diff options
| -rw-r--r-- | README.md | 4 | ||||
| -rw-r--r-- | javascript/edit-attention.js | 41 | ||||
| -rw-r--r-- | javascript/progressbar.js | 15 | ||||
| -rw-r--r-- | launch.py | 11 | ||||
| -rw-r--r-- | modules/extras.py | 2 | ||||
| -rw-r--r-- | modules/hypernetwork.py | 88 | ||||
| -rw-r--r-- | modules/images.py | 18 | ||||
| -rw-r--r-- | modules/processing.py | 50 | ||||
| -rw-r--r-- | modules/prompt_parser.py | 249 | ||||
| -rw-r--r-- | modules/sd_hijack.py | 10 | ||||
| -rw-r--r-- | modules/sd_hijack_optimizations.py | 25 | ||||
| -rw-r--r-- | modules/sd_models.py | 8 | ||||
| -rw-r--r-- | modules/sd_samplers.py | 106 | ||||
| -rw-r--r-- | modules/shared.py | 43 | ||||
| -rw-r--r-- | modules/textual_inversion/preprocess.py | 29 | ||||
| -rw-r--r-- | modules/ui.py | 62 | ||||
| -rw-r--r-- | scripts/img2imgalt.py | 3 | ||||
| -rw-r--r-- | scripts/outpainting_mk_2.py | 5 | ||||
| -rw-r--r-- | scripts/xy_grid.py | 18 | ||||
| -rw-r--r-- | style.css | 8 | ||||
| -rw-r--r-- | webui.py | 7 |
21 files changed, 615 insertions, 187 deletions
@@ -16,6 +16,7 @@ Check the [custom scripts](https://github.com/AUTOMATIC1111/stable-diffusion-web - Attention, specify parts of text that the model should pay more attention to
- a man in a ((tuxedo)) - will pay more attention to tuxedo
- a man in a (tuxedo:1.21) - alternative syntax
+ - select text and press ctrl+up or ctrl+down to aduotmatically adjust attention to selected text
- Loopback, run img2img processing multiple times
- X/Y plot, a way to draw a 2 dimensional plot of images with different parameters
- Textual Inversion
@@ -61,6 +62,9 @@ Check the [custom scripts](https://github.com/AUTOMATIC1111/stable-diffusion-web - Reloading checkpoints on the fly
- Checkpoint Merger, a tab that allows you to merge two checkpoints into one
- [Custom scripts](https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Custom-Scripts) with many extensions from community
+- [Composable-Diffusion](https://energy-based-model.github.io/Compositional-Visual-Generation-with-Composable-Diffusion-Models/), a way to use multiple prompts at once
+ - separate prompts using uppercase `AND`
+ - also supports weights for prompts: `a cat :1.2 AND a dog AND a penguin :2.2`
## Installation and Running
Make sure the required [dependencies](https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Dependencies) are met and follow the instructions available for both [NVidia](https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Install-and-Run-on-NVidia-GPUs) (recommended) and [AMD](https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Install-and-Run-on-AMD-GPUs) GPUs.
diff --git a/javascript/edit-attention.js b/javascript/edit-attention.js new file mode 100644 index 00000000..c67ed579 --- /dev/null +++ b/javascript/edit-attention.js @@ -0,0 +1,41 @@ +addEventListener('keydown', (event) => {
+ let target = event.originalTarget;
+ if (!target.hasAttribute("placeholder")) return;
+ if (!target.placeholder.toLowerCase().includes("prompt")) return;
+
+ let plus = "ArrowUp"
+ let minus = "ArrowDown"
+ if (event.key != plus && event.key != minus) return;
+
+ selectionStart = target.selectionStart;
+ selectionEnd = target.selectionEnd;
+ if(selectionStart == selectionEnd) return;
+
+ event.preventDefault();
+
+ if (selectionStart == 0 || target.value[selectionStart - 1] != "(") {
+ target.value = target.value.slice(0, selectionStart) +
+ "(" + target.value.slice(selectionStart, selectionEnd) + ":1.0)" +
+ target.value.slice(selectionEnd);
+
+ target.focus();
+ target.selectionStart = selectionStart + 1;
+ target.selectionEnd = selectionEnd + 1;
+
+ } else {
+ end = target.value.slice(selectionEnd + 1).indexOf(")") + 1;
+ weight = parseFloat(target.value.slice(selectionEnd + 1, selectionEnd + 1 + end));
+ if (event.key == minus) weight -= 0.1;
+ if (event.key == plus) weight += 0.1;
+
+ weight = parseFloat(weight.toPrecision(12));
+
+ target.value = target.value.slice(0, selectionEnd + 1) +
+ weight +
+ target.value.slice(selectionEnd + 1 + end - 1);
+
+ target.focus();
+ target.selectionStart = selectionStart;
+ target.selectionEnd = selectionEnd;
+ }
+});
diff --git a/javascript/progressbar.js b/javascript/progressbar.js index 1e297abb..f9e9290e 100644 --- a/javascript/progressbar.js +++ b/javascript/progressbar.js @@ -4,6 +4,21 @@ global_progressbars = {} function check_progressbar(id_part, id_progressbar, id_progressbar_span, id_interrupt, id_preview, id_gallery){ var progressbar = gradioApp().getElementById(id_progressbar) var interrupt = gradioApp().getElementById(id_interrupt) + + if(opts.show_progress_in_title && progressbar && progressbar.offsetParent){ + if(progressbar.innerText){ + let newtitle = 'Stable Diffusion - ' + progressbar.innerText + if(document.title != newtitle){ + document.title = newtitle; + } + }else{ + let newtitle = 'Stable Diffusion' + if(document.title != newtitle){ + document.title = newtitle; + } + } + } + if(progressbar!= null && progressbar != global_progressbars[id_progressbar]){ global_progressbars[id_progressbar] = progressbar @@ -19,7 +19,7 @@ clip_package = os.environ.get('CLIP_PACKAGE', "git+https://github.com/openai/CLI stable_diffusion_commit_hash = os.environ.get('STABLE_DIFFUSION_COMMIT_HASH', "69ae4b35e0a0f6ee1af8bb9a5d0016ccb27e36dc")
taming_transformers_commit_hash = os.environ.get('TAMING_TRANSFORMERS_COMMIT_HASH', "24268930bf1dce879235a7fddd0b2355b84d7ea6")
-k_diffusion_commit_hash = os.environ.get('K_DIFFUSION_COMMIT_HASH', "a7ec1974d4ccb394c2dca275f42cd97490618924")
+k_diffusion_commit_hash = os.environ.get('K_DIFFUSION_COMMIT_HASH', "f4e99857772fc3a126ba886aadf795a332774878")
codeformer_commit_hash = os.environ.get('CODEFORMER_COMMIT_HASH', "c5b4593074ba6214284d6acd5f1719b6c5d739af")
blip_commit_hash = os.environ.get('BLIP_COMMIT_HASH', "48211a1594f1321b00f14c9f7a5b4813144b2fb9")
@@ -86,6 +86,15 @@ def git_clone(url, dir, name, commithash=None): # TODO clone into temporary dir and move if successful
if os.path.exists(dir):
+ if commithash is None:
+ return
+
+ current_hash = run(f'"{git}" -C {dir} rev-parse HEAD', None, f"Couldn't determine {name}'s hash: {commithash}").strip()
+ if current_hash == commithash:
+ return
+
+ run(f'"{git}" -C {dir} fetch', f"Fetching updates for {name}...", f"Couldn't fetch {name}")
+ run(f'"{git}" -C {dir} checkout {commithash}', f"Checking out commint for {name} with hash: {commithash}...", f"Couldn't checkout commit {commithash} for {name}")
return
run(f'"{git}" clone "{url}" "{dir}"', f"Cloning {name} into {dir}...", f"Couldn't clone {name}")
diff --git a/modules/extras.py b/modules/extras.py index 6a0d5cb0..1d9e64e5 100644 --- a/modules/extras.py +++ b/modules/extras.py @@ -100,6 +100,8 @@ def run_extras(extras_mode, image, image_folder, gfpgan_visibility, codeformer_v outputs.append(image)
+ devices.torch_gc()
+
return outputs, plaintext_to_html(info), ''
diff --git a/modules/hypernetwork.py b/modules/hypernetwork.py new file mode 100644 index 00000000..c7b86682 --- /dev/null +++ b/modules/hypernetwork.py @@ -0,0 +1,88 @@ +import glob
+import os
+import sys
+import traceback
+
+import torch
+
+from ldm.util import default
+from modules import devices, shared
+import torch
+from torch import einsum
+from einops import rearrange, repeat
+
+
+class HypernetworkModule(torch.nn.Module):
+ def __init__(self, dim, state_dict):
+ super().__init__()
+
+ self.linear1 = torch.nn.Linear(dim, dim * 2)
+ self.linear2 = torch.nn.Linear(dim * 2, dim)
+
+ self.load_state_dict(state_dict, strict=True)
+ self.to(devices.device)
+
+ def forward(self, x):
+ return x + (self.linear2(self.linear1(x)))
+
+
+class Hypernetwork:
+ filename = None
+ name = None
+
+ def __init__(self, filename):
+ self.filename = filename
+ self.name = os.path.splitext(os.path.basename(filename))[0]
+ self.layers = {}
+
+ state_dict = torch.load(filename, map_location='cpu')
+ for size, sd in state_dict.items():
+ self.layers[size] = (HypernetworkModule(size, sd[0]), HypernetworkModule(size, sd[1]))
+
+
+def load_hypernetworks(path):
+ res = {}
+
+ for filename in glob.iglob(path + '**/*.pt', recursive=True):
+ try:
+ hn = Hypernetwork(filename)
+ res[hn.name] = hn
+ except Exception:
+ print(f"Error loading hypernetwork {filename}", file=sys.stderr)
+ print(traceback.format_exc(), file=sys.stderr)
+
+ return res
+
+
+def attention_CrossAttention_forward(self, x, context=None, mask=None):
+ h = self.heads
+
+ q = self.to_q(x)
+ context = default(context, x)
+
+ hypernetwork = shared.selected_hypernetwork()
+ hypernetwork_layers = (hypernetwork.layers if hypernetwork is not None else {}).get(context.shape[2], None)
+
+ if hypernetwork_layers is not None:
+ k = self.to_k(hypernetwork_layers[0](context))
+ v = self.to_v(hypernetwork_layers[1](context))
+ else:
+ k = self.to_k(context)
+ v = self.to_v(context)
+
+ q, k, v = map(lambda t: rearrange(t, 'b n (h d) -> (b h) n d', h=h), (q, k, v))
+
+ sim = einsum('b i d, b j d -> b i j', q, k) * self.scale
+
+ if mask is not None:
+ mask = rearrange(mask, 'b ... -> b (...)')
+ max_neg_value = -torch.finfo(sim.dtype).max
+ mask = repeat(mask, 'b j -> (b h) () j', h=h)
+ sim.masked_fill_(~mask, max_neg_value)
+
+ # attention, what we cannot get enough of
+ attn = sim.softmax(dim=-1)
+
+ out = einsum('b i j, b j d -> b i d', attn, v)
+ out = rearrange(out, '(b h) n d -> b n (h d)', h=h)
+ return self.to_out(out)
diff --git a/modules/images.py b/modules/images.py index c2fadab9..29c5ee24 100644 --- a/modules/images.py +++ b/modules/images.py @@ -292,18 +292,13 @@ def apply_filename_pattern(x, p, seed, prompt): x = x.replace("[cfg]", str(p.cfg_scale))
x = x.replace("[width]", str(p.width))
x = x.replace("[height]", str(p.height))
-
- #currently disabled if using the save button, will work otherwise
- # if enabled it will cause a bug because styles is not included in the save_files data dictionary
- if hasattr(p, "styles"):
- x = x.replace("[styles]", sanitize_filename_part(", ".join([x for x in p.styles if not x == "None"]) or "None", replace_spaces=False))
-
+ x = x.replace("[styles]", sanitize_filename_part(", ".join([x for x in p.styles if not x == "None"]) or "None", replace_spaces=False))
x = x.replace("[sampler]", sanitize_filename_part(sd_samplers.samplers[p.sampler_index].name, replace_spaces=False))
- x = x.replace("[model_hash]", shared.sd_model.sd_model_hash)
+ x = x.replace("[model_hash]", getattr(p, "sd_model_hash", shared.sd_model.sd_model_hash))
x = x.replace("[date]", datetime.date.today().isoformat())
x = x.replace("[datetime]", datetime.datetime.now().strftime("%Y%m%d%H%M%S"))
- x = x.replace("[job_timestamp]", shared.state.job_timestamp)
+ x = x.replace("[job_timestamp]", getattr(p, "job_timestamp", shared.state.job_timestamp))
# Apply [prompt] at last. Because it may contain any replacement word.^M
if prompt is not None:
@@ -353,7 +348,7 @@ def get_next_sequence_number(path, basename): return result + 1
-def save_image(image, path, basename, seed=None, prompt=None, extension='png', info=None, short_filename=False, no_prompt=False, grid=False, pnginfo_section_name='parameters', p=None, existing_info=None, forced_filename=None, suffix=""):
+def save_image(image, path, basename, seed=None, prompt=None, extension='png', info=None, short_filename=False, no_prompt=False, grid=False, pnginfo_section_name='parameters', p=None, existing_info=None, forced_filename=None, suffix="", save_to_dirs=None):
if short_filename or prompt is None or seed is None:
file_decoration = ""
elif opts.save_to_dirs:
@@ -377,7 +372,8 @@ def save_image(image, path, basename, seed=None, prompt=None, extension='png', i else:
pnginfo = None
- save_to_dirs = (grid and opts.grid_save_to_dirs) or (not grid and opts.save_to_dirs and not no_prompt)
+ if save_to_dirs is None:
+ save_to_dirs = (grid and opts.grid_save_to_dirs) or (not grid and opts.save_to_dirs and not no_prompt)
if save_to_dirs:
dirname = apply_filename_pattern(opts.directories_filename_pattern or "[prompt_words]", p, seed, prompt).strip('\\ /')
@@ -431,4 +427,4 @@ def save_image(image, path, basename, seed=None, prompt=None, extension='png', i with open(f"{fullfn_without_extension}.txt", "w", encoding="utf8") as file:
file.write(info + "\n")
-
+ return fullfn
diff --git a/modules/processing.py b/modules/processing.py index 8180c63d..f773a30e 100644 --- a/modules/processing.py +++ b/modules/processing.py @@ -11,9 +11,8 @@ import cv2 from skimage import exposure
import modules.sd_hijack
-from modules import devices, prompt_parser, masking
+from modules import devices, prompt_parser, masking, sd_samplers, lowvram
from modules.sd_hijack import model_hijack
-from modules.sd_samplers import samplers, samplers_for_img2img
from modules.shared import opts, cmd_opts, state
import modules.shared as shared
import modules.face_restoration
@@ -84,7 +83,7 @@ class StableDiffusionProcessing: self.s_tmin = opts.s_tmin
self.s_tmax = float('inf') # not representable as a standard ui option
self.s_noise = opts.s_noise
-
+
if not seed_enable_extras:
self.subseed = -1
self.subseed_strength = 0
@@ -110,7 +109,7 @@ class Processed: self.width = p.width
self.height = p.height
self.sampler_index = p.sampler_index
- self.sampler = samplers[p.sampler_index].name
+ self.sampler = sd_samplers.samplers[p.sampler_index].name
self.cfg_scale = p.cfg_scale
self.steps = p.steps
self.batch_size = p.batch_size
@@ -122,6 +121,8 @@ class Processed: self.denoising_strength = getattr(p, 'denoising_strength', None)
self.extra_generation_params = p.extra_generation_params
self.index_of_first_image = index_of_first_image
+ self.styles = p.styles
+ self.job_timestamp = state.job_timestamp
self.eta = p.eta
self.ddim_discretize = p.ddim_discretize
@@ -166,6 +167,8 @@ class Processed: "extra_generation_params": self.extra_generation_params,
"index_of_first_image": self.index_of_first_image,
"infotexts": self.infotexts,
+ "styles": self.styles,
+ "job_timestamp": self.job_timestamp,
}
return json.dumps(obj)
@@ -265,7 +268,7 @@ def create_infotext(p, all_prompts, all_seeds, all_subseeds, comments, iteration generation_params = {
"Steps": p.steps,
- "Sampler": samplers[p.sampler_index].name,
+ "Sampler": sd_samplers.samplers[p.sampler_index].name,
"CFG scale": p.cfg_scale,
"Seed": all_seeds[index],
"Face restoration": (opts.face_restoration_model if p.restore_faces else None),
@@ -296,7 +299,7 @@ def process_images(p: StableDiffusionProcessing) -> Processed: assert(len(p.prompt) > 0)
else:
assert p.prompt is not None
-
+
devices.torch_gc()
seed = get_fixed_seed(p.seed)
@@ -359,8 +362,8 @@ def process_images(p: StableDiffusionProcessing) -> Processed: #uc = p.sd_model.get_learned_conditioning(len(prompts) * [p.negative_prompt])
#c = p.sd_model.get_learned_conditioning(prompts)
with devices.autocast():
- uc = prompt_parser.get_learned_conditioning(len(prompts) * [p.negative_prompt], p.steps)
- c = prompt_parser.get_learned_conditioning(prompts, p.steps)
+ uc = prompt_parser.get_learned_conditioning(shared.sd_model, len(prompts) * [p.negative_prompt], p.steps)
+ c = prompt_parser.get_multicond_learned_conditioning(shared.sd_model, prompts, p.steps)
if len(model_hijack.comments) > 0:
for comment in model_hijack.comments:
@@ -383,6 +386,13 @@ def process_images(p: StableDiffusionProcessing) -> Processed: x_samples_ddim = p.sd_model.decode_first_stage(samples_ddim)
x_samples_ddim = torch.clamp((x_samples_ddim + 1.0) / 2.0, min=0.0, max=1.0)
+ del samples_ddim
+
+ if shared.cmd_opts.lowvram or shared.cmd_opts.medvram:
+ lowvram.send_everything_to_cpu()
+
+ devices.torch_gc()
+
if opts.filter_nsfw:
import modules.safety as safety
x_samples_ddim = modules.safety.censor_batch(x_samples_ddim)
@@ -424,9 +434,15 @@ def process_images(p: StableDiffusionProcessing) -> Processed: if opts.samples_save and not p.do_not_save_samples:
images.save_image(image, p.outpath_samples, "", seeds[i], prompts[i], opts.samples_format, info=infotext(n, i), p=p)
- infotexts.append(infotext(n, i))
+ text = infotext(n, i)
+ infotexts.append(text)
+ image.info["parameters"] = text
output_images.append(image)
+ del x_samples_ddim
+
+ devices.torch_gc()
+
state.nextjob()
p.color_corrections = None
@@ -437,7 +453,9 @@ def process_images(p: StableDiffusionProcessing) -> Processed: grid = images.image_grid(output_images, p.batch_size)
if opts.return_grid:
- infotexts.insert(0, infotext())
+ text = infotext()
+ infotexts.insert(0, text)
+ grid.info["parameters"] = text
output_images.insert(0, grid)
index_of_first_image = 1
@@ -478,7 +496,7 @@ class StableDiffusionProcessingTxt2Img(StableDiffusionProcessing): self.firstphase_height_truncated = int(scale * self.height)
def sample(self, conditioning, unconditional_conditioning, seeds, subseeds, subseed_strength):
- self.sampler = samplers[self.sampler_index].constructor(self.sd_model)
+ self.sampler = sd_samplers.create_sampler_with_index(sd_samplers.samplers, self.sampler_index, self.sd_model)
if not self.enable_hr:
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)
@@ -521,13 +539,14 @@ class StableDiffusionProcessingTxt2Img(StableDiffusionProcessing): shared.state.nextjob()
- self.sampler = samplers[self.sampler_index].constructor(self.sd_model)
+ self.sampler = sd_samplers.create_sampler_with_index(sd_samplers.samplers, self.sampler_index, self.sd_model)
+
noise = create_random_tensors(samples.shape[1:], seeds=seeds, subseeds=subseeds, subseed_strength=subseed_strength, seed_resize_from_h=self.seed_resize_from_h, seed_resize_from_w=self.seed_resize_from_w, p=self)
# GC now before running the next img2img to prevent running out of memory
x = None
devices.torch_gc()
-
+
samples = self.sampler.sample_img2img(self, samples, noise, conditioning, unconditional_conditioning, steps=self.steps)
return samples
@@ -556,7 +575,7 @@ class StableDiffusionProcessingImg2Img(StableDiffusionProcessing): self.nmask = None
def init(self, all_prompts, all_seeds, all_subseeds):
- self.sampler = samplers_for_img2img[self.sampler_index].constructor(self.sd_model)
+ self.sampler = sd_samplers.create_sampler_with_index(sd_samplers.samplers_for_img2img, self.sampler_index, self.sd_model)
crop_region = None
if self.image_mask is not None:
@@ -663,4 +682,7 @@ class StableDiffusionProcessingImg2Img(StableDiffusionProcessing): if self.mask is not None:
samples = samples * self.nmask + self.init_latent * self.mask
+ del x
+ devices.torch_gc()
+
return samples
diff --git a/modules/prompt_parser.py b/modules/prompt_parser.py index 5d58c4ed..f00256f2 100644 --- a/modules/prompt_parser.py +++ b/modules/prompt_parser.py @@ -1,10 +1,7 @@ import re
from collections import namedtuple
-import torch
-from lark import Lark, Transformer, Visitor
-import functools
-
-import modules.shared as shared
+from typing import List
+import lark
# a prompt like this: "fantasy landscape with a [mountain:lake:0.25] and [an oak:a christmas tree:0.75][ in foreground::0.6][ in background:0.25] [shoddy:masterful:0.5]"
# will be represented with prompt_schedule like this (assuming steps=100):
@@ -14,25 +11,48 @@ import modules.shared as shared # [75, 'fantasy landscape with a lake and an oak in background masterful']
# [100, 'fantasy landscape with a lake and a christmas tree in background masterful']
+schedule_parser = lark.Lark(r"""
+!start: (prompt | /[][():]/+)*
+prompt: (emphasized | scheduled | plain | WHITESPACE)*
+!emphasized: "(" prompt ")"
+ | "(" prompt ":" prompt ")"
+ | "[" prompt "]"
+scheduled: "[" [prompt ":"] prompt ":" [WHITESPACE] NUMBER "]"
+WHITESPACE: /\s+/
+plain: /([^\\\[\]():]|\\.)+/
+%import common.SIGNED_NUMBER -> NUMBER
+""")
def get_learned_conditioning_prompt_schedules(prompts, steps):
- grammar = r"""
- start: prompt
- prompt: (emphasized | scheduled | weighted | plain)*
- !emphasized: "(" prompt ")"
- | "(" prompt ":" prompt ")"
- | "[" prompt "]"
- scheduled: "[" (prompt ":")? prompt ":" NUMBER "]"
- !weighted: "{" weighted_item ("|" weighted_item)* "}"
- !weighted_item: prompt (":" prompt)?
- plain: /([^\\\[\](){}:|]|\\.)+/
- %import common.SIGNED_NUMBER -> NUMBER
"""
- parser = Lark(grammar, parser='lalr')
+ >>> g = lambda p: get_learned_conditioning_prompt_schedules([p], 10)[0]
+ >>> g("test")
+ [[10, 'test']]
+ >>> g("a [b:3]")
+ [[3, 'a '], [10, 'a b']]
+ >>> g("a [b: 3]")
+ [[3, 'a '], [10, 'a b']]
+ >>> g("a [[[b]]:2]")
+ [[2, 'a '], [10, 'a [[b]]']]
+ >>> g("[(a:2):3]")
+ [[3, ''], [10, '(a:2)']]
+ >>> g("a [b : c : 1] d")
+ [[1, 'a b d'], [10, 'a c d']]
+ >>> g("a[b:[c:d:2]:1]e")
+ [[1, 'abe'], [2, 'ace'], [10, 'ade']]
+ >>> g("a [unbalanced")
+ [[10, 'a [unbalanced']]
+ >>> g("a [b:.5] c")
+ [[5, 'a c'], [10, 'a b c']]
+ >>> g("a [{b|d{:.5] c") # not handling this right now
+ [[5, 'a c'], [10, 'a {b|d{ c']]
+ >>> g("((a][:b:c [d:3]")
+ [[3, '((a][:b:c '], [10, '((a][:b:c d']]
+ """
def collect_steps(steps, tree):
l = [steps]
- class CollectSteps(Visitor):
+ class CollectSteps(lark.Visitor):
def scheduled(self, tree):
tree.children[-1] = float(tree.children[-1])
if tree.children[-1] < 1:
@@ -43,13 +63,10 @@ def get_learned_conditioning_prompt_schedules(prompts, steps): return sorted(set(l))
def at_step(step, tree):
- class AtStep(Transformer):
+ class AtStep(lark.Transformer):
def scheduled(self, args):
- if len(args) == 2:
- before, after, when = (), *args
- else:
- before, after, when = args
- yield before if step <= when else after
+ before, after, _, when = args
+ yield before or () if step <= when else after
def start(self, args):
def flatten(x):
if type(x) == str:
@@ -57,16 +74,22 @@ def get_learned_conditioning_prompt_schedules(prompts, steps): else:
for gen in x:
yield from flatten(gen)
- return ''.join(flatten(args[0]))
+ return ''.join(flatten(args))
def plain(self, args):
yield args[0].value
def __default__(self, data, children, meta):
for child in children:
yield from child
return AtStep().transform(tree)
-
+
def get_schedule(prompt):
- tree = parser.parse(prompt)
+ try:
+ tree = schedule_parser.parse(prompt)
+ except lark.exceptions.LarkError as e:
+ if 0:
+ import traceback
+ traceback.print_exc()
+ return [[steps, prompt]]
return [[t, at_step(t, tree)] for t in collect_steps(steps, tree)]
promptdict = {prompt: get_schedule(prompt) for prompt in set(prompts)}
@@ -74,11 +97,26 @@ def get_learned_conditioning_prompt_schedules(prompts, steps): ScheduledPromptConditioning = namedtuple("ScheduledPromptConditioning", ["end_at_step", "cond"])
-ScheduledPromptBatch = namedtuple("ScheduledPromptBatch", ["shape", "schedules"])
-def get_learned_conditioning(prompts, steps):
+def get_learned_conditioning(model, prompts, steps):
+ """converts a list of prompts into a list of prompt schedules - each schedule is a list of ScheduledPromptConditioning, specifying the comdition (cond),
+ and the sampling step at which this condition is to be replaced by the next one.
+ Input:
+ (model, ['a red crown', 'a [blue:green:5] jeweled crown'], 20)
+
+ Output:
+ [
+ [
+ ScheduledPromptConditioning(end_at_step=20, cond=tensor([[-0.3886, 0.0229, -0.0523, ..., -0.4901, -0.3066, 0.0674], ..., [ 0.3317, -0.5102, -0.4066, ..., 0.4119, -0.7647, -1.0160]], device='cuda:0'))
+ ],
+ [
+ ScheduledPromptConditioning(end_at_step=5, cond=tensor([[-0.3886, 0.0229, -0.0522, ..., -0.4901, -0.3067, 0.0673], ..., [-0.0192, 0.3867, -0.4644, ..., 0.1135, -0.3696, -0.4625]], device='cuda:0')),
+ ScheduledPromptConditioning(end_at_step=20, cond=tensor([[-0.3886, 0.0229, -0.0522, ..., -0.4901, -0.3067, 0.0673], ..., [-0.7352, -0.4356, -0.7888, ..., 0.6994, -0.4312, -1.2593]], device='cuda:0'))
+ ]
+ ]
+ """
res = []
prompt_schedules = get_learned_conditioning_prompt_schedules(prompts, steps)
@@ -92,7 +130,7 @@ def get_learned_conditioning(prompts, steps): continue
texts = [x[1] for x in prompt_schedule]
- conds = shared.sd_model.get_learned_conditioning(texts)
+ conds = model.get_learned_conditioning(texts)
cond_schedule = []
for i, (end_at_step, text) in enumerate(prompt_schedule):
@@ -101,22 +139,109 @@ def get_learned_conditioning(prompts, steps): cache[prompt] = cond_schedule
res.append(cond_schedule)
- return ScheduledPromptBatch((len(prompts),) + res[0][0].cond.shape, res)
+ return res
+
+
+re_AND = re.compile(r"\bAND\b")
+re_weight = re.compile(r"^(.*?)(?:\s*:\s*([-+]?(?:\d+\.?|\d*\.\d+)))?\s*$")
+
+def get_multicond_prompt_list(prompts):
+ res_indexes = []
+
+ prompt_flat_list = []
+ prompt_indexes = {}
+
+ for prompt in prompts:
+ subprompts = re_AND.split(prompt)
+
+ indexes = []
+ for subprompt in subprompts:
+ match = re_weight.search(subprompt)
+
+ text, weight = match.groups() if match is not None else (subprompt, 1.0)
+
+ weight = float(weight) if weight is not None else 1.0
+
+ index = prompt_indexes.get(text, None)
+ if index is None:
+ index = len(prompt_flat_list)
+ prompt_flat_list.append(text)
+ prompt_indexes[text] = index
+
+ indexes.append((index, weight))
+
+ res_indexes.append(indexes)
+
+ return res_indexes, prompt_flat_list, prompt_indexes
+
+
+class ComposableScheduledPromptConditioning:
+ def __init__(self, schedules, weight=1.0):
+ self.schedules: List[ScheduledPromptConditioning] = schedules
+ self.weight: float = weight
+
+
+class MulticondLearnedConditioning:
+ def __init__(self, shape, batch):
+ self.shape: tuple = shape # the shape field is needed to send this object to DDIM/PLMS
+ self.batch: List[List[ComposableScheduledPromptConditioning]] = batch
+def get_multicond_learned_conditioning(model, prompts, steps) -> MulticondLearnedConditioning:
+ """same as get_learned_conditioning, but returns a list of ScheduledPromptConditioning along with the weight objects for each prompt.
+ For each prompt, the list is obtained by splitting the prompt using the AND separator.
-def reconstruct_cond_batch(c: ScheduledPromptBatch, current_step):
- res = torch.zeros(c.shape, device=shared.device, dtype=next(shared.sd_model.parameters()).dtype)
- for i, cond_schedule in enumerate(c.schedules):
+ https://energy-based-model.github.io/Compositional-Visual-Generation-with-Composable-Diffusion-Models/
+ """
+
+ res_indexes, prompt_flat_list, prompt_indexes = get_multicond_prompt_list(prompts)
+
+ learned_conditioning = get_learned_conditioning(model, prompt_flat_list, steps)
+
+ res = []
+ for indexes in res_indexes:
+ res.append([ComposableScheduledPromptConditioning(learned_conditioning[i], weight) for i, weight in indexes])
+
+ return MulticondLearnedConditioning(shape=(len(prompts),), batch=res)
+
+
+def reconstruct_cond_batch(c: List[List[ScheduledPromptConditioning]], current_step):
+ param = c[0][0].cond
+ res = torch.zeros((len(c),) + param.shape, device=param.device, dtype=param.dtype)
+ for i, cond_schedule in enumerate(c):
target_index = 0
- for curret_index, (end_at, cond) in enumerate(cond_schedule):
+ for current, (end_at, cond) in enumerate(cond_schedule):
if current_step <= end_at:
- target_index = curret_index
+ target_index = current
break
res[i] = cond_schedule[target_index].cond
return res
+def reconstruct_multicond_batch(c: MulticondLearnedConditioning, current_step):
+ param = c.batch[0][0].schedules[0].cond
+
+ tensors = []
+ conds_list = []
+
+ for batch_no, composable_prompts in enumerate(c.batch):
+ conds_for_batch = []
+
+ for cond_index, composable_prompt in enumerate(composable_prompts):
+ target_index = 0
+ for current, (end_at, cond) in enumerate(composable_prompt.schedules):
+ if current_step <= end_at:
+ target_index = current
+ break
+
+ conds_for_batch.append((len(tensors), composable_prompt.weight))
+ tensors.append(composable_prompt.schedules[target_index].cond)
+
+ conds_list.append(conds_for_batch)
+
+ return conds_list, torch.stack(tensors).to(device=param.device, dtype=param.dtype)
+
+
re_attention = re.compile(r"""
\\\(|
\\\)|
@@ -148,23 +273,26 @@ def parse_prompt_attention(text): \\ - literal character '\'
anything else - just text
- Example:
-
- 'a (((house:1.3)) [on] a (hill:0.5), sun, (((sky))).'
-
- produces:
-
- [
- ['a ', 1.0],
- ['house', 1.5730000000000004],
- [' ', 1.1],
- ['on', 1.0],
- [' a ', 1.1],
- ['hill', 0.55],
- [', sun, ', 1.1],
- ['sky', 1.4641000000000006],
- ['.', 1.1]
- ]
+ >>> parse_prompt_attention('normal text')
+ [['normal text', 1.0]]
+ >>> parse_prompt_attention('an (important) word')
+ [['an ', 1.0], ['important', 1.1], [' word', 1.0]]
+ >>> parse_prompt_attention('(unbalanced')
+ [['unbalanced', 1.1]]
+ >>> parse_prompt_attention('\(literal\]')
+ [['(literal]', 1.0]]
+ >>> parse_prompt_attention('(unnecessary)(parens)')
+ [['unnecessaryparens', 1.1]]
+ >>> parse_prompt_attention('a (((house:1.3)) [on] a (hill:0.5), sun, (((sky))).')
+ [['a ', 1.0],
+ ['house', 1.5730000000000004],
+ [' ', 1.1],
+ ['on', 1.0],
+ [' a ', 1.1],
+ ['hill', 0.55],
+ [', sun, ', 1.1],
+ ['sky', 1.4641000000000006],
+ ['.', 1.1]]
"""
res = []
@@ -206,4 +334,19 @@ def parse_prompt_attention(text): if len(res) == 0:
res = [["", 1.0]]
+ # merge runs of identical weights
+ i = 0
+ while i + 1 < len(res):
+ if res[i][1] == res[i + 1][1]:
+ res[i][0] += res[i + 1][0]
+ res.pop(i + 1)
+ else:
+ i += 1
+
return res
+
+if __name__ == "__main__":
+ import doctest
+ doctest.testmod(optionflags=doctest.NORMALIZE_WHITESPACE)
+else:
+ import torch # doctest faster
diff --git a/modules/sd_hijack.py b/modules/sd_hijack.py index 3fa06242..d68f89cc 100644 --- a/modules/sd_hijack.py +++ b/modules/sd_hijack.py @@ -5,9 +5,10 @@ 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
+from modules import prompt_parser, devices, sd_hijack_optimizations, shared, hypernetwork
from modules.shared import opts, device, cmd_opts
import ldm.modules.attention
@@ -19,16 +20,19 @@ diffusionmodules_model_AttnBlock_forward = ldm.modules.diffusionmodules.model.At def apply_optimizations():
+ undo_optimizations()
+
+ ldm.modules.diffusionmodules.model.nonlinearity = silu
+
if cmd_opts.opt_split_attention_v1:
ldm.modules.attention.CrossAttention.forward = sd_hijack_optimizations.split_cross_attention_forward_v1
elif not cmd_opts.disable_opt_split_attention and (cmd_opts.opt_split_attention or torch.cuda.is_available()):
ldm.modules.attention.CrossAttention.forward = sd_hijack_optimizations.split_cross_attention_forward
- ldm.modules.diffusionmodules.model.nonlinearity = sd_hijack_optimizations.nonlinearity_hijack
ldm.modules.diffusionmodules.model.AttnBlock.forward = sd_hijack_optimizations.cross_attention_attnblock_forward
def undo_optimizations():
- ldm.modules.attention.CrossAttention.forward = attention_CrossAttention_forward
+ ldm.modules.attention.CrossAttention.forward = hypernetwork.attention_CrossAttention_forward
ldm.modules.diffusionmodules.model.nonlinearity = diffusionmodules_model_nonlinearity
ldm.modules.diffusionmodules.model.AttnBlock.forward = diffusionmodules_model_AttnBlock_forward
diff --git a/modules/sd_hijack_optimizations.py b/modules/sd_hijack_optimizations.py index 9c079e57..d9cca485 100644 --- a/modules/sd_hijack_optimizations.py +++ b/modules/sd_hijack_optimizations.py @@ -5,6 +5,8 @@ from torch import einsum from ldm.util import default
from einops import rearrange
+from modules import shared
+
# see https://github.com/basujindal/stable-diffusion/pull/117 for discussion
def split_cross_attention_forward_v1(self, x, context=None, mask=None):
@@ -42,8 +44,19 @@ def split_cross_attention_forward(self, x, context=None, mask=None): q_in = self.to_q(x)
context = default(context, x)
- k_in = self.to_k(context) * self.scale
- v_in = self.to_v(context)
+
+ hypernetwork = shared.selected_hypernetwork()
+ hypernetwork_layers = (hypernetwork.layers if hypernetwork is not None else {}).get(context.shape[2], None)
+
+ if hypernetwork_layers is not None:
+ k_in = self.to_k(hypernetwork_layers[0](context))
+ v_in = self.to_v(hypernetwork_layers[1](context))
+ else:
+ k_in = self.to_k(context)
+ v_in = self.to_v(context)
+
+ k_in *= self.scale
+
del context, x
q, k, v = map(lambda t: rearrange(t, 'b n (h d) -> (b h) n d', h=h), (q_in, k_in, v_in))
@@ -92,14 +105,6 @@ def split_cross_attention_forward(self, x, context=None, mask=None): return self.to_out(r2)
-def nonlinearity_hijack(x):
- # swish
- t = torch.sigmoid(x)
- x *= t
- del t
-
- return x
-
def cross_attention_attnblock_forward(self, x):
h_ = x
h_ = self.norm(h_)
diff --git a/modules/sd_models.py b/modules/sd_models.py index 5f992064..8f794b47 100644 --- a/modules/sd_models.py +++ b/modules/sd_models.py @@ -134,6 +134,14 @@ def load_model_weights(model, checkpoint_file, sd_model_hash): devices.dtype = torch.float32 if shared.cmd_opts.no_half else torch.float16
+ vae_file = os.path.splitext(checkpoint_file)[0] + ".vae.pt"
+ if os.path.exists(vae_file):
+ print(f"Loading VAE weights from: {vae_file}")
+ vae_ckpt = torch.load(vae_file, map_location="cpu")
+ vae_dict = {k: v for k, v in vae_ckpt["state_dict"].items() if k[0:4] != "loss"}
+
+ model.first_stage_model.load_state_dict(vae_dict)
+
model.sd_model_hash = sd_model_hash
model.sd_model_checkpint = checkpoint_file
diff --git a/modules/sd_samplers.py b/modules/sd_samplers.py index dbf570d2..df17e93c 100644 --- a/modules/sd_samplers.py +++ b/modules/sd_samplers.py @@ -13,31 +13,57 @@ from modules.shared import opts, cmd_opts, state import modules.shared as shared
-SamplerData = namedtuple('SamplerData', ['name', 'constructor', 'aliases'])
+SamplerData = namedtuple('SamplerData', ['name', 'constructor', 'aliases', 'options'])
samplers_k_diffusion = [
- ('Euler a', 'sample_euler_ancestral', ['k_euler_a']),
- ('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']),
- ('DPM fast', 'sample_dpm_fast', ['k_dpm_fast']),
- ('DPM adaptive', 'sample_dpm_adaptive', ['k_dpm_ad']),
+ ('Euler a', 'sample_euler_ancestral', ['k_euler_a'], {}),
+ ('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'], {}),
+ ('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'}),
]
samplers_data_k_diffusion = [
- SamplerData(label, lambda model, funcname=funcname: KDiffusionSampler(funcname, model), aliases)
- for label, funcname, aliases in samplers_k_diffusion
+ SamplerData(label, lambda model, funcname=funcname: KDiffusionSampler(funcname, model), aliases, options)
+ for label, funcname, aliases, options in samplers_k_diffusion
if hasattr(k_diffusion.sampling, funcname)
]
-samplers = [
+all_samplers = [
*samplers_data_k_diffusion,
- SamplerData('DDIM', lambda model: VanillaStableDiffusionSampler(ldm.models.diffusion.ddim.DDIMSampler, model), []),
- SamplerData('PLMS', lambda model: VanillaStableDiffusionSampler(ldm.models.diffusion.plms.PLMSSampler, model), []),
+ SamplerData('DDIM', lambda model: VanillaStableDiffusionSampler(ldm.models.diffusion.ddim.DDIMSampler, model), [], {}),
+ SamplerData('PLMS', lambda model: VanillaStableDiffusionSampler(ldm.models.diffusion.plms.PLMSSampler, model), [], {}),
]
-samplers_for_img2img = [x for x in samplers if x.name not in ['PLMS', 'DPM fast', 'DPM adaptive']]
+
+samplers = []
+samplers_for_img2img = []
+
+
+def create_sampler_with_index(list_of_configs, index, model):
+ config = list_of_configs[index]
+ sampler = config.constructor(model)
+ sampler.config = config
+
+ return sampler
+
+
+def set_samplers():
+ global samplers, samplers_for_img2img
+
+ hidden = set(opts.hide_samplers)
+ hidden_img2img = set(opts.hide_samplers + ['PLMS', 'DPM fast', 'DPM adaptive'])
+
+ samplers = [x for x in all_samplers if x.name not in hidden]
+ samplers_for_img2img = [x for x in all_samplers if x.name not in hidden_img2img]
+
+
+set_samplers()
sampler_extra_params = {
'sample_euler': ['s_churn', 's_tmin', 's_tmax', 's_noise'],
@@ -104,14 +130,18 @@ class VanillaStableDiffusionSampler: self.step = 0
self.eta = None
self.default_eta = 0.0
+ self.config = None
def number_of_needed_noises(self, p):
return 0
def p_sample_ddim_hook(self, x_dec, cond, ts, unconditional_conditioning, *args, **kwargs):
- cond = prompt_parser.reconstruct_cond_batch(cond, self.step)
+ conds_list, tensor = prompt_parser.reconstruct_multicond_batch(cond, self.step)
unconditional_conditioning = prompt_parser.reconstruct_cond_batch(unconditional_conditioning, self.step)
+ assert all([len(conds) == 1 for conds in conds_list]), 'composition via AND is not supported for DDIM/PLMS samplers'
+ cond = tensor
+
if self.mask is not None:
img_orig = self.sampler.model.q_sample(self.init_latent, ts)
x_dec = img_orig * self.mask + self.nmask * x_dec
@@ -183,19 +213,31 @@ class CFGDenoiser(torch.nn.Module): self.step = 0
def forward(self, x, sigma, uncond, cond, cond_scale):
- cond = prompt_parser.reconstruct_cond_batch(cond, self.step)
+ conds_list, tensor = prompt_parser.reconstruct_multicond_batch(cond, self.step)
uncond = prompt_parser.reconstruct_cond_batch(uncond, self.step)
+ batch_size = len(conds_list)
+ repeats = [len(conds_list[i]) for i in range(batch_size)]
+
+ x_in = torch.cat([torch.stack([x[i] for _ in range(n)]) for i, n in enumerate(repeats)] + [x])
+ sigma_in = torch.cat([torch.stack([sigma[i] for _ in range(n)]) for i, n in enumerate(repeats)] + [sigma])
+ cond_in = torch.cat([tensor, uncond])
+
if shared.batch_cond_uncond:
- x_in = torch.cat([x] * 2)
- sigma_in = torch.cat([sigma] * 2)
- cond_in = torch.cat([uncond, cond])
- uncond, cond = self.inner_model(x_in, sigma_in, cond=cond_in).chunk(2)
- denoised = uncond + (cond - uncond) * cond_scale
+ x_out = self.inner_model(x_in, sigma_in, cond=cond_in)
else:
- uncond = self.inner_model(x, sigma, cond=uncond)
- cond = self.inner_model(x, sigma, cond=cond)
- denoised = uncond + (cond - uncond) * cond_scale
+ x_out = torch.zeros_like(x_in)
+ for batch_offset in range(0, x_out.shape[0], batch_size):
+ a = batch_offset
+ b = a + batch_size
+ x_out[a:b] = self.inner_model(x_in[a:b], sigma_in[a:b], cond=cond_in[a:b])
+
+ denoised_uncond = x_out[-batch_size:]
+ denoised = torch.clone(denoised_uncond)
+
+ for i, conds in enumerate(conds_list):
+ for cond_index, weight in conds:
+ denoised[i] += (x_out[cond_index] - denoised_uncond[i]) * (weight * cond_scale)
if self.mask is not None:
denoised = self.init_latent * self.mask + self.nmask * denoised
@@ -250,6 +292,7 @@ class KDiffusionSampler: self.stop_at = None
self.eta = None
self.default_eta = 1.0
+ self.config = None
def callback_state(self, d):
store_latent(d["denoised"])
@@ -295,9 +338,11 @@ class KDiffusionSampler: steps, t_enc = setup_img2img_steps(p, steps)
if p.sampler_noise_scheduler_override:
- sigmas = p.sampler_noise_scheduler_override(steps)
+ 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.model_wrap.get_sigmas(steps)
noise = noise * sigmas[steps - t_enc - 1]
xi = x + noise
@@ -314,9 +359,12 @@ class KDiffusionSampler: steps = steps or p.steps
if p.sampler_noise_scheduler_override:
- sigmas = p.sampler_noise_scheduler_override(steps)
+ 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.model_wrap.get_sigmas(steps)
+
x = x * sigmas[0]
extra_params_kwargs = self.initialize(p)
diff --git a/modules/shared.py b/modules/shared.py index ff4e5fa3..879d8424 100644 --- a/modules/shared.py +++ b/modules/shared.py @@ -13,11 +13,11 @@ import modules.memmon import modules.sd_models
import modules.styles
import modules.devices as devices
-from modules.paths import script_path, sd_path
+from modules import sd_samplers, hypernetwork
+from modules.paths import models_path, script_path, sd_path
sd_model_file = os.path.join(script_path, 'model.ckpt')
default_sd_model_file = sd_model_file
-model_path = os.path.join(script_path, 'models')
parser = argparse.ArgumentParser()
parser.add_argument("--config", type=str, default=os.path.join(sd_path, "configs/stable-diffusion/v1-inference.yaml"), help="path to config which constructs model",)
parser.add_argument("--ckpt", type=str, default=sd_model_file, help="path to checkpoint of stable diffusion model; if specified, this checkpoint will be added to the list of checkpoints and loaded",)
@@ -35,14 +35,14 @@ parser.add_argument("--always-batch-cond-uncond", action='store_true', help="dis parser.add_argument("--unload-gfpgan", action='store_true', help="does not do anything.")
parser.add_argument("--precision", type=str, help="evaluate at this precision", choices=["full", "autocast"], default="autocast")
parser.add_argument("--share", action='store_true', help="use share=True for gradio and make the UI accessible through their site (doesn't work for me but you might have better luck)")
-parser.add_argument("--codeformer-models-path", type=str, help="Path to directory with codeformer model file(s).", default=os.path.join(model_path, 'Codeformer'))
-parser.add_argument("--gfpgan-models-path", type=str, help="Path to directory with GFPGAN model file(s).", default=os.path.join(model_path, 'GFPGAN'))
-parser.add_argument("--esrgan-models-path", type=str, help="Path to directory with ESRGAN model file(s).", default=os.path.join(model_path, 'ESRGAN'))
-parser.add_argument("--bsrgan-models-path", type=str, help="Path to directory with BSRGAN model file(s).", default=os.path.join(model_path, 'BSRGAN'))
-parser.add_argument("--realesrgan-models-path", type=str, help="Path to directory with RealESRGAN model file(s).", default=os.path.join(model_path, 'RealESRGAN'))
-parser.add_argument("--scunet-models-path", type=str, help="Path to directory with ScuNET model file(s).", default=os.path.join(model_path, 'ScuNET'))
-parser.add_argument("--swinir-models-path", type=str, help="Path to directory with SwinIR model file(s).", default=os.path.join(model_path, 'SwinIR'))
-parser.add_argument("--ldsr-models-path", type=str, help="Path to directory with LDSR model file(s).", default=os.path.join(model_path, 'LDSR'))
+parser.add_argument("--codeformer-models-path", type=str, help="Path to directory with codeformer model file(s).", default=os.path.join(models_path, 'Codeformer'))
+parser.add_argument("--gfpgan-models-path", type=str, help="Path to directory with GFPGAN model file(s).", default=os.path.join(models_path, 'GFPGAN'))
+parser.add_argument("--esrgan-models-path", type=str, help="Path to directory with ESRGAN model file(s).", default=os.path.join(models_path, 'ESRGAN'))
+parser.add_argument("--bsrgan-models-path", type=str, help="Path to directory with BSRGAN model file(s).", default=os.path.join(models_path, 'BSRGAN'))
+parser.add_argument("--realesrgan-models-path", type=str, help="Path to directory with RealESRGAN model file(s).", default=os.path.join(models_path, 'RealESRGAN'))
+parser.add_argument("--scunet-models-path", type=str, help="Path to directory with ScuNET model file(s).", default=os.path.join(models_path, 'ScuNET'))
+parser.add_argument("--swinir-models-path", type=str, help="Path to directory with SwinIR model file(s).", default=os.path.join(models_path, 'SwinIR'))
+parser.add_argument("--ldsr-models-path", type=str, help="Path to directory with LDSR model file(s).", default=os.path.join(models_path, 'LDSR'))
parser.add_argument("--opt-split-attention", action='store_true', help="force-enables cross-attention layer optimization. By default, it's on for torch.cuda and off for other torch devices.")
parser.add_argument("--disable-opt-split-attention", action='store_true', help="force-disables cross-attention layer optimization")
parser.add_argument("--opt-split-attention-v1", action='store_true', help="enable older version of split attention optimization that does not consume all the VRAM it can find")
@@ -55,6 +55,7 @@ parser.add_argument("--hide-ui-dir-config", action='store_true', help="hide dire parser.add_argument("--ui-settings-file", type=str, help="filename to use for ui settings", default=os.path.join(script_path, 'config.json'))
parser.add_argument("--gradio-debug", action='store_true', help="launch gradio with --debug option")
parser.add_argument("--gradio-auth", type=str, help='set gradio authentication like "username:password"; or comma-delimit multiple like "u1:p1,u2:p2,u3:p3"', default=None)
+parser.add_argument("--gradio-img2img-tool", type=str, help='gradio image uploader tool: can be either editor for ctopping, or color-sketch for drawing', choices=["color-sketch", "editor"], default="editor")
parser.add_argument("--opt-channelslast", action='store_true', help="change memory type for stable diffusion to channels last")
parser.add_argument("--styles-file", type=str, help="filename to use for styles", default=os.path.join(script_path, 'styles.csv'))
parser.add_argument("--autolaunch", action='store_true', help="open the webui URL in the system's default browser upon launch", default=False)
@@ -75,6 +76,12 @@ parallel_processing_allowed = not cmd_opts.lowvram and not cmd_opts.medvram config_filename = cmd_opts.ui_settings_file
+hypernetworks = hypernetwork.load_hypernetworks(os.path.join(models_path, 'hypernetworks'))
+
+
+def selected_hypernetwork():
+ return hypernetworks.get(opts.sd_hypernetwork, None)
+
class State:
interrupted = False
@@ -205,6 +212,7 @@ options_templates.update(options_section(('system', "System"), { options_templates.update(options_section(('sd', "Stable Diffusion"), {
"sd_model_checkpoint": OptionInfo(None, "Stable Diffusion checkpoint", gr.Dropdown, lambda: {"choices": modules.sd_models.checkpoint_tiles()}),
+ "sd_hypernetwork": OptionInfo("None", "Stable Diffusion finetune hypernetwork", gr.Dropdown, lambda: {"choices": ["None"] + [x for x in hypernetworks.keys()]}),
"img2img_color_correction": OptionInfo(False, "Apply color correction to img2img results to match original colors."),
"save_images_before_color_correction": OptionInfo(False, "Save a copy of image before applying color correction to img2img results"),
"img2img_fix_steps": OptionInfo(False, "With img2img, do exactly the amount of steps the slider specifies (normally you'd do less with less denoising)."),
@@ -234,17 +242,20 @@ options_templates.update(options_section(('ui', "User interface"), { "font": OptionInfo("", "Font for image grids that have text"),
"js_modal_lightbox": OptionInfo(True, "Enable full page image viewer"),
"js_modal_lightbox_initialy_zoomed": OptionInfo(True, "Show images zoomed in by default in full page image viewer"),
+ "show_progress_in_title": OptionInfo(True, "Show generation progress in window title."),
}))
options_templates.update(options_section(('sampler-params', "Sampler parameters"), {
- "eta_ddim": OptionInfo(0.0, "eta (noise multiplier) for DDIM", gr.Slider, {"minimum": 0.0, "maximum": 1.0, "step": 0.01}),
- "eta_ancestral": OptionInfo(1.0, "eta (noise multiplier) for ancestral samplers", gr.Slider, {"minimum": 0.0, "maximum": 1.0, "step": 0.01}),
- "ddim_discretize": OptionInfo('uniform', "img2img DDIM discretize", gr.Radio, {"choices": ['uniform', 'quad']}),
- 's_churn': OptionInfo(0.0, "sigma churn", gr.Slider, {"minimum": 0.0, "maximum": 1.0, "step": 0.01}),
- 's_tmin': OptionInfo(0.0, "sigma tmin", gr.Slider, {"minimum": 0.0, "maximum": 1.0, "step": 0.01}),
- 's_noise': OptionInfo(1.0, "sigma noise", gr.Slider, {"minimum": 0.0, "maximum": 1.0, "step": 0.01}),
+ "hide_samplers": OptionInfo([], "Hide samplers in user interface (requires restart)", gr.CheckboxGroup, lambda: {"choices": [x.name for x in sd_samplers.all_samplers]}),
+ "eta_ddim": OptionInfo(0.0, "eta (noise multiplier) for DDIM", gr.Slider, {"minimum": 0.0, "maximum": 1.0, "step": 0.01}),
+ "eta_ancestral": OptionInfo(1.0, "eta (noise multiplier) for ancestral samplers", gr.Slider, {"minimum": 0.0, "maximum": 1.0, "step": 0.01}),
+ "ddim_discretize": OptionInfo('uniform', "img2img DDIM discretize", gr.Radio, {"choices": ['uniform', 'quad']}),
+ 's_churn': OptionInfo(0.0, "sigma churn", gr.Slider, {"minimum": 0.0, "maximum": 1.0, "step": 0.01}),
+ 's_tmin': OptionInfo(0.0, "sigma tmin", gr.Slider, {"minimum": 0.0, "maximum": 1.0, "step": 0.01}),
+ 's_noise': OptionInfo(1.0, "sigma noise", gr.Slider, {"minimum": 0.0, "maximum": 1.0, "step": 0.01}),
}))
+
class Options:
data = None
data_labels = options_templates
diff --git a/modules/textual_inversion/preprocess.py b/modules/textual_inversion/preprocess.py index f545a993..f1c002a2 100644 --- a/modules/textual_inversion/preprocess.py +++ b/modules/textual_inversion/preprocess.py @@ -1,5 +1,7 @@ import os
from PIL import Image, ImageOps
+import platform
+import sys
import tqdm
from modules import shared, images
@@ -10,7 +12,7 @@ def preprocess(process_src, process_dst, process_flip, process_split, process_ca src = os.path.abspath(process_src)
dst = os.path.abspath(process_dst)
- assert src != dst, 'same directory specified as source and desitnation'
+ assert src != dst, 'same directory specified as source and destination'
os.makedirs(dst, exist_ok=True)
@@ -25,6 +27,7 @@ def preprocess(process_src, process_dst, process_flip, process_split, process_ca def save_pic_with_caption(image, index):
if process_caption:
caption = "-" + shared.interrogator.generate_caption(image)
+ caption = sanitize_caption(os.path.join(dst, f"{index:05}-{subindex[0]}"), caption, ".png")
else:
caption = filename
caption = os.path.splitext(caption)[0]
@@ -75,3 +78,27 @@ def preprocess(process_src, process_dst, process_flip, process_split, process_ca if process_caption:
shared.interrogator.send_blip_to_ram()
+
+def sanitize_caption(base_path, original_caption, suffix):
+ operating_system = platform.system().lower()
+ if (operating_system == "windows"):
+ invalid_path_characters = "\\/:*?\"<>|"
+ max_path_length = 259
+ else:
+ invalid_path_characters = "/" #linux/macos
+ max_path_length = 1023
+ caption = original_caption
+ for invalid_character in invalid_path_characters:
+ caption = caption.replace(invalid_character, "")
+ fixed_path_length = len(base_path) + len(suffix)
+ if fixed_path_length + len(caption) <= max_path_length:
+ return caption
+ caption_tokens = caption.split()
+ new_caption = ""
+ for token in caption_tokens:
+ last_caption = new_caption
+ new_caption = new_caption + token + " "
+ if (len(new_caption) + fixed_path_length - 1 > max_path_length):
+ break
+ print(f"\nPath will be too long. Truncated caption: {original_caption}\nto: {last_caption}", file=sys.stderr)
+ return last_caption.strip()
diff --git a/modules/ui.py b/modules/ui.py index ae98219a..4e136ef5 100644 --- a/modules/ui.py +++ b/modules/ui.py @@ -35,8 +35,8 @@ import modules.gfpgan_model import modules.codeformer_model
import modules.styles
import modules.generation_parameters_copypaste
-from modules.prompt_parser import get_learned_conditioning_prompt_schedules
-from modules.images import apply_filename_pattern, get_next_sequence_number
+from modules import prompt_parser
+from modules.images import save_image
import modules.textual_inversion.ui
# this is a fix for Windows users. Without it, javascript files will be served with text/html content-type and the bowser will not show any UI
@@ -115,20 +115,13 @@ def save_files(js_data, images, index): p = MyObject(data)
path = opts.outdir_save
save_to_dirs = opts.use_save_to_dirs_for_ui
-
- if save_to_dirs:
- dirname = apply_filename_pattern(opts.directories_filename_pattern or "[prompt_words]", p, p.seed, p.prompt)
- path = os.path.join(opts.outdir_save, dirname)
-
- os.makedirs(path, exist_ok=True)
-
+ extension: str = opts.samples_format
+ start_index = 0
if index > -1 and opts.save_selected_only and (index >= data["index_of_first_image"]): # ensures we are looking at a specific non-grid picture, and we have save_selected_only
images = [images[index]]
- infotexts = [data["infotexts"][index]]
- else:
- infotexts = data["infotexts"]
+ start_index = index
with open(os.path.join(opts.outdir_save, "log.csv"), "a", encoding="utf8", newline='') as file:
at_start = file.tell() == 0
@@ -136,37 +129,18 @@ def save_files(js_data, images, index): if at_start:
writer.writerow(["prompt", "seed", "width", "height", "sampler", "cfgs", "steps", "filename", "negative_prompt"])
- file_decoration = opts.samples_filename_pattern or "[seed]-[prompt_spaces]"
- if file_decoration != "":
- file_decoration = "-" + file_decoration.lower()
- file_decoration = apply_filename_pattern(file_decoration, p, p.seed, p.prompt)
- truncated = (file_decoration[:240] + '..') if len(file_decoration) > 240 else file_decoration
- filename_base = truncated
- extension = opts.samples_format.lower()
-
- basecount = get_next_sequence_number(path, "")
- for i, filedata in enumerate(images):
- file_number = f"{basecount+i:05}"
- filename = file_number + filename_base + f".{extension}"
- filepath = os.path.join(path, filename)
-
-
+ for image_index, filedata in enumerate(images, start_index):
if filedata.startswith("data:image/png;base64,"):
filedata = filedata[len("data:image/png;base64,"):]
image = Image.open(io.BytesIO(base64.decodebytes(filedata.encode('utf-8'))))
- if opts.enable_pnginfo and extension == 'png':
- pnginfo = PngImagePlugin.PngInfo()
- pnginfo.add_text('parameters', infotexts[i])
- image.save(filepath, pnginfo=pnginfo)
- else:
- image.save(filepath, quality=opts.jpeg_quality)
- if opts.enable_pnginfo and extension in ("jpg", "jpeg", "webp"):
- piexif.insert(piexif.dump({"Exif": {
- piexif.ExifIFD.UserComment: piexif.helper.UserComment.dump(infotexts[i], encoding="unicode")
- }}), filepath)
+ is_grid = image_index < p.index_of_first_image
+ i = 0 if is_grid else (image_index - p.index_of_first_image)
+ fullfn = save_image(image, path, "", seed=p.all_seeds[i], prompt=p.all_prompts[i], extension=extension, info=p.infotexts[image_index], grid=is_grid, p=p, save_to_dirs=save_to_dirs)
+
+ filename = os.path.relpath(fullfn, path)
filenames.append(filename)
writer.writerow([data["prompt"], data["seed"], data["width"], data["height"], data["sampler"], data["cfg_scale"], data["steps"], filenames[0], data["negative_prompt"]])
@@ -197,6 +171,11 @@ def wrap_gradio_call(func, extra_outputs=None): res = extra_outputs_array + [f"<div class='error'>{plaintext_to_html(type(e).__name__+': '+str(e))}</div>"]
elapsed = time.perf_counter() - t
+ elapsed_m = int(elapsed // 60)
+ elapsed_s = elapsed % 60
+ elapsed_text = f"{elapsed_s:.2f}s"
+ if (elapsed_m > 0):
+ elapsed_text = f"{elapsed_m}m "+elapsed_text
if run_memmon:
mem_stats = {k: -(v//-(1024*1024)) for k, v in shared.mem_mon.stop().items()}
@@ -211,7 +190,7 @@ def wrap_gradio_call(func, extra_outputs=None): vram_html = ''
# last item is always HTML
- res[-1] += f"<div class='performance'><p class='time'>Time taken: <wbr>{elapsed:.2f}s</p>{vram_html}</div>"
+ res[-1] += f"<div class='performance'><p class='time'>Time taken: <wbr>{elapsed_text}</p>{vram_html}</div>"
shared.state.interrupted = False
shared.state.job_count = 0
@@ -395,7 +374,9 @@ def connect_reuse_seed(seed: gr.Number, reuse_seed: gr.Button, generation_info: def update_token_counter(text, steps):
try:
- prompt_schedules = get_learned_conditioning_prompt_schedules([text], steps)
+ _, prompt_flat_list, _ = prompt_parser.get_multicond_prompt_list([text])
+ prompt_schedules = prompt_parser.get_learned_conditioning_prompt_schedules(prompt_flat_list, steps)
+
except Exception:
# a parsing error can happen here during typing, and we don't want to bother the user with
# messages related to it in console
@@ -652,7 +633,7 @@ def create_ui(wrap_gradio_gpu_call): 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", show_label=False, source="upload", interactive=True, type="pil")
+ 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)
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="sketch", image_mode="RGBA")
@@ -1219,6 +1200,7 @@ def create_ui(wrap_gradio_gpu_call): )
def request_restart():
+ shared.state.interrupt()
settings_interface.gradio_ref.do_restart = True
restart_gradio.click(
diff --git a/scripts/img2imgalt.py b/scripts/img2imgalt.py index 0ef137f7..f9894cb0 100644 --- a/scripts/img2imgalt.py +++ b/scripts/img2imgalt.py @@ -8,7 +8,6 @@ import gradio as gr from modules import processing, shared, sd_samplers, prompt_parser
from modules.processing import Processed
-from modules.sd_samplers import samplers
from modules.shared import opts, cmd_opts, state
import torch
@@ -159,7 +158,7 @@ class Script(scripts.Script): combined_noise = ((1 - randomness) * rec_noise + randomness * rand_noise) / ((randomness**2 + (1-randomness)**2) ** 0.5)
- sampler = samplers[p.sampler_index].constructor(p.sd_model)
+ sampler = sd_samplers.create_sampler_with_index(sd_samplers.samplers, p.sampler_index, p.sd_model)
sigmas = sampler.model_wrap.get_sigmas(p.steps)
diff --git a/scripts/outpainting_mk_2.py b/scripts/outpainting_mk_2.py index 11613ca3..a6468e09 100644 --- a/scripts/outpainting_mk_2.py +++ b/scripts/outpainting_mk_2.py @@ -85,8 +85,11 @@ def get_matched_noise(_np_src_image, np_mask_rgb, noise_q=1, color_variation=0.0 src_dist = np.absolute(src_fft)
src_phase = src_fft / src_dist
+ # create a generator with a static seed to make outpainting deterministic / only follow global seed
+ rng = np.random.default_rng(0)
+
noise_window = _get_gaussian_window(width, height, mode=1) # start with simple gaussian noise
- noise_rgb = np.random.random_sample((width, height, num_channels))
+ noise_rgb = rng.random((width, height, num_channels))
noise_grey = (np.sum(noise_rgb, axis=2) / 3.)
noise_rgb *= color_variation # the colorfulness of the starting noise is blended to greyscale with a parameter
for c in range(num_channels):
diff --git a/scripts/xy_grid.py b/scripts/xy_grid.py index 1237e754..c0c364df 100644 --- a/scripts/xy_grid.py +++ b/scripts/xy_grid.py @@ -1,8 +1,9 @@ from collections import namedtuple
from copy import copy
-from itertools import permutations
+from itertools import permutations, chain
import random
-
+import csv
+from io import StringIO
from PIL import Image
import numpy as np
@@ -76,6 +77,11 @@ def apply_checkpoint(p, x, xs): modules.sd_models.reload_model_weights(shared.sd_model, info)
+def apply_hypernetwork(p, x, xs):
+ hn = shared.hypernetworks.get(x, None)
+ opts.data["sd_hypernetwork"] = hn.name if hn is not None else 'None'
+
+
def format_value_add_label(p, opt, x):
if type(x) == float:
x = round(x, 8)
@@ -121,6 +127,7 @@ axis_options = [ AxisOption("Prompt order", str_permutations, apply_order, format_value_join_list),
AxisOption("Sampler", str, apply_sampler, format_value),
AxisOption("Checkpoint name", str, apply_checkpoint, format_value),
+ AxisOption("Hypernetwork", str, apply_hypernetwork, format_value),
AxisOption("Sigma Churn", float, apply_field("s_churn"), format_value_add_label),
AxisOption("Sigma min", float, apply_field("s_tmin"), format_value_add_label),
AxisOption("Sigma max", float, apply_field("s_tmax"), format_value_add_label),
@@ -168,7 +175,6 @@ re_range_float = re.compile(r"\s*([+-]?\s*\d+(?:.\d*)?)\s*-\s*([+-]?\s*\d+(?:.\d re_range_count = re.compile(r"\s*([+-]?\s*\d+)\s*-\s*([+-]?\s*\d+)(?:\s*\[(\d+)\s*\])?\s*")
re_range_count_float = re.compile(r"\s*([+-]?\s*\d+(?:.\d*)?)\s*-\s*([+-]?\s*\d+(?:.\d*)?)(?:\s*\[(\d+(?:.\d*)?)\s*\])?\s*")
-
class Script(scripts.Script):
def title(self):
return "X/Y plot"
@@ -193,11 +199,13 @@ class Script(scripts.Script): modules.processing.fix_seed(p)
p.batch_size = 1
+ initial_hn = opts.sd_hypernetwork
+
def process_axis(opt, vals):
if opt.label == 'Nothing':
return [0]
- valslist = [x.strip() for x in vals.split(",")]
+ valslist = [x.strip() for x in chain.from_iterable(csv.reader(StringIO(vals)))]
if opt.type == int:
valslist_ext = []
@@ -300,4 +308,6 @@ class Script(scripts.Script): # restore checkpoint in case it was changed by axes
modules.sd_models.reload_model_weights(shared.sd_model)
+ opts.data["sd_hypernetwork"] = initial_hn
+
return processed
@@ -408,3 +408,11 @@ input[type="range"]{ .red {
color: red;
}
+
+.gallery-item {
+ --tw-bg-opacity: 0 !important;
+}
+
+#img2img_image div.h-60{
+ height: 480px;
+}
\ No newline at end of file @@ -2,11 +2,12 @@ import os import threading
import time
import importlib
-from modules import devices
-from modules.paths import script_path
import signal
import threading
+from modules.paths import script_path
+
+from modules import devices, sd_samplers
import modules.codeformer_model as codeformer
import modules.extras
import modules.face_restoration
@@ -109,6 +110,8 @@ def webui(): time.sleep(0.5)
break
+ sd_samplers.set_samplers()
+
print('Reloading Custom Scripts')
modules.scripts.reload_scripts(os.path.join(script_path, "scripts"))
print('Reloading modules: modules.ui')
|