] to make it pay less attention"),
"use_old_emphasis_implementation": OptionInfo(False, "Use old emphasis implementation. Can be useful to reproduce old seeds."),
"enable_batch_seeds": OptionInfo(True, "Make K-diffusion samplers produce same images in a batch as when making a single image"),
"comma_padding_backtrack": OptionInfo(20, "Increase coherency by padding from the last comma within n tokens when using more than 75 tokens", gr.Slider, {"minimum": 0, "maximum": 74, "step": 1 }),
- "filter_nsfw": OptionInfo(False, "Filter NSFW content"),
'CLIP_stop_at_last_layers': OptionInfo(1, "Clip skip", gr.Slider, {"minimum": 1, "maximum": 12, "step": 1}),
"random_artist_categories": OptionInfo([], "Allowed categories for random artists selection when using the Roll button", gr.CheckboxGroup, {"choices": artist_db.categories()}),
}))
@@ -395,6 +398,7 @@ options_templates.update(options_section(('ui', "User interface"), {
"add_model_name_to_info": OptionInfo(False, "Add model name to generation information"),
"disable_weights_auto_swap": OptionInfo(False, "When reading generation parameters from text into UI (from PNG info or pasted text), do not change the selected model/checkpoint."),
"send_seed": OptionInfo(True, "Send seed when sending prompt or image to other interface"),
+ "send_size": OptionInfo(True, "Send size when sending prompt or image to another interface"),
"font": OptionInfo("", "Font for image grids that have text"),
"js_modal_lightbox": OptionInfo(True, "Enable full page image viewer"),
"js_modal_lightbox_initially_zoomed": OptionInfo(True, "Show images zoomed in by default in full page image viewer"),
diff --git a/modules/textual_inversion/dataset.py b/modules/textual_inversion/dataset.py
index 2dc64c3c..88d68c76 100644
--- a/modules/textual_inversion/dataset.py
+++ b/modules/textual_inversion/dataset.py
@@ -28,9 +28,9 @@ class DatasetEntry:
class PersonalizedBase(Dataset):
- def __init__(self, data_root, width, height, repeats, flip_p=0.5, placeholder_token="*", model=None, cond_model=None, device=None, template_file=None, include_cond=False, batch_size=1, gradient_step=1, shuffle_tags=False, tag_drop_out=0, latent_sampling_method='once'):
+ def __init__(self, data_root, width, height, repeats, flip_p=0.5, placeholder_token="*", model=None, cond_model=None, device=None, template_file=None, include_cond=False, batch_size=1, gradient_step=1, shuffle_tags=False, tag_drop_out=0, latent_sampling_method='once'):
re_word = re.compile(shared.opts.dataset_filename_word_regex) if len(shared.opts.dataset_filename_word_regex) > 0 else None
-
+
self.placeholder_token = placeholder_token
self.width = width
@@ -50,14 +50,14 @@ class PersonalizedBase(Dataset):
self.image_paths = [os.path.join(data_root, file_path) for file_path in os.listdir(data_root)]
-
+
self.shuffle_tags = shuffle_tags
self.tag_drop_out = tag_drop_out
print("Preparing dataset...")
for path in tqdm.tqdm(self.image_paths):
if shared.state.interrupted:
- raise Exception("inturrupted")
+ raise Exception("interrupted")
try:
image = Image.open(path).convert('RGB').resize((self.width, self.height), PIL.Image.BICUBIC)
except Exception:
@@ -144,7 +144,7 @@ class PersonalizedDataLoader(DataLoader):
self.collate_fn = collate_wrapper_random
else:
self.collate_fn = collate_wrapper
-
+
class BatchLoader:
def __init__(self, data):
diff --git a/modules/textual_inversion/textual_inversion.py b/modules/textual_inversion/textual_inversion.py
index e28c357a..daf3997b 100644
--- a/modules/textual_inversion/textual_inversion.py
+++ b/modules/textual_inversion/textual_inversion.py
@@ -133,7 +133,7 @@ class EmbeddingDatabase:
process_file(fullfn, fn)
except Exception:
- print(f"Error loading emedding {fn}:", file=sys.stderr)
+ print(f"Error loading embedding {fn}:", file=sys.stderr)
print(traceback.format_exc(), file=sys.stderr)
continue
@@ -194,7 +194,7 @@ def write_loss(log_directory, filename, step, epoch_len, values):
csv_writer.writeheader()
epoch = (step - 1) // epoch_len
- epoch_step = (step - 1) % epoch_len
+ epoch_step = (step - 1) % epoch_len
csv_writer.writerow({
"step": step,
@@ -270,9 +270,9 @@ def train_embedding(embedding_name, learn_rate, batch_size, gradient_step, data_
# dataset loading may take a while, so input validations and early returns should be done before this
shared.state.textinfo = f"Preparing dataset from {html.escape(data_root)}..."
old_parallel_processing_allowed = shared.parallel_processing_allowed
-
+
pin_memory = shared.opts.pin_memory
-
+
ds = modules.textual_inversion.dataset.PersonalizedBase(data_root=data_root, width=training_width, height=training_height, repeats=shared.opts.training_image_repeats_per_epoch, placeholder_token=embedding_name, model=shared.sd_model, cond_model=shared.sd_model.cond_stage_model, device=devices.device, template_file=template_file, batch_size=batch_size, gradient_step=gradient_step, shuffle_tags=shuffle_tags, tag_drop_out=tag_drop_out, latent_sampling_method=latent_sampling_method)
latent_sampling_method = ds.latent_sampling_method
@@ -295,12 +295,12 @@ def train_embedding(embedding_name, learn_rate, batch_size, gradient_step, data_
loss_step = 0
_loss_step = 0 #internal
-
+
last_saved_file = "<none>"
last_saved_image = "<none>"
forced_filename = "<none>"
embedding_yet_to_be_embedded = False
-
+
pbar = tqdm.tqdm(total=steps - initial_step)
try:
for i in range((steps-initial_step) * gradient_step):
@@ -327,10 +327,10 @@ def train_embedding(embedding_name, learn_rate, batch_size, gradient_step, data_
c = shared.sd_model.cond_stage_model(batch.cond_text)
loss = shared.sd_model(x, c)[0] / gradient_step
del x
-
+
_loss_step += loss.item()
scaler.scale(loss).backward()
-
+
# go back until we reach gradient accumulation steps
if (j + 1) % gradient_step != 0:
continue
diff --git a/modules/ui.py b/modules/ui.py
index fe4abe05..faba69a4 100644
--- a/modules/ui.py
+++ b/modules/ui.py
@@ -49,10 +49,14 @@ if not cmd_opts.share and not cmd_opts.listen:
gradio.utils.version_check = lambda: None
gradio.utils.get_local_ip_address = lambda: '127.0.0.1'
-if cmd_opts.ngrok != None:
+if cmd_opts.ngrok is not None:
import modules.ngrok as ngrok
print('ngrok authtoken detected, trying to connect...')
- ngrok.connect(cmd_opts.ngrok, cmd_opts.port if cmd_opts.port != None else 7860, cmd_opts.ngrok_region)
+ ngrok.connect(
+ cmd_opts.ngrok,
+ cmd_opts.port if cmd_opts.port is not None else 7860,
+ cmd_opts.ngrok_region
+ )
def gr_show(visible=True):
@@ -82,6 +86,7 @@ folder_symbol = '\U0001f4c2' # 📂
refresh_symbol = '\U0001f504' # 🔄
save_style_symbol = '\U0001f4be' # 💾
apply_style_symbol = '\U0001f4cb' # 📋
+clear_prompt_symbol = '\U0001F5D1' # 🗑️
def plaintext_to_html(text):
@@ -302,8 +307,8 @@ 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])
@@ -316,6 +321,17 @@ def create_seed_inputs():
return seed, reuse_seed, subseed, reuse_subseed, subseed_strength, seed_resize_from_h, seed_resize_from_w, seed_checkbox
+
+def connect_clear_prompt(button):
+ """Given clear button, prompt, and token_counter objects, setup clear prompt button click event"""
+ button.click(
+ _js="clear_prompt",
+ fn=None,
+ inputs=[],
+ outputs=[],
+ )
+
+
def connect_reuse_seed(seed: gr.Number, reuse_seed: gr.Button, generation_info: gr.Textbox, dummy_component, is_subseed):
""" Connects a 'reuse (sub)seed' button's click event so that it copies last used
(sub)seed value from generation info the to the seed field. If copying subseed and subseed strength
@@ -391,10 +407,17 @@ def create_toprow(is_img2img):
paste = gr.Button(value=paste_symbol, elem_id="paste")
save_style = gr.Button(value=save_style_symbol, elem_id="style_create")
prompt_style_apply = gr.Button(value=apply_style_symbol, elem_id="style_apply")
-
+ clear_prompt_button = gr.Button(value=clear_prompt_symbol, elem_id=f"{id_part}_clear_prompt")
token_counter = gr.HTML(value="<span></span>", elem_id=f"{id_part}_token_counter")
token_button = gr.Button(visible=False, elem_id=f"{id_part}_token_button")
+ clear_prompt_button.click(
+ fn=lambda *x: x,
+ _js="confirm_clear_prompt",
+ inputs=[prompt, negative_prompt],
+ outputs=[prompt, negative_prompt],
+ )
+
button_interrogate = None
button_deepbooru = None
if is_img2img:
@@ -616,10 +639,14 @@ def create_ui():
modules.scripts.scripts_txt2img.initialize_scripts(is_img2img=False)
with gr.Blocks(analytics_enabled=False) as txt2img_interface:
- txt2img_prompt, roll, txt2img_prompt_style, txt2img_negative_prompt, txt2img_prompt_style2, submit, _, _, txt2img_prompt_style_apply, txt2img_save_style, txt2img_paste, token_counter, token_button = create_toprow(is_img2img=False)
+ txt2img_prompt, roll, txt2img_prompt_style, txt2img_negative_prompt, txt2img_prompt_style2, submit, _, _,txt2img_prompt_style_apply, txt2img_save_style, txt2img_paste, token_counter, token_button = create_toprow(is_img2img=False)
+
dummy_component = gr.Label(visible=False)
txt_prompt_img = gr.File(label="", elem_id="txt2img_prompt_image", file_count="single", type="bytes", visible=False)
+
+
+
with gr.Row(elem_id='txt2img_progress_row'):
with gr.Column(scale=1):
pass
@@ -635,8 +662,8 @@ 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)
@@ -644,8 +671,8 @@ 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):
@@ -770,7 +797,8 @@ def create_ui():
modules.scripts.scripts_img2img.initialize_scripts(is_img2img=True)
with gr.Blocks(analytics_enabled=False) as img2img_interface:
- img2img_prompt, roll, img2img_prompt_style, img2img_negative_prompt, img2img_prompt_style2, submit, img2img_interrogate, img2img_deepbooru, img2img_prompt_style_apply, img2img_save_style, img2img_paste, token_counter, token_button = create_toprow(is_img2img=True)
+ img2img_prompt, roll, img2img_prompt_style, img2img_negative_prompt, img2img_prompt_style2, submit, img2img_interrogate, img2img_deepbooru, img2img_prompt_style_apply, img2img_save_style, img2img_paste,token_counter, token_button = create_toprow(is_img2img=True)
+
with gr.Row(elem_id='img2img_progress_row'):
img2img_prompt_img = gr.File(label="", elem_id="img2img_prompt_image", file_count="single", type="bytes", visible=False)
@@ -788,7 +816,7 @@ def create_ui():
with gr.Tabs(elem_id="mode_img2img") as tabs_img2img_mode:
with gr.TabItem('img2img', id='img2img'):
- 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)
+ 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, image_mode="RGBA").style(height=480)
with gr.TabItem('Inpaint', id='inpaint'):
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)
@@ -835,8 +863,8 @@ 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)
@@ -1171,8 +1199,8 @@ 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():
@@ -1230,8 +1258,8 @@ 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)
@@ -1450,7 +1478,7 @@ 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 b487ac25..eec9586f 100644
--- a/modules/ui_extensions.py
+++ b/modules/ui_extensions.py
@@ -9,6 +9,8 @@ import git
import gradio as gr
import html
+import shutil
+import errno
from modules import extensions, shared, paths
@@ -138,7 +140,18 @@ def install_extension_from_url(dirname, url):
repo = git.Repo.clone_from(url, tmpdir)
repo.remote().fetch()
- os.rename(tmpdir, target_dir)
+ try:
+ os.rename(tmpdir, target_dir)
+ except OSError as err:
+ # TODO what does this do on windows? I think it'll be a different error code but I don't have a system to check it
+ # Shouldn't cause any new issues at least but we probably want to handle it there too.
+ if err.errno == errno.EXDEV:
+ # Cross device link, typical in docker or when tmp/ and extensions/ are on different file systems
+ # Since we can't use a rename, do the slower but more versitile shutil.move()
+ shutil.move(tmpdir, target_dir)
+ else:
+ # Something else, not enough free space, permissions, etc. rethrow it so that it gets handled.
+ raise(err)
import launch
launch.run_extension_installer(target_dir)
@@ -206,12 +219,13 @@ def refresh_available_extensions_from_data(hide_tags):
if url is None:
continue
+ existing = installed_extension_urls.get(normalize_git_url(url), None)
+ extension_tags = extension_tags + ["installed"] if existing else extension_tags
+
if len([x for x in extension_tags if x in tags_to_hide]) > 0:
hidden += 1
continue
- existing = installed_extension_urls.get(normalize_git_url(url), None)
-
install_code = f"""<input onclick="install_extension_from_index(this, '{html.escape(url)}')" type="button" value="{"Install" if not existing else "Installed"}" {"disabled=disabled" if existing else ""} class="gr-button gr-button-lg gr-button-secondary">"""
tags_text = ", ".join([f"<span class='extension-tag' title='{tags.get(x, '')}'>{x}</span>" for x in extension_tags])
@@ -222,7 +236,11 @@ def refresh_available_extensions_from_data(hide_tags):
<td>{html.escape(description)}</td>
<td>{install_code}</td>
</tr>
- """
+
+ """
+
+ for tag in [x for x in extension_tags if x not in tags]:
+ tags[tag] = tag
code += """
</tbody>
@@ -272,7 +290,7 @@ def create_ui():
install_extension_button = gr.Button(elem_id="install_extension_button", visible=False)
with gr.Row():
- hide_tags = gr.CheckboxGroup(value=["ads", "localization"], label="Hide extensions with tags", choices=["script", "ads", "localization"])
+ hide_tags = gr.CheckboxGroup(value=["ads", "localization", "installed"], label="Hide extensions with tags", choices=["script", "ads", "localization", "installed"])
install_result = gr.HTML()
available_extensions_table = gr.HTML()
diff --git a/requirements.txt b/requirements.txt
index 5f3d9623..5bed694e 100644
--- a/requirements.txt
+++ b/requirements.txt
@@ -1,6 +1,6 @@
+blendmodes
accelerate
basicsr
-diffusers
fairscale==0.4.4
fonts
font-roboto
@@ -30,3 +30,4 @@ inflection
GitPython
torchsde
safetensors
+psutil; sys_platform == 'darwin'
diff --git a/requirements_versions.txt b/requirements_versions.txt
index 035fa82f..c126c8c4 100644
--- a/requirements_versions.txt
+++ b/requirements_versions.txt
@@ -1,5 +1,5 @@
+blendmodes==2022
transformers==4.19.2
-diffusers==0.3.0
accelerate==0.12.0
basicsr==1.4.2
gfpgan==1.3.8
@@ -27,3 +27,4 @@ inflection==0.5.1
GitPython==3.1.27
torchsde==0.2.5
safetensors==0.2.5
+httpcore<=0.15
diff --git a/script.js b/script.js
index 8b3b67e3..9748ec90 100644
--- a/script.js
+++ b/script.js
@@ -1,5 +1,6 @@
-function gradioApp(){
- return document.getElementsByTagName('gradio-app')[0].shadowRoot;
+function gradioApp() {
+ const gradioShadowRoot = document.getElementsByTagName('gradio-app')[0].shadowRoot
+ return !!gradioShadowRoot ? gradioShadowRoot : document;
}
function get_uiCurrentTab() {
@@ -82,4 +83,4 @@ function uiElementIsVisible(el) {
}
}
return isVisible;
-} \ No newline at end of file
+}
diff --git a/scripts/prompt_matrix.py b/scripts/prompt_matrix.py
index 5fd952e9..4c79eaef 100644
--- a/scripts/prompt_matrix.py
+++ b/scripts/prompt_matrix.py
@@ -18,7 +18,7 @@ def draw_xy_grid(xs, ys, x_label, y_label, cell):
ver_texts = [[images.GridAnnotation(y_label(y))] for y in ys]
hor_texts = [[images.GridAnnotation(x_label(x))] for x in xs]
- first_pocessed = None
+ first_processed = None
state.job_count = len(xs) * len(ys)
@@ -27,17 +27,17 @@ def draw_xy_grid(xs, ys, x_label, y_label, cell):
state.job = f"{ix + iy * len(xs) + 1} out of {len(xs) * len(ys)}"
processed = cell(x, y)
- if first_pocessed is None:
- first_pocessed = processed
+ if first_processed is None:
+ first_processed = processed
res.append(processed.images[0])
grid = images.image_grid(res, rows=len(ys))
grid = images.draw_grid_annotations(grid, res[0].width, res[0].height, hor_texts, ver_texts)
- first_pocessed.images = [grid]
+ first_processed.images = [grid]
- return first_pocessed
+ return first_processed
class Script(scripts.Script):
@@ -46,10 +46,11 @@ class Script(scripts.Script):
def ui(self, is_img2img):
put_at_start = gr.Checkbox(label='Put variable parts at start of prompt', value=False)
+ different_seeds = gr.Checkbox(label='Use different seed for each picture', value=False)
- return [put_at_start]
+ return [put_at_start, different_seeds]
- def run(self, p, put_at_start):
+ def run(self, p, put_at_start, different_seeds):
modules.processing.fix_seed(p)
original_prompt = p.prompt[0] if type(p.prompt) == list else p.prompt
@@ -73,7 +74,7 @@ class Script(scripts.Script):
print(f"Prompt matrix will create {len(all_prompts)} images using a total of {p.n_iter} batches.")
p.prompt = all_prompts
- p.seed = [p.seed for _ in all_prompts]
+ p.seed = [p.seed + (i if different_seeds else 0) for i in range(len(all_prompts))]
p.prompt_for_display = original_prompt
processed = process_images(p)
diff --git a/scripts/sd_upscale.py b/scripts/sd_upscale.py
index 01074291..28bd96b3 100644
--- a/scripts/sd_upscale.py
+++ b/scripts/sd_upscale.py
@@ -17,13 +17,14 @@ class Script(scripts.Script):
return is_img2img
def ui(self, is_img2img):
- info = gr.HTML("<p style=\"margin-bottom:0.75em\">Will upscale the image to twice the dimensions; use width and height sliders to set tile size</p>")
+ info = gr.HTML("<p style=\"margin-bottom:0.75em\">Will upscale the image by the selected scale factor; use width and height sliders to set tile size</p>")
overlap = gr.Slider(minimum=0, maximum=256, step=16, label='Tile overlap', value=64)
+ scale_factor = gr.Slider(minimum=1, maximum=4, step=1, label='Scale Factor', value=2)
upscaler_index = gr.Radio(label='Upscaler', choices=[x.name for x in shared.sd_upscalers], value=shared.sd_upscalers[0].name, type="index")
- return [info, overlap, upscaler_index]
+ return [info, overlap, upscaler_index, scale_factor]
- def run(self, p, _, overlap, upscaler_index):
+ def run(self, p, _, overlap, upscaler_index, scale_factor):
processing.fix_seed(p)
upscaler = shared.sd_upscalers[upscaler_index]
@@ -34,9 +35,9 @@ class Script(scripts.Script):
seed = p.seed
init_img = p.init_images[0]
-
- if(upscaler.name != "None"):
- img = upscaler.scaler.upscale(init_img, 2, upscaler.data_path)
+
+ if (upscaler.name != "None"):
+ img = upscaler.scaler.upscale(init_img, scale_factor, upscaler.data_path)
else:
img = init_img
@@ -69,7 +70,7 @@ class Script(scripts.Script):
work_results = []
for i in range(batch_count):
p.batch_size = batch_size
- p.init_images = work[i*batch_size:(i+1)*batch_size]
+ p.init_images = work[i * batch_size:(i + 1) * batch_size]
state.job = f"Batch {i + 1 + n * batch_count} out of {state.job_count}"
processed = processing.process_images(p)
diff --git a/scripts/xy_grid.py b/scripts/xy_grid.py
index d402c281..3e0b2805 100644
--- a/scripts/xy_grid.py
+++ b/scripts/xy_grid.py
@@ -10,13 +10,16 @@ import numpy as np
import modules.scripts as scripts
import gradio as gr
-from modules import images, sd_samplers
+from modules import images, paths, sd_samplers
from modules.hypernetworks import hypernetwork
from modules.processing import process_images, Processed, StableDiffusionProcessingTxt2Img
from modules.shared import opts, cmd_opts, state
import modules.shared as shared
import modules.sd_samplers
import modules.sd_models
+import modules.sd_vae
+import glob
+import os
import re
@@ -114,6 +117,38 @@ def apply_clip_skip(p, x, xs):
opts.data["CLIP_stop_at_last_layers"] = x
+def apply_upscale_latent_space(p, x, xs):
+ if x.lower().strip() != '0':
+ opts.data["use_scale_latent_for_hires_fix"] = True
+ else:
+ opts.data["use_scale_latent_for_hires_fix"] = False
+
+
+def find_vae(name: str):
+ if name.lower() in ['auto', 'none']:
+ return name
+ else:
+ vae_path = os.path.abspath(os.path.join(paths.models_path, 'VAE'))
+ found = glob.glob(os.path.join(vae_path, f'**/{name}.*pt'), recursive=True)
+ if found:
+ return found[0]
+ else:
+ return 'auto'
+
+
+def apply_vae(p, x, xs):
+ if x.lower().strip() == 'none':
+ modules.sd_vae.reload_vae_weights(shared.sd_model, vae_file='None')
+ else:
+ found = find_vae(x)
+ if found:
+ v = modules.sd_vae.reload_vae_weights(shared.sd_model, vae_file=found)
+
+
+def apply_styles(p: StableDiffusionProcessingTxt2Img, x: str, _):
+ p.styles = x.split(',')
+
+
def format_value_add_label(p, opt, x):
if type(x) == float:
x = round(x, 8)
@@ -167,7 +202,10 @@ axis_options = [
AxisOption("Eta", float, apply_field("eta"), format_value_add_label, None),
AxisOption("Clip skip", int, apply_clip_skip, format_value_add_label, None),
AxisOption("Denoising", float, apply_field("denoising_strength"), format_value_add_label, None),
+ AxisOption("Upscale latent space for hires.", str, apply_upscale_latent_space, format_value_add_label, None),
AxisOption("Cond. Image Mask Weight", float, apply_field("inpainting_mask_weight"), format_value_add_label, None),
+ AxisOption("VAE", str, apply_vae, format_value_add_label, None),
+ AxisOption("Styles", str, apply_styles, format_value_add_label, None),
]
@@ -229,14 +267,18 @@ class SharedSettingsStackHelper(object):
self.CLIP_stop_at_last_layers = opts.CLIP_stop_at_last_layers
self.hypernetwork = opts.sd_hypernetwork
self.model = shared.sd_model
+ self.use_scale_latent_for_hires_fix = opts.use_scale_latent_for_hires_fix
+ self.vae = opts.sd_vae
def __exit__(self, exc_type, exc_value, tb):
modules.sd_models.reload_model_weights(self.model)
+ modules.sd_vae.reload_vae_weights(self.model, vae_file=find_vae(self.vae))
hypernetwork.load_hypernetwork(self.hypernetwork)
hypernetwork.apply_strength()
opts.data["CLIP_stop_at_last_layers"] = self.CLIP_stop_at_last_layers
+ opts.data["use_scale_latent_for_hires_fix"] = self.use_scale_latent_for_hires_fix
re_range = re.compile(r"\s*([+-]?\s*\d+)\s*-\s*([+-]?\s*\d+)(?:\s*\(([+-]\d+)\s*\))?\s*")
diff --git a/style.css b/style.css
index e2b71f25..3ad78006 100644
--- a/style.css
+++ b/style.css
@@ -114,7 +114,7 @@
padding: 0.4em 0;
}
-#roll, #paste, #style_create, #style_apply{
+#roll_col > button {
min-width: 2em;
min-height: 2em;
max-width: 2em;
diff --git a/webui-macos-env.sh b/webui-macos-env.sh
index 68d1f754..95ca9c55 100644
--- a/webui-macos-env.sh
+++ b/webui-macos-env.sh
@@ -4,6 +4,12 @@
# 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"
diff --git a/webui.py b/webui.py
index 78204d11..3aee8792 100644
--- a/webui.py
+++ b/webui.py
@@ -8,6 +8,7 @@ from fastapi import FastAPI
from fastapi.middleware.cors import CORSMiddleware
from fastapi.middleware.gzip import GZipMiddleware
+from modules import import_hook
from modules.call_queue import wrap_queued_call, queue_lock, wrap_gradio_gpu_call
from modules.paths import script_path
@@ -137,7 +138,7 @@ def webui():
shared.demo = modules.ui.create_ui()
- app, local_url, share_url = shared.demo.launch(
+ app, local_url, share_url = shared.demo.queue(default_enabled=False).launch(
share=cmd_opts.share,
server_name=server_name,
server_port=cmd_opts.port,
@@ -153,8 +154,8 @@ def webui():
# gradio uses a very open CORS policy via app.user_middleware, which makes it possible for
# an attacker to trick the user into opening a malicious HTML page, which makes a request to the
- # running web ui and do whatever the attcker wants, including installing an extension and
- # runnnig its code. We disable this here. Suggested by RyotaK.
+ # running web ui and do whatever the attacker wants, including installing an extension and
+ # running its code. We disable this here. Suggested by RyotaK.
app.user_middleware = [x for x in app.user_middleware if x.cls.__name__ != 'CORSMiddleware']
setup_cors(app)