Mirrored_Repos/InvokeAI
1
0
Fork 0
mirror of https://github.com/invoke-ai/InvokeAI synced 2024-08-30 20:32:17 +00:00
Code Issues Packages Projects Releases Wiki Activity

Merge branch 'main' into maryhipp/preselected-image

Browse Source
This commit is contained in:
blessedcoolant
2023-08-29 01:12:09 +12:00
parent e60af40c8d 502570e083
commit 3c37245804
54 changed files with 1395 additions and 705 deletions
Download Patch File Download Diff File Expand all files Collapse all files

22
invokeai/app/api_app.py
View File

@ -1,11 +1,11 @@
# Copyright (c) 2022-2023 Kyle Schouviller (https://github.com/kyle0654) and the InvokeAI Team
import asyncio
from inspect import signature
import logging
import uvicorn
import socket
from inspect import signature
from pathlib import Path
import uvicorn
from fastapi import FastAPI
from fastapi.middleware.cors import CORSMiddleware
from fastapi.openapi.docs import get_redoc_html, get_swagger_ui_html
@ -13,7 +13,6 @@ from fastapi.openapi.utils import get_openapi
from fastapi.staticfiles import StaticFiles
from fastapi_events.handlers.local import local_handler
from fastapi_events.middleware import EventHandlerASGIMiddleware
from pathlib import Path
from pydantic.schema import schema
from .services.config import InvokeAIAppConfig
@ -30,9 +29,12 @@ from .api.sockets import SocketIO
from .invocations.baseinvocation import BaseInvocation, _InputField, _OutputField, UIConfigBase
import torch
# noinspection PyUnresolvedReferences
import invokeai.backend.util.hotfixes # noqa: F401 (monkeypatching on import)
if torch.backends.mps.is_available():
# noinspection PyUnresolvedReferences
import invokeai.backend.util.mps_fixes # noqa: F401 (monkeypatching on import)
@ -40,7 +42,6 @@ app_config = InvokeAIAppConfig.get_config()
app_config.parse_args()
logger = InvokeAILogger.getLogger(config=app_config)
# fix for windows mimetypes registry entries being borked
# see https://github.com/invoke-ai/InvokeAI/discussions/3684#discussioncomment-6391352
mimetypes.add_type("application/javascript", ".js")
@ -208,6 +209,17 @@ def invoke_api():
check_invokeai_root(app_config) # note, may exit with an exception if root not set up
if app_config.dev_reload:
try:
import jurigged
except ImportError as e:
logger.error(
'Can\'t start `--dev_reload` because jurigged is not found; `pip install -e ".[dev]"` to include development dependencies.',
exc_info=e,
)
else:
jurigged.watch(logger=InvokeAILogger.getLogger(name="jurigged").info)
port = find_port(app_config.port)
if port != app_config.port:
logger.warn(f"Port {app_config.port} in use, using port {port}")

4
invokeai/app/invocations/image.py
View File

@ -375,6 +375,9 @@ class ImageResizeInvocation(BaseInvocation):
width: int = InputField(default=512, ge=64, multiple_of=8, description="The width to resize to (px)")
height: int = InputField(default=512, ge=64, multiple_of=8, description="The height to resize to (px)")
resample_mode: PIL_RESAMPLING_MODES = InputField(default="bicubic", description="The resampling mode")
metadata: Optional[CoreMetadata] = InputField(
default=None, description=FieldDescriptions.core_metadata, ui_hidden=True
)
def invoke(self, context: InvocationContext) -> ImageOutput:
image = context.services.images.get_pil_image(self.image.image_name)
@ -393,6 +396,7 @@ class ImageResizeInvocation(BaseInvocation):
node_id=self.id,
session_id=context.graph_execution_state_id,
is_intermediate=self.is_intermediate,
metadata=self.metadata.dict() if self.metadata else None,
)
return ImageOutput(

155
invokeai/app/invocations/latent.py
View File

@ -21,6 +21,8 @@ from torchvision.transforms.functional import resize as tv_resize
from invokeai.app.invocations.metadata import CoreMetadata
from invokeai.app.invocations.primitives import (
DenoiseMaskField,
DenoiseMaskOutput,
ImageField,
ImageOutput,
LatentsField,
@ -31,8 +33,8 @@ from invokeai.app.util.controlnet_utils import prepare_control_image
from invokeai.app.util.step_callback import stable_diffusion_step_callback
from invokeai.backend.model_management.models import ModelType, SilenceWarnings
from ...backend.model_management.models import BaseModelType
from ...backend.model_management.lora import ModelPatcher
from ...backend.model_management.models import BaseModelType
from ...backend.stable_diffusion import PipelineIntermediateState
from ...backend.stable_diffusion.diffusers_pipeline import (
ConditioningData,
@ -44,16 +46,7 @@ from ...backend.stable_diffusion.diffusion.shared_invokeai_diffusion import Post
from ...backend.stable_diffusion.schedulers import SCHEDULER_MAP
from ...backend.util.devices import choose_precision, choose_torch_device
from ..models.image import ImageCategory, ResourceOrigin
from .baseinvocation import (
BaseInvocation,
FieldDescriptions,
Input,
InputField,
InvocationContext,
UIType,
tags,
title,
)
from .baseinvocation import BaseInvocation, FieldDescriptions, Input, InputField, InvocationContext, UIType, tags, title
from .compel import ConditioningField
from .controlnet_image_processors import ControlField
from .model import ModelInfo, UNetField, VaeField
@ -64,6 +57,72 @@ DEFAULT_PRECISION = choose_precision(choose_torch_device())
SAMPLER_NAME_VALUES = Literal[tuple(list(SCHEDULER_MAP.keys()))]
@title("Create Denoise Mask")
@tags("mask", "denoise")
class CreateDenoiseMaskInvocation(BaseInvocation):
"""Creates mask for denoising model run."""
# Metadata
type: Literal["create_denoise_mask"] = "create_denoise_mask"
# Inputs
vae: VaeField = InputField(description=FieldDescriptions.vae, input=Input.Connection, ui_order=0)
image: Optional[ImageField] = InputField(default=None, description="Image which will be masked", ui_order=1)
mask: ImageField = InputField(description="The mask to use when pasting", ui_order=2)
tiled: bool = InputField(default=False, description=FieldDescriptions.tiled, ui_order=3)
fp32: bool = InputField(default=DEFAULT_PRECISION == "float32", description=FieldDescriptions.fp32, ui_order=4)
def prep_mask_tensor(self, mask_image):
if mask_image.mode != "L":
mask_image = mask_image.convert("L")
mask_tensor = image_resized_to_grid_as_tensor(mask_image, normalize=False)
if mask_tensor.dim() == 3:
mask_tensor = mask_tensor.unsqueeze(0)
# if shape is not None:
# mask_tensor = tv_resize(mask_tensor, shape, T.InterpolationMode.BILINEAR)
return mask_tensor
@torch.no_grad()
def invoke(self, context: InvocationContext) -> DenoiseMaskOutput:
if self.image is not None:
image = context.services.images.get_pil_image(self.image.image_name)
image = image_resized_to_grid_as_tensor(image.convert("RGB"))
if image.dim() == 3:
image = image.unsqueeze(0)
else:
image = None
mask = self.prep_mask_tensor(
context.services.images.get_pil_image(self.mask.image_name),
)
if image is not None:
vae_info = context.services.model_manager.get_model(
**self.vae.vae.dict(),
context=context,
)
img_mask = tv_resize(mask, image.shape[-2:], T.InterpolationMode.BILINEAR, antialias=False)
masked_image = image * torch.where(img_mask < 0.5, 0.0, 1.0)
# TODO:
masked_latents = ImageToLatentsInvocation.vae_encode(vae_info, self.fp32, self.tiled, masked_image.clone())
masked_latents_name = f"{context.graph_execution_state_id}__{self.id}_masked_latents"
context.services.latents.save(masked_latents_name, masked_latents)
else:
masked_latents_name = None
mask_name = f"{context.graph_execution_state_id}__{self.id}_mask"
context.services.latents.save(mask_name, mask)
return DenoiseMaskOutput(
denoise_mask=DenoiseMaskField(
mask_name=mask_name,
masked_latents_name=masked_latents_name,
),
)
def get_scheduler(
context: InvocationContext,
scheduler_info: ModelInfo,
@ -126,10 +185,8 @@ class DenoiseLatentsInvocation(BaseInvocation):
control: Union[ControlField, list[ControlField]] = InputField(
default=None, description=FieldDescriptions.control, input=Input.Connection, ui_order=5
)
latents: Optional[LatentsField] = InputField(
description=FieldDescriptions.latents, input=Input.Connection, ui_order=4
)
mask: Optional[ImageField] = InputField(
latents: Optional[LatentsField] = InputField(description=FieldDescriptions.latents, input=Input.Connection)
denoise_mask: Optional[DenoiseMaskField] = InputField(
default=None,
description=FieldDescriptions.mask,
)
@ -342,19 +399,18 @@ class DenoiseLatentsInvocation(BaseInvocation):
return num_inference_steps, timesteps, init_timestep
def prep_mask_tensor(self, mask, context, lantents):
if mask is None:
return None
def prep_inpaint_mask(self, context, latents):
if self.denoise_mask is None:
return None, None
mask_image = context.services.images.get_pil_image(mask.image_name)
if mask_image.mode != "L":
# FIXME: why do we get passed an RGB image here? We can only use single-channel.
mask_image = mask_image.convert("L")
mask_tensor = image_resized_to_grid_as_tensor(mask_image, normalize=False)
if mask_tensor.dim() == 3:
mask_tensor = mask_tensor.unsqueeze(0)
mask_tensor = tv_resize(mask_tensor, lantents.shape[-2:], T.InterpolationMode.BILINEAR)
return 1 - mask_tensor
mask = context.services.latents.get(self.denoise_mask.mask_name)
mask = tv_resize(mask, latents.shape[-2:], T.InterpolationMode.BILINEAR, antialias=False)
if self.denoise_mask.masked_latents_name is not None:
masked_latents = context.services.latents.get(self.denoise_mask.masked_latents_name)
else:
masked_latents = None
return 1 - mask, masked_latents
@torch.no_grad()
def invoke(self, context: InvocationContext) -> LatentsOutput:
@ -375,7 +431,7 @@ class DenoiseLatentsInvocation(BaseInvocation):
if seed is None:
seed = 0
mask = self.prep_mask_tensor(self.mask, context, latents)
mask, masked_latents = self.prep_inpaint_mask(context, latents)
# Get the source node id (we are invoking the prepared node)
graph_execution_state = context.services.graph_execution_manager.get(context.graph_execution_state_id)
@ -406,6 +462,8 @@ class DenoiseLatentsInvocation(BaseInvocation):
noise = noise.to(device=unet.device, dtype=unet.dtype)
if mask is not None:
mask = mask.to(device=unet.device, dtype=unet.dtype)
if masked_latents is not None:
masked_latents = masked_latents.to(device=unet.device, dtype=unet.dtype)
scheduler = get_scheduler(
context=context,
@ -442,6 +500,7 @@ class DenoiseLatentsInvocation(BaseInvocation):
noise=noise,
seed=seed,
mask=mask,
masked_latents=masked_latents,
num_inference_steps=num_inference_steps,
conditioning_data=conditioning_data,
control_data=control_data, # list[ControlNetData]
@ -663,26 +722,11 @@ class ImageToLatentsInvocation(BaseInvocation):
tiled: bool = InputField(default=False, description=FieldDescriptions.tiled)
fp32: bool = InputField(default=DEFAULT_PRECISION == "float32", description=FieldDescriptions.fp32)
@torch.no_grad()
def invoke(self, context: InvocationContext) -> LatentsOutput:
# image = context.services.images.get(
# self.image.image_type, self.image.image_name
# )
image = context.services.images.get_pil_image(self.image.image_name)
# vae_info = context.services.model_manager.get_model(**self.vae.vae.dict())
vae_info = context.services.model_manager.get_model(
**self.vae.vae.dict(),
context=context,
)
image_tensor = image_resized_to_grid_as_tensor(image.convert("RGB"))
if image_tensor.dim() == 3:
image_tensor = einops.rearrange(image_tensor, "c h w -> 1 c h w")
@staticmethod
def vae_encode(vae_info, upcast, tiled, image_tensor):
with vae_info as vae:
orig_dtype = vae.dtype
if self.fp32:
if upcast:
vae.to(dtype=torch.float32)
use_torch_2_0_or_xformers = isinstance(
@ -707,7 +751,7 @@ class ImageToLatentsInvocation(BaseInvocation):
vae.to(dtype=torch.float16)
# latents = latents.half()
if self.tiled:
if tiled:
vae.enable_tiling()
else:
vae.disable_tiling()
@ -721,6 +765,23 @@ class ImageToLatentsInvocation(BaseInvocation):
latents = vae.config.scaling_factor * latents
latents = latents.to(dtype=orig_dtype)
return latents
@torch.no_grad()
def invoke(self, context: InvocationContext) -> LatentsOutput:
image = context.services.images.get_pil_image(self.image.image_name)
vae_info = context.services.model_manager.get_model(
**self.vae.vae.dict(),
context=context,
)
image_tensor = image_resized_to_grid_as_tensor(image.convert("RGB"))
if image_tensor.dim() == 3:
image_tensor = einops.rearrange(image_tensor, "c h w -> 1 c h w")
latents = self.vae_encode(vae_info, self.fp32, self.tiled, image_tensor)
name = f"{context.graph_execution_state_id}__{self.id}"
latents = latents.to("cpu")
context.services.latents.save(name, latents)

19
invokeai/app/invocations/primitives.py
View File

@ -294,6 +294,25 @@ class ImageCollectionInvocation(BaseInvocation):
return ImageCollectionOutput(collection=self.collection)
# endregion
# region DenoiseMask
class DenoiseMaskField(BaseModel):
"""An inpaint mask field"""
mask_name: str = Field(description="The name of the mask image")
masked_latents_name: Optional[str] = Field(description="The name of the masked image latents")
class DenoiseMaskOutput(BaseInvocationOutput):
"""Base class for nodes that output a single image"""
type: Literal["denoise_mask_output"] = "denoise_mask_output"
denoise_mask: DenoiseMaskField = OutputField(description="Mask for denoise model run")
# endregion
# region Latents

8
invokeai/app/services/config/invokeai_config.py
View File

@ -169,11 +169,13 @@ two configs are kept in separate sections of the config file:
"""
from __future__ import annotations
import os
from omegaconf import OmegaConf, DictConfig
from pathlib import Path
from typing import ClassVar, Dict, List, Literal, Union, get_type_hints, Optional
from omegaconf import OmegaConf, DictConfig
from pydantic import Field, parse_obj_as
from typing import ClassVar, Dict, List, Literal, Union, Optional, get_type_hints
from .base import InvokeAISettings
@ -233,6 +235,8 @@ class InvokeAIAppConfig(InvokeAISettings):
log_format : Literal['plain', 'color', 'syslog', 'legacy'] = Field(default="color", description='Log format. Use "plain" for text-only, "color" for colorized output, "legacy" for 2.3-style logging and "syslog" for syslog-style', category="Logging")
log_level : Literal["debug", "info", "warning", "error", "critical"] = Field(default="info", description="Emit logging messages at this level or higher", category="Logging")
dev_reload : bool = Field(default=False, description="Automatically reload when Python sources are changed.", category="Development")
version : bool = Field(default=False, description="Show InvokeAI version and exit", category="Other")
# CACHE

20
invokeai/backend/install/invokeai_configure.py
View File

@ -50,7 +50,7 @@ from invokeai.frontend.install.model_install import addModelsForm, process_and_e
# TO DO - Move all the frontend code into invokeai.frontend.install
from invokeai.frontend.install.widgets import (
SingleSelectColumns,
SingleSelectColumnsSimple,
MultiSelectColumns,
CenteredButtonPress,
FileBox,
@ -354,7 +354,6 @@ Use cursor arrows to make a checkbox selection, and space to toggle.
device = old_opts.device
attention_type = old_opts.attention_type
attention_slice_size = old_opts.attention_slice_size
self.nextrely += 1
self.add_widget_intelligent(
npyscreen.TitleFixedText,
@ -385,7 +384,7 @@ Use cursor arrows to make a checkbox selection, and space to toggle.
)
self.nextrely -= 2
self.precision = self.add_widget_intelligent(
SingleSelectColumns,
SingleSelectColumnsSimple,
columns=len(PRECISION_CHOICES),
name="Precision",
values=PRECISION_CHOICES,
@ -406,10 +405,10 @@ Use cursor arrows to make a checkbox selection, and space to toggle.
)
self.nextrely -= 2
self.device = self.add_widget_intelligent(
SingleSelectColumns,
SingleSelectColumnsSimple,
columns=len(DEVICE_CHOICES),
values=DEVICE_CHOICES,
value=DEVICE_CHOICES.index(device),
value=[DEVICE_CHOICES.index(device)],
begin_entry_at=3,
relx=30,
max_height=2,
@ -426,10 +425,10 @@ Use cursor arrows to make a checkbox selection, and space to toggle.
)
self.nextrely -= 2
self.attention_type = self.add_widget_intelligent(
SingleSelectColumns,
SingleSelectColumnsSimple,
columns=len(ATTENTION_CHOICES),
values=ATTENTION_CHOICES,
value=ATTENTION_CHOICES.index(attention_type),
value=[ATTENTION_CHOICES.index(attention_type)],
begin_entry_at=3,
max_height=2,
relx=30,
@ -448,17 +447,16 @@ Use cursor arrows to make a checkbox selection, and space to toggle.
)
self.nextrely -= 2
self.attention_slice_size = self.add_widget_intelligent(
SingleSelectColumns,
SingleSelectColumnsSimple,
columns=len(ATTENTION_SLICE_CHOICES),
values=ATTENTION_SLICE_CHOICES,
value=ATTENTION_SLICE_CHOICES.index(attention_slice_size),
value=[ATTENTION_SLICE_CHOICES.index(attention_slice_size)],
relx=30,
hidden=attention_type != "sliced",
max_height=2,
max_width=110,
scroll_exit=True,
)
self.add_widget_intelligent(
npyscreen.TitleFixedText,
name="Model RAM cache size (GB). Make this at least large enough to hold a single full model.",
@ -707,8 +705,6 @@ def initialize_rootdir(root: Path, yes_to_all: bool = False):
path = dest / "core"
path.mkdir(parents=True, exist_ok=True)
maybe_create_models_yaml(root)
def maybe_create_models_yaml(root: Path):
models_yaml = root / "configs" / "models.yaml"

41
invokeai/backend/stable_diffusion/diffusers_pipeline.py
View File

@ -144,7 +144,7 @@ def image_resized_to_grid_as_tensor(image: PIL.Image.Image, normalize: bool = Tr
w, h = trim_to_multiple_of(*image.size, multiple_of=multiple_of)
transformation = T.Compose(
[
T.Resize((h, w), T.InterpolationMode.LANCZOS),
T.Resize((h, w), T.InterpolationMode.LANCZOS, antialias=True),
T.ToTensor(),
]
)
@ -358,6 +358,7 @@ class StableDiffusionGeneratorPipeline(StableDiffusionPipeline):
callback: Callable[[PipelineIntermediateState], None] = None,
control_data: List[ControlNetData] = None,
mask: Optional[torch.Tensor] = None,
masked_latents: Optional[torch.Tensor] = None,
seed: Optional[int] = None,
) -> tuple[torch.Tensor, Optional[AttentionMapSaver]]:
if init_timestep.shape[0] == 0:
@ -376,28 +377,28 @@ class StableDiffusionGeneratorPipeline(StableDiffusionPipeline):
latents = self.scheduler.add_noise(latents, noise, batched_t)
if mask is not None:
# if no noise provided, noisify unmasked area based on seed(or 0 as fallback)
if noise is None:
noise = torch.randn(
orig_latents.shape,
dtype=torch.float32,
device="cpu",
generator=torch.Generator(device="cpu").manual_seed(seed or 0),
).to(device=orig_latents.device, dtype=orig_latents.dtype)
latents = self.scheduler.add_noise(latents, noise, batched_t)
latents = torch.lerp(
orig_latents, latents.to(dtype=orig_latents.dtype), mask.to(dtype=orig_latents.dtype)
)
if is_inpainting_model(self.unet):
# You'd think the inpainting model wouldn't be paying attention to the area it is going to repaint
# (that's why there's a mask!) but it seems to really want that blanked out.
# masked_latents = latents * torch.where(mask < 0.5, 1, 0) TODO: inpaint/outpaint/infill
if masked_latents is None:
raise Exception("Source image required for inpaint mask when inpaint model used!")
# TODO: we should probably pass this in so we don't have to try/finally around setting it.
self.invokeai_diffuser.model_forward_callback = AddsMaskLatents(self._unet_forward, mask, orig_latents)
self.invokeai_diffuser.model_forward_callback = AddsMaskLatents(
self._unet_forward, mask, masked_latents
)
else:
# if no noise provided, noisify unmasked area based on seed(or 0 as fallback)
if noise is None:
noise = torch.randn(
orig_latents.shape,
dtype=torch.float32,
device="cpu",
generator=torch.Generator(device="cpu").manual_seed(seed or 0),
).to(device=orig_latents.device, dtype=orig_latents.dtype)
latents = self.scheduler.add_noise(latents, noise, batched_t)
latents = torch.lerp(
orig_latents, latents.to(dtype=orig_latents.dtype), mask.to(dtype=orig_latents.dtype)
)
additional_guidance.append(AddsMaskGuidance(mask, orig_latents, self.scheduler, noise))
try:

20
invokeai/frontend/install/widgets.py
View File

@ -177,6 +177,8 @@ class FloatTitleSlider(npyscreen.TitleText):
class SelectColumnBase:
"""Base class for selection widget arranged in columns."""
def make_contained_widgets(self):
self._my_widgets = []
column_width = self.width // self.columns
@ -253,6 +255,7 @@ class MultiSelectColumns(SelectColumnBase, npyscreen.MultiSelect):
class SingleSelectWithChanged(npyscreen.SelectOne):
def __init__(self, *args, **kwargs):
super().__init__(*args, **kwargs)
self.on_changed = None
def h_select(self, ch):
super().h_select(ch)
@ -260,7 +263,9 @@ class SingleSelectWithChanged(npyscreen.SelectOne):
self.on_changed(self.value)
class SingleSelectColumns(SelectColumnBase, SingleSelectWithChanged):
class SingleSelectColumnsSimple(SelectColumnBase, SingleSelectWithChanged):
"""Row of radio buttons. Spacebar to select."""
def __init__(self, screen, columns: int = 1, values: list = [], **keywords):
self.columns = columns
self.value_cnt = len(values)
@ -268,12 +273,6 @@ class SingleSelectColumns(SelectColumnBase, SingleSelectWithChanged):
self.on_changed = None
super().__init__(screen, values=values, **keywords)
def when_value_edited(self):
self.h_select(self.cursor_line)
def when_cursor_moved(self):
self.h_select(self.cursor_line)
def h_cursor_line_right(self, ch):
self.h_exit_down("bye bye")
@ -281,6 +280,13 @@ class SingleSelectColumns(SelectColumnBase, SingleSelectWithChanged):
self.h_exit_up("bye bye")
class SingleSelectColumns(SingleSelectColumnsSimple):
"""Row of radio buttons. When tabbing over a selection, it is auto selected."""
def when_cursor_moved(self):
self.h_select(self.cursor_line)
class TextBoxInner(npyscreen.MultiLineEdit):
def __init__(self, *args, **kwargs):
super().__init__(*args, **kwargs)

14
invokeai/frontend/web/src/features/nodes/components/flow/nodes/Invocation/fields/InputFieldRenderer.tsx
View File

@ -10,6 +10,7 @@ import ColorInputField from './inputs/ColorInputField';
import ConditioningInputField from './inputs/ConditioningInputField';
import ControlInputField from './inputs/ControlInputField';
import ControlNetModelInputField from './inputs/ControlNetModelInputField';
import DenoiseMaskInputField from './inputs/DenoiseMaskInputField';
import EnumInputField from './inputs/EnumInputField';
import ImageCollectionInputField from './inputs/ImageCollectionInputField';
import ImageInputField from './inputs/ImageInputField';
@ -105,6 +106,19 @@ const InputFieldRenderer = ({ nodeId, fieldName }: InputFieldProps) => {
);
}
if (
field?.type === 'DenoiseMaskField' &&
fieldTemplate?.type === 'DenoiseMaskField'
) {
return (
<DenoiseMaskInputField
nodeId={nodeId}
field={field}
fieldTemplate={fieldTemplate}
/>
);
}
if (
field?.type === 'ConditioningField' &&
fieldTemplate?.type === 'ConditioningField'

17
invokeai/frontend/web/src/features/nodes/components/flow/nodes/Invocation/fields/inputs/DenoiseMaskInputField.tsx Normal file
View File

@ -0,0 +1,17 @@
import {
DenoiseMaskInputFieldTemplate,
DenoiseMaskInputFieldValue,
FieldComponentProps,
} from 'features/nodes/types/types';
import { memo } from 'react';
const DenoiseMaskInputFieldComponent = (
_props: FieldComponentProps<
DenoiseMaskInputFieldValue,
DenoiseMaskInputFieldTemplate
>
) => {
return null;
};
export default memo(DenoiseMaskInputFieldComponent);

5
invokeai/frontend/web/src/features/nodes/types/constants.ts
View File

@ -59,6 +59,11 @@ export const FIELDS: Record<FieldType, FieldUIConfig> = {
description: 'Images may be passed between nodes.',
color: 'purple.500',
},
DenoiseMaskField: {
title: 'Denoise Mask',
description: 'Denoise Mask may be passed between nodes',
color: 'red.700',
},
LatentsField: {
title: 'Latents',
description: 'Latents may be passed between nodes.',

22
invokeai/frontend/web/src/features/nodes/types/types.ts
View File

@ -64,6 +64,7 @@ export const zFieldType = z.enum([
'string',
'array',
'ImageField',
'DenoiseMaskField',
'LatentsField',
'ConditioningField',
'ControlField',
@ -120,6 +121,7 @@ export type InputFieldTemplate =
| StringInputFieldTemplate
| BooleanInputFieldTemplate
| ImageInputFieldTemplate
| DenoiseMaskInputFieldTemplate
| LatentsInputFieldTemplate
| ConditioningInputFieldTemplate
| UNetInputFieldTemplate
@ -205,6 +207,12 @@ export const zConditioningField = z.object({
});
export type ConditioningField = z.infer<typeof zConditioningField>;
export const zDenoiseMaskField = z.object({
mask_name: z.string().trim().min(1),
masked_latents_name: z.string().trim().min(1).optional(),
});
export type DenoiseMaskFieldValue = z.infer<typeof zDenoiseMaskField>;
export const zIntegerInputFieldValue = zInputFieldValueBase.extend({
type: z.literal('integer'),
value: z.number().optional(),
@ -241,6 +249,14 @@ export const zLatentsInputFieldValue = zInputFieldValueBase.extend({
});
export type LatentsInputFieldValue = z.infer<typeof zLatentsInputFieldValue>;
export const zDenoiseMaskInputFieldValue = zInputFieldValueBase.extend({
type: z.literal('DenoiseMaskField'),
value: zDenoiseMaskField.optional(),
});
export type DenoiseMaskInputFieldValue = z.infer<
typeof zDenoiseMaskInputFieldValue
>;
export const zConditioningInputFieldValue = zInputFieldValueBase.extend({
type: z.literal('ConditioningField'),
value: zConditioningField.optional(),
@ -459,6 +475,7 @@ export const zInputFieldValue = z.discriminatedUnion('type', [
zBooleanInputFieldValue,
zImageInputFieldValue,
zLatentsInputFieldValue,
zDenoiseMaskInputFieldValue,
zConditioningInputFieldValue,
zUNetInputFieldValue,
zClipInputFieldValue,
@ -532,6 +549,11 @@ export type ImageCollectionInputFieldTemplate = InputFieldTemplateBase & {
type: 'ImageCollection';
};
export type DenoiseMaskInputFieldTemplate = InputFieldTemplateBase & {
default: undefined;
type: 'DenoiseMaskField';
};
export type LatentsInputFieldTemplate = InputFieldTemplateBase & {
default: string;
type: 'LatentsField';

20
invokeai/frontend/web/src/features/nodes/util/fieldTemplateBuilders.ts
View File

@ -8,6 +8,7 @@ import {
ConditioningInputFieldTemplate,
ControlInputFieldTemplate,
ControlNetModelInputFieldTemplate,
DenoiseMaskInputFieldTemplate,
EnumInputFieldTemplate,
FieldType,
FloatInputFieldTemplate,
@ -263,6 +264,19 @@ const buildImageCollectionInputFieldTemplate = ({
return template;
};
const buildDenoiseMaskInputFieldTemplate = ({
schemaObject,
baseField,
}: BuildInputFieldArg): DenoiseMaskInputFieldTemplate => {
const template: DenoiseMaskInputFieldTemplate = {
...baseField,
type: 'DenoiseMaskField',
default: schemaObject.default ?? undefined,
};
return template;
};
const buildLatentsInputFieldTemplate = ({
schemaObject,
baseField,
@ -498,6 +512,12 @@ export const buildInputFieldTemplate = (
baseField,
});
}
if (fieldType === 'DenoiseMaskField') {
return buildDenoiseMaskInputFieldTemplate({
schemaObject: fieldSchema,
baseField,
});
}
if (fieldType === 'LatentsField') {
return buildLatentsInputFieldTemplate({
schemaObject: fieldSchema,

4
invokeai/frontend/web/src/features/nodes/util/fieldValueBuilders.ts
View File

@ -49,6 +49,10 @@ export const buildInputFieldValue = (
fieldValue.value = [];
}
if (template.type === 'DenoiseMaskField') {
fieldValue.value = undefined;
}
if (template.type === 'LatentsField') {
fieldValue.value = undefined;
}

18
invokeai/frontend/web/src/features/nodes/util/graphBuilders/addVAEToGraph.ts
View File

@ -9,6 +9,7 @@ import {
CANVAS_TEXT_TO_IMAGE_GRAPH,
IMAGE_TO_IMAGE_GRAPH,
IMAGE_TO_LATENTS,
INPAINT_CREATE_MASK,
INPAINT_IMAGE,
LATENTS_TO_IMAGE,
MAIN_MODEL_LOADER,
@ -30,6 +31,11 @@ export const addVAEToGraph = (
modelLoaderNodeId: string = MAIN_MODEL_LOADER
): void => {
const { vae } = state.generation;
const { boundingBoxScaleMethod } = state.canvas;
const isUsingScaledDimensions = ['auto', 'manual'].includes(
boundingBoxScaleMethod
);
const isAutoVae = !vae;
const metadataAccumulator = graph.nodes[METADATA_ACCUMULATOR] as
@ -76,7 +82,7 @@ export const addVAEToGraph = (
field: isAutoVae && isOnnxModel ? 'vae_decoder' : 'vae',
},
destination: {
node_id: CANVAS_OUTPUT,
node_id: isUsingScaledDimensions ? LATENTS_TO_IMAGE : CANVAS_OUTPUT,
field: 'vae',
},
});
@ -117,6 +123,16 @@ export const addVAEToGraph = (
field: 'vae',
},
},
{
source: {
node_id: isAutoVae ? modelLoaderNodeId : VAE_LOADER,
field: isAutoVae && isOnnxModel ? 'vae_decoder' : 'vae',
},
destination: {
node_id: INPAINT_CREATE_MASK,
field: 'vae',
},
},
{
source: {
node_id: isAutoVae ? modelLoaderNodeId : VAE_LOADER,

160
invokeai/frontend/web/src/features/nodes/util/graphBuilders/buildCanvasImageToImageGraph.ts
View File

@ -2,11 +2,7 @@ import { logger } from 'app/logging/logger';
import { RootState } from 'app/store/store';
import { NonNullableGraph } from 'features/nodes/types/types';
import { initialGenerationState } from 'features/parameters/store/generationSlice';
import {
ImageDTO,
ImageResizeInvocation,
ImageToLatentsInvocation,
} from 'services/api/types';
import { ImageDTO, ImageToLatentsInvocation } from 'services/api/types';
import { addControlNetToLinearGraph } from './addControlNetToLinearGraph';
import { addDynamicPromptsToGraph } from './addDynamicPromptsToGraph';
import { addLoRAsToGraph } from './addLoRAsToGraph';
@ -19,12 +15,13 @@ import {
CLIP_SKIP,
DENOISE_LATENTS,
IMAGE_TO_LATENTS,
IMG2IMG_RESIZE,
LATENTS_TO_IMAGE,
MAIN_MODEL_LOADER,
METADATA_ACCUMULATOR,
NEGATIVE_CONDITIONING,
NOISE,
POSITIVE_CONDITIONING,
RESIZE,
} from './constants';
/**
@ -43,6 +40,7 @@ export const buildCanvasImageToImageGraph = (
scheduler,
steps,
img2imgStrength: strength,
vaePrecision,
clipSkip,
shouldUseCpuNoise,
shouldUseNoiseSettings,
@ -51,7 +49,15 @@ export const buildCanvasImageToImageGraph = (
// The bounding box determines width and height, not the width and height params
const { width, height } = state.canvas.boundingBoxDimensions;
const { shouldAutoSave } = state.canvas;
const {
scaledBoundingBoxDimensions,
boundingBoxScaleMethod,
shouldAutoSave,
} = state.canvas;
const isUsingScaledDimensions = ['auto', 'manual'].includes(
boundingBoxScaleMethod
);
if (!model) {
log.error('No model found in state');
@ -104,15 +110,17 @@ export const buildCanvasImageToImageGraph = (
id: NOISE,
is_intermediate: true,
use_cpu,
width: !isUsingScaledDimensions
? width
: scaledBoundingBoxDimensions.width,
height: !isUsingScaledDimensions
? height
: scaledBoundingBoxDimensions.height,
},
[IMAGE_TO_LATENTS]: {
type: 'i2l',
id: IMAGE_TO_LATENTS,
is_intermediate: true,
// must be set manually later, bc `fit` parameter may require a resize node inserted
// image: {
// image_name: initialImage.image_name,
// },
},
[DENOISE_LATENTS]: {
type: 'denoise_latents',
@ -214,82 +222,84 @@ export const buildCanvasImageToImageGraph = (
field: 'latents',
},
},
// Decode the denoised latents to an image
],
};
// Decode Latents To Image & Handle Scaled Before Processing
if (isUsingScaledDimensions) {
graph.nodes[IMG2IMG_RESIZE] = {
id: IMG2IMG_RESIZE,
type: 'img_resize',
is_intermediate: true,
image: initialImage,
width: scaledBoundingBoxDimensions.width,
height: scaledBoundingBoxDimensions.height,
};
graph.nodes[LATENTS_TO_IMAGE] = {
id: LATENTS_TO_IMAGE,
type: 'l2i',
is_intermediate: true,
fp32: vaePrecision === 'fp32' ? true : false,
};
graph.nodes[CANVAS_OUTPUT] = {
id: CANVAS_OUTPUT,
type: 'img_resize',
is_intermediate: !shouldAutoSave,
width: width,
height: height,
};
graph.edges.push(
{
source: {
node_id: IMG2IMG_RESIZE,
field: 'image',
},
destination: {
node_id: IMAGE_TO_LATENTS,
field: 'image',
},
},
{
source: {
node_id: DENOISE_LATENTS,
field: 'latents',
},
destination: {
node_id: CANVAS_OUTPUT,
node_id: LATENTS_TO_IMAGE,
field: 'latents',
},
},
],
};
// handle `fit`
if (initialImage.width !== width || initialImage.height !== height) {
// The init image needs to be resized to the specified width and height before being passed to `IMAGE_TO_LATENTS`
// Create a resize node, explicitly setting its image
const resizeNode: ImageResizeInvocation = {
id: RESIZE,
type: 'img_resize',
image: {
image_name: initialImage.image_name,
},
is_intermediate: true,
width,
height,
};
graph.nodes[RESIZE] = resizeNode;
// The `RESIZE` node then passes its image to `IMAGE_TO_LATENTS`
graph.edges.push({
source: { node_id: RESIZE, field: 'image' },
destination: {
node_id: IMAGE_TO_LATENTS,
field: 'image',
},
});
// The `RESIZE` node also passes its width and height to `NOISE`
graph.edges.push({
source: { node_id: RESIZE, field: 'width' },
destination: {
node_id: NOISE,
field: 'width',
},
});
graph.edges.push({
source: { node_id: RESIZE, field: 'height' },
destination: {
node_id: NOISE,
field: 'height',
},
});
{
source: {
node_id: LATENTS_TO_IMAGE,
field: 'image',
},
destination: {
node_id: CANVAS_OUTPUT,
field: 'image',
},
}
);
} else {
// We are not resizing, so we need to set the image on the `IMAGE_TO_LATENTS` node explicitly
(graph.nodes[IMAGE_TO_LATENTS] as ImageToLatentsInvocation).image = {
image_name: initialImage.image_name,
graph.nodes[CANVAS_OUTPUT] = {
type: 'l2i',
id: CANVAS_OUTPUT,
is_intermediate: !shouldAutoSave,
fp32: vaePrecision === 'fp32' ? true : false,
};
// Pass the image's dimensions to the `NOISE` node
(graph.nodes[IMAGE_TO_LATENTS] as ImageToLatentsInvocation).image =
initialImage;
graph.edges.push({
source: { node_id: IMAGE_TO_LATENTS, field: 'width' },
destination: {
node_id: NOISE,
field: 'width',
source: {
node_id: DENOISE_LATENTS,
field: 'latents',
},
});
graph.edges.push({
source: { node_id: IMAGE_TO_LATENTS, field: 'height' },
destination: {
node_id: NOISE,
field: 'height',
node_id: CANVAS_OUTPUT,
field: 'latents',
},
});
}
@ -300,8 +310,10 @@ export const buildCanvasImageToImageGraph = (
type: 'metadata_accumulator',
generation_mode: 'img2img',
cfg_scale,
height,
width,
width: !isUsingScaledDimensions ? width : scaledBoundingBoxDimensions.width,
height: !isUsingScaledDimensions
? height
: scaledBoundingBoxDimensions.height,
positive_prompt: '', // set in addDynamicPromptsToGraph
negative_prompt: negativePrompt,
model,

39
invokeai/frontend/web/src/features/nodes/util/graphBuilders/buildCanvasInpaintGraph.ts
View File

@ -2,6 +2,7 @@ import { logger } from 'app/logging/logger';
import { RootState } from 'app/store/store';
import { NonNullableGraph } from 'features/nodes/types/types';
import {
CreateDenoiseMaskInvocation,
ImageBlurInvocation,
ImageDTO,
ImageToLatentsInvocation,
@ -15,13 +16,14 @@ import { addNSFWCheckerToGraph } from './addNSFWCheckerToGraph';
import { addVAEToGraph } from './addVAEToGraph';
import { addWatermarkerToGraph } from './addWatermarkerToGraph';
import {
CANVAS_INPAINT_GRAPH,
CANVAS_OUTPUT,
CANVAS_COHERENCE_DENOISE_LATENTS,
CANVAS_COHERENCE_NOISE,
CANVAS_COHERENCE_NOISE_INCREMENT,
CANVAS_INPAINT_GRAPH,
CANVAS_OUTPUT,
CLIP_SKIP,
DENOISE_LATENTS,
INPAINT_CREATE_MASK,
INPAINT_IMAGE,
INPAINT_IMAGE_RESIZE_DOWN,
INPAINT_IMAGE_RESIZE_UP,
@ -127,6 +129,12 @@ export const buildCanvasInpaintGraph = (
is_intermediate: true,
fp32: vaePrecision === 'fp32' ? true : false,
},
[INPAINT_CREATE_MASK]: {
type: 'create_denoise_mask',
id: INPAINT_CREATE_MASK,
is_intermediate: true,
fp32: vaePrecision === 'fp32' ? true : false,
},
[NOISE]: {
type: 'noise',
id: NOISE,
@ -276,16 +284,27 @@ export const buildCanvasInpaintGraph = (
field: 'latents',
},
},
// Create Inpaint Mask
{
source: {
node_id: MASK_BLUR,
field: 'image',
},
destination: {
node_id: DENOISE_LATENTS,
node_id: INPAINT_CREATE_MASK,
field: 'mask',
},
},
{
source: {
node_id: INPAINT_CREATE_MASK,
field: 'denoise_mask',
},
destination: {
node_id: DENOISE_LATENTS,
field: 'denoise_mask',
},
},
// Iterate
{
source: {
@ -459,6 +478,16 @@ export const buildCanvasInpaintGraph = (
field: 'image',
},
},
{
source: {
node_id: INPAINT_IMAGE_RESIZE_UP,
field: 'image',
},
destination: {
node_id: INPAINT_CREATE_MASK,
field: 'image',
},
},
// Color Correct The Inpainted Result
{
source: {
@ -516,6 +545,10 @@ export const buildCanvasInpaintGraph = (
...(graph.nodes[MASK_BLUR] as ImageBlurInvocation),
image: canvasMaskImage,
};
graph.nodes[INPAINT_CREATE_MASK] = {
...(graph.nodes[INPAINT_CREATE_MASK] as CreateDenoiseMaskInvocation),
image: canvasInitImage,
};
graph.edges.push(
// Color Correct The Inpainted Result

44
invokeai/frontend/web/src/features/nodes/util/graphBuilders/buildCanvasOutpaintGraph.ts
View File

@ -17,13 +17,14 @@ import { addNSFWCheckerToGraph } from './addNSFWCheckerToGraph';
import { addVAEToGraph } from './addVAEToGraph';
import { addWatermarkerToGraph } from './addWatermarkerToGraph';
import {
CANVAS_OUTPAINT_GRAPH,
CANVAS_OUTPUT,
CANVAS_COHERENCE_DENOISE_LATENTS,
CANVAS_COHERENCE_NOISE,
CANVAS_COHERENCE_NOISE_INCREMENT,
CANVAS_OUTPAINT_GRAPH,
CANVAS_OUTPUT,
CLIP_SKIP,
DENOISE_LATENTS,
INPAINT_CREATE_MASK,
INPAINT_IMAGE,
INPAINT_IMAGE_RESIZE_DOWN,
INPAINT_IMAGE_RESIZE_UP,
@ -153,6 +154,12 @@ export const buildCanvasOutpaintGraph = (
use_cpu,
is_intermediate: true,
},
[INPAINT_CREATE_MASK]: {
type: 'create_denoise_mask',
id: INPAINT_CREATE_MASK,
is_intermediate: true,
fp32: vaePrecision === 'fp32' ? true : false,
},
[DENOISE_LATENTS]: {
type: 'denoise_latents',
id: DENOISE_LATENTS,
@ -317,16 +324,27 @@ export const buildCanvasOutpaintGraph = (
field: 'latents',
},
},
// Create Inpaint Mask
{
source: {
node_id: MASK_BLUR,
field: 'image',
},
destination: {
node_id: DENOISE_LATENTS,
node_id: INPAINT_CREATE_MASK,
field: 'mask',
},
},
{
source: {
node_id: INPAINT_CREATE_MASK,
field: 'denoise_mask',
},
destination: {
node_id: DENOISE_LATENTS,
field: 'denoise_mask',
},
},
// Iterate
{
source: {
@ -522,6 +540,16 @@ export const buildCanvasOutpaintGraph = (
field: 'image',
},
},
{
source: {
node_id: INPAINT_INFILL,
field: 'image',
},
destination: {
node_id: INPAINT_CREATE_MASK,
field: 'image',
},
},
// Take combined mask and resize and then blur
{
source: {
@ -640,6 +668,16 @@ export const buildCanvasOutpaintGraph = (
field: 'image',
},
},
{
source: {
node_id: INPAINT_INFILL,
field: 'image',
},
destination: {
node_id: INPAINT_CREATE_MASK,
field: 'image',
},
},
// Color Correct The Inpainted Result
{
source: {

165
invokeai/frontend/web/src/features/nodes/util/graphBuilders/buildCanvasSDXLImageToImageGraph.ts
View File

@ -2,11 +2,7 @@ import { logger } from 'app/logging/logger';
import { RootState } from 'app/store/store';
import { NonNullableGraph } from 'features/nodes/types/types';
import { initialGenerationState } from 'features/parameters/store/generationSlice';
import {
ImageDTO,
ImageResizeInvocation,
ImageToLatentsInvocation,
} from 'services/api/types';
import { ImageDTO, ImageToLatentsInvocation } from 'services/api/types';
import { addControlNetToLinearGraph } from './addControlNetToLinearGraph';
import { addDynamicPromptsToGraph } from './addDynamicPromptsToGraph';
import { addNSFWCheckerToGraph } from './addNSFWCheckerToGraph';
@ -17,11 +13,12 @@ import { addWatermarkerToGraph } from './addWatermarkerToGraph';
import {
CANVAS_OUTPUT,
IMAGE_TO_LATENTS,
IMG2IMG_RESIZE,
LATENTS_TO_IMAGE,
METADATA_ACCUMULATOR,
NEGATIVE_CONDITIONING,
NOISE,
POSITIVE_CONDITIONING,
RESIZE,
SDXL_CANVAS_IMAGE_TO_IMAGE_GRAPH,
SDXL_DENOISE_LATENTS,
SDXL_MODEL_LOADER,
@ -59,7 +56,15 @@ export const buildCanvasSDXLImageToImageGraph = (
// The bounding box determines width and height, not the width and height params
const { width, height } = state.canvas.boundingBoxDimensions;
const { shouldAutoSave } = state.canvas;
const {
scaledBoundingBoxDimensions,
boundingBoxScaleMethod,
shouldAutoSave,
} = state.canvas;
const isUsingScaledDimensions = ['auto', 'manual'].includes(
boundingBoxScaleMethod
);
if (!model) {
log.error('No model found in state');
@ -109,16 +114,18 @@ export const buildCanvasSDXLImageToImageGraph = (
id: NOISE,
is_intermediate: true,
use_cpu,
width: !isUsingScaledDimensions
? width
: scaledBoundingBoxDimensions.width,
height: !isUsingScaledDimensions
? height
: scaledBoundingBoxDimensions.height,
},
[IMAGE_TO_LATENTS]: {
type: 'i2l',
id: IMAGE_TO_LATENTS,
is_intermediate: true,
fp32: vaePrecision === 'fp32' ? true : false,
// must be set manually later, bc `fit` parameter may require a resize node inserted
// image: {
// image_name: initialImage.image_name,
// },
},
[SDXL_DENOISE_LATENTS]: {
type: 'denoise_latents',
@ -132,12 +139,6 @@ export const buildCanvasSDXLImageToImageGraph = (
: 1 - strength,
denoising_end: shouldUseSDXLRefiner ? refinerStart : 1,
},
[CANVAS_OUTPUT]: {
type: 'l2i',
id: CANVAS_OUTPUT,
is_intermediate: !shouldAutoSave,
fp32: vaePrecision === 'fp32' ? true : false,
},
},
edges: [
// Connect Model Loader To UNet & CLIP
@ -232,82 +233,84 @@ export const buildCanvasSDXLImageToImageGraph = (
field: 'latents',
},
},
// Decode denoised latents to an image
],
};
// Decode Latents To Image & Handle Scaled Before Processing
if (isUsingScaledDimensions) {
graph.nodes[IMG2IMG_RESIZE] = {
id: IMG2IMG_RESIZE,
type: 'img_resize',
is_intermediate: true,
image: initialImage,
width: scaledBoundingBoxDimensions.width,
height: scaledBoundingBoxDimensions.height,
};
graph.nodes[LATENTS_TO_IMAGE] = {
id: LATENTS_TO_IMAGE,
type: 'l2i',
is_intermediate: true,
fp32: vaePrecision === 'fp32' ? true : false,
};
graph.nodes[CANVAS_OUTPUT] = {
id: CANVAS_OUTPUT,
type: 'img_resize',
is_intermediate: !shouldAutoSave,
width: width,
height: height,
};
graph.edges.push(
{
source: {
node_id: IMG2IMG_RESIZE,
field: 'image',
},
destination: {
node_id: IMAGE_TO_LATENTS,
field: 'image',
},
},
{
source: {
node_id: SDXL_DENOISE_LATENTS,
field: 'latents',
},
destination: {
node_id: CANVAS_OUTPUT,
node_id: LATENTS_TO_IMAGE,
field: 'latents',
},
},
],
};
// handle `fit`
if (initialImage.width !== width || initialImage.height !== height) {
// The init image needs to be resized to the specified width and height before being passed to `IMAGE_TO_LATENTS`
// Create a resize node, explicitly setting its image
const resizeNode: ImageResizeInvocation = {
id: RESIZE,
type: 'img_resize',
image: {
image_name: initialImage.image_name,
},
is_intermediate: true,
width,
height,
};
graph.nodes[RESIZE] = resizeNode;
// The `RESIZE` node then passes its image to `IMAGE_TO_LATENTS`
graph.edges.push({
source: { node_id: RESIZE, field: 'image' },
destination: {
node_id: IMAGE_TO_LATENTS,
field: 'image',
},
});
// The `RESIZE` node also passes its width and height to `NOISE`
graph.edges.push({
source: { node_id: RESIZE, field: 'width' },
destination: {
node_id: NOISE,
field: 'width',
},
});
graph.edges.push({
source: { node_id: RESIZE, field: 'height' },
destination: {
node_id: NOISE,
field: 'height',
},
});
{
source: {
node_id: LATENTS_TO_IMAGE,
field: 'image',
},
destination: {
node_id: CANVAS_OUTPUT,
field: 'image',
},
}
);
} else {
// We are not resizing, so we need to set the image on the `IMAGE_TO_LATENTS` node explicitly
(graph.nodes[IMAGE_TO_LATENTS] as ImageToLatentsInvocation).image = {
image_name: initialImage.image_name,
graph.nodes[CANVAS_OUTPUT] = {
type: 'l2i',
id: CANVAS_OUTPUT,
is_intermediate: !shouldAutoSave,
fp32: vaePrecision === 'fp32' ? true : false,
};
// Pass the image's dimensions to the `NOISE` node
(graph.nodes[IMAGE_TO_LATENTS] as ImageToLatentsInvocation).image =
initialImage;
graph.edges.push({
source: { node_id: IMAGE_TO_LATENTS, field: 'width' },
destination: {
node_id: NOISE,
field: 'width',
source: {
node_id: SDXL_DENOISE_LATENTS,
field: 'latents',
},
});
graph.edges.push({
source: { node_id: IMAGE_TO_LATENTS, field: 'height' },
destination: {
node_id: NOISE,
field: 'height',
node_id: CANVAS_OUTPUT,
field: 'latents',
},
});
}
@ -318,8 +321,10 @@ export const buildCanvasSDXLImageToImageGraph = (
type: 'metadata_accumulator',
generation_mode: 'img2img',
cfg_scale,
height,
width,
width: !isUsingScaledDimensions ? width : scaledBoundingBoxDimensions.width,
height: !isUsingScaledDimensions
? height
: scaledBoundingBoxDimensions.height,
positive_prompt: '', // set in addDynamicPromptsToGraph
negative_prompt: negativePrompt,
model,

37
invokeai/frontend/web/src/features/nodes/util/graphBuilders/buildCanvasSDXLInpaintGraph.ts
View File

@ -2,6 +2,7 @@ import { logger } from 'app/logging/logger';
import { RootState } from 'app/store/store';
import { NonNullableGraph } from 'features/nodes/types/types';
import {
CreateDenoiseMaskInvocation,
ImageBlurInvocation,
ImageDTO,
ImageToLatentsInvocation,
@ -16,10 +17,11 @@ import { addSDXLRefinerToGraph } from './addSDXLRefinerToGraph';
import { addVAEToGraph } from './addVAEToGraph';
import { addWatermarkerToGraph } from './addWatermarkerToGraph';
import {
CANVAS_OUTPUT,
CANVAS_COHERENCE_DENOISE_LATENTS,
CANVAS_COHERENCE_NOISE,
CANVAS_COHERENCE_NOISE_INCREMENT,
CANVAS_OUTPUT,
INPAINT_CREATE_MASK,
INPAINT_IMAGE,
INPAINT_IMAGE_RESIZE_DOWN,
INPAINT_IMAGE_RESIZE_UP,
@ -136,6 +138,12 @@ export const buildCanvasSDXLInpaintGraph = (
use_cpu,
is_intermediate: true,
},
[INPAINT_CREATE_MASK]: {
type: 'create_denoise_mask',
id: INPAINT_CREATE_MASK,
is_intermediate: true,
fp32: vaePrecision === 'fp32' ? true : false,
},
[SDXL_DENOISE_LATENTS]: {
type: 'denoise_latents',
id: SDXL_DENOISE_LATENTS,
@ -290,16 +298,27 @@ export const buildCanvasSDXLInpaintGraph = (
field: 'latents',
},
},
// Create Inpaint Mask
{
source: {
node_id: MASK_BLUR,
field: 'image',
},
destination: {
node_id: SDXL_DENOISE_LATENTS,
node_id: INPAINT_CREATE_MASK,
field: 'mask',
},
},
{
source: {
node_id: INPAINT_CREATE_MASK,
field: 'denoise_mask',
},
destination: {
node_id: SDXL_DENOISE_LATENTS,
field: 'denoise_mask',
},
},
// Iterate
{
source: {
@ -473,6 +492,16 @@ export const buildCanvasSDXLInpaintGraph = (
field: 'image',
},
},
{
source: {
node_id: INPAINT_IMAGE_RESIZE_UP,
field: 'image',
},
destination: {
node_id: INPAINT_CREATE_MASK,
field: 'image',
},
},
// Color Correct The Inpainted Result
{
source: {
@ -530,6 +559,10 @@ export const buildCanvasSDXLInpaintGraph = (
...(graph.nodes[MASK_BLUR] as ImageBlurInvocation),
image: canvasMaskImage,
};
graph.nodes[INPAINT_CREATE_MASK] = {
...(graph.nodes[INPAINT_CREATE_MASK] as CreateDenoiseMaskInvocation),
image: canvasInitImage,
};
graph.edges.push(
// Color Correct The Inpainted Result

42
invokeai/frontend/web/src/features/nodes/util/graphBuilders/buildCanvasSDXLOutpaintGraph.ts
View File

@ -18,10 +18,11 @@ import { addSDXLRefinerToGraph } from './addSDXLRefinerToGraph';
import { addVAEToGraph } from './addVAEToGraph';
import { addWatermarkerToGraph } from './addWatermarkerToGraph';
import {
CANVAS_OUTPUT,
CANVAS_COHERENCE_DENOISE_LATENTS,
CANVAS_COHERENCE_NOISE,
CANVAS_COHERENCE_NOISE_INCREMENT,
CANVAS_OUTPUT,
INPAINT_CREATE_MASK,
INPAINT_IMAGE,
INPAINT_IMAGE_RESIZE_DOWN,
INPAINT_IMAGE_RESIZE_UP,
@ -156,6 +157,12 @@ export const buildCanvasSDXLOutpaintGraph = (
use_cpu,
is_intermediate: true,
},
[INPAINT_CREATE_MASK]: {
type: 'create_denoise_mask',
id: INPAINT_CREATE_MASK,
is_intermediate: true,
fp32: vaePrecision === 'fp32' ? true : false,
},
[SDXL_DENOISE_LATENTS]: {
type: 'denoise_latents',
id: SDXL_DENOISE_LATENTS,
@ -331,16 +338,27 @@ export const buildCanvasSDXLOutpaintGraph = (
field: 'latents',
},
},
// Create Inpaint Mask
{
source: {
node_id: MASK_BLUR,
field: 'image',
},
destination: {
node_id: SDXL_DENOISE_LATENTS,
node_id: INPAINT_CREATE_MASK,
field: 'mask',
},
},
{
source: {
node_id: INPAINT_CREATE_MASK,
field: 'denoise_mask',
},
destination: {
node_id: SDXL_DENOISE_LATENTS,
field: 'denoise_mask',
},
},
// Iterate
{
source: {
@ -537,6 +555,16 @@ export const buildCanvasSDXLOutpaintGraph = (
field: 'image',
},
},
{
source: {
node_id: INPAINT_INFILL,
field: 'image',
},
destination: {
node_id: INPAINT_CREATE_MASK,
field: 'image',
},
},
// Take combined mask and resize and then blur
{
source: {
@ -655,6 +683,16 @@ export const buildCanvasSDXLOutpaintGraph = (
field: 'image',
},
},
{
source: {
node_id: INPAINT_INFILL,
field: 'image',
},
destination: {
node_id: INPAINT_CREATE_MASK,
field: 'image',
},
},
// Color Correct The Inpainted Result
{
source: {

87
invokeai/frontend/web/src/features/nodes/util/graphBuilders/buildCanvasSDXLTextToImageGraph.ts
View File

@ -15,6 +15,7 @@ import { addVAEToGraph } from './addVAEToGraph';
import { addWatermarkerToGraph } from './addWatermarkerToGraph';
import {
CANVAS_OUTPUT,
LATENTS_TO_IMAGE,
METADATA_ACCUMULATOR,
NEGATIVE_CONDITIONING,
NOISE,
@ -49,7 +50,15 @@ export const buildCanvasSDXLTextToImageGraph = (
// The bounding box determines width and height, not the width and height params
const { width, height } = state.canvas.boundingBoxDimensions;
const { shouldAutoSave } = state.canvas;
const {
scaledBoundingBoxDimensions,
boundingBoxScaleMethod,
shouldAutoSave,
} = state.canvas;
const isUsingScaledDimensions = ['auto', 'manual'].includes(
boundingBoxScaleMethod
);
const { shouldUseSDXLRefiner, refinerStart, shouldConcatSDXLStylePrompt } =
state.sdxl;
@ -136,17 +145,15 @@ export const buildCanvasSDXLTextToImageGraph = (
type: 'noise',
id: NOISE,
is_intermediate: true,
width,
height,
width: !isUsingScaledDimensions
? width
: scaledBoundingBoxDimensions.width,
height: !isUsingScaledDimensions
? height
: scaledBoundingBoxDimensions.height,
use_cpu,
},
[t2lNode.id]: t2lNode,
[CANVAS_OUTPUT]: {
type: isUsingOnnxModel ? 'l2i_onnx' : 'l2i',
id: CANVAS_OUTPUT,
is_intermediate: !shouldAutoSave,
fp32: vaePrecision === 'fp32' ? true : false,
},
},
edges: [
// Connect Model Loader to UNet and CLIP
@ -231,19 +238,67 @@ export const buildCanvasSDXLTextToImageGraph = (
field: 'noise',
},
},
// Decode Denoised Latents To Image
],
};
// Decode Latents To Image & Handle Scaled Before Processing
if (isUsingScaledDimensions) {
graph.nodes[LATENTS_TO_IMAGE] = {
id: LATENTS_TO_IMAGE,
type: isUsingOnnxModel ? 'l2i_onnx' : 'l2i',
is_intermediate: true,
fp32: vaePrecision === 'fp32' ? true : false,
};
graph.nodes[CANVAS_OUTPUT] = {
id: CANVAS_OUTPUT,
type: 'img_resize',
is_intermediate: !shouldAutoSave,
width: width,
height: height,
};
graph.edges.push(
{
source: {
node_id: SDXL_DENOISE_LATENTS,
field: 'latents',
},
destination: {
node_id: CANVAS_OUTPUT,
node_id: LATENTS_TO_IMAGE,
field: 'latents',
},
},
],
};
{
source: {
node_id: LATENTS_TO_IMAGE,
field: 'image',
},
destination: {
node_id: CANVAS_OUTPUT,
field: 'image',
},
}
);
} else {
graph.nodes[CANVAS_OUTPUT] = {
type: isUsingOnnxModel ? 'l2i_onnx' : 'l2i',
id: CANVAS_OUTPUT,
is_intermediate: !shouldAutoSave,
fp32: vaePrecision === 'fp32' ? true : false,
};
graph.edges.push({
source: {
node_id: SDXL_DENOISE_LATENTS,
field: 'latents',
},
destination: {
node_id: CANVAS_OUTPUT,
field: 'latents',
},
});
}
// add metadata accumulator, which is only mostly populated - some fields are added later
graph.nodes[METADATA_ACCUMULATOR] = {
@ -251,8 +306,10 @@ export const buildCanvasSDXLTextToImageGraph = (
type: 'metadata_accumulator',
generation_mode: 'txt2img',
cfg_scale,
height,
width,
width: !isUsingScaledDimensions ? width : scaledBoundingBoxDimensions.width,
height: !isUsingScaledDimensions
? height
: scaledBoundingBoxDimensions.height,
positive_prompt: '', // set in addDynamicPromptsToGraph
negative_prompt: negativePrompt,
model,

87
invokeai/frontend/web/src/features/nodes/util/graphBuilders/buildCanvasTextToImageGraph.ts
View File

@ -17,6 +17,7 @@ import {
CANVAS_TEXT_TO_IMAGE_GRAPH,
CLIP_SKIP,
DENOISE_LATENTS,
LATENTS_TO_IMAGE,
MAIN_MODEL_LOADER,
METADATA_ACCUMULATOR,
NEGATIVE_CONDITIONING,
@ -39,6 +40,7 @@ export const buildCanvasTextToImageGraph = (
cfgScale: cfg_scale,
scheduler,
steps,
vaePrecision,
clipSkip,
shouldUseCpuNoise,
shouldUseNoiseSettings,
@ -47,7 +49,15 @@ export const buildCanvasTextToImageGraph = (
// The bounding box determines width and height, not the width and height params
const { width, height } = state.canvas.boundingBoxDimensions;
const { shouldAutoSave } = state.canvas;
const {
scaledBoundingBoxDimensions,
boundingBoxScaleMethod,
shouldAutoSave,
} = state.canvas;
const isUsingScaledDimensions = ['auto', 'manual'].includes(
boundingBoxScaleMethod
);
if (!model) {
log.error('No model found in state');
@ -131,16 +141,15 @@ export const buildCanvasTextToImageGraph = (
type: 'noise',
id: NOISE,
is_intermediate: true,
width,
height,
width: !isUsingScaledDimensions
? width
: scaledBoundingBoxDimensions.width,
height: !isUsingScaledDimensions
? height
: scaledBoundingBoxDimensions.height,
use_cpu,
},
[t2lNode.id]: t2lNode,
[CANVAS_OUTPUT]: {
type: isUsingOnnxModel ? 'l2i_onnx' : 'l2i',
id: CANVAS_OUTPUT,
is_intermediate: !shouldAutoSave,
},
},
edges: [
// Connect Model Loader to UNet & CLIP Skip
@ -216,19 +225,67 @@ export const buildCanvasTextToImageGraph = (
field: 'noise',
},
},
// Decode denoised latents to image
],
};
// Decode Latents To Image & Handle Scaled Before Processing
if (isUsingScaledDimensions) {
graph.nodes[LATENTS_TO_IMAGE] = {
id: LATENTS_TO_IMAGE,
type: isUsingOnnxModel ? 'l2i_onnx' : 'l2i',
is_intermediate: true,
fp32: vaePrecision === 'fp32' ? true : false,
};
graph.nodes[CANVAS_OUTPUT] = {
id: CANVAS_OUTPUT,
type: 'img_resize',
is_intermediate: !shouldAutoSave,
width: width,
height: height,
};
graph.edges.push(
{
source: {
node_id: DENOISE_LATENTS,
field: 'latents',
},
destination: {
node_id: CANVAS_OUTPUT,
node_id: LATENTS_TO_IMAGE,
field: 'latents',
},
},
],
};
{
source: {
node_id: LATENTS_TO_IMAGE,
field: 'image',
},
destination: {
node_id: CANVAS_OUTPUT,
field: 'image',
},
}
);
} else {
graph.nodes[CANVAS_OUTPUT] = {
type: isUsingOnnxModel ? 'l2i_onnx' : 'l2i',
id: CANVAS_OUTPUT,
is_intermediate: !shouldAutoSave,
fp32: vaePrecision === 'fp32' ? true : false,
};
graph.edges.push({
source: {
node_id: DENOISE_LATENTS,
field: 'latents',
},
destination: {
node_id: CANVAS_OUTPUT,
field: 'latents',
},
});
}
// add metadata accumulator, which is only mostly populated - some fields are added later
graph.nodes[METADATA_ACCUMULATOR] = {
@ -236,8 +293,10 @@ export const buildCanvasTextToImageGraph = (
type: 'metadata_accumulator',
generation_mode: 'txt2img',
cfg_scale,
height,
width,
width: !isUsingScaledDimensions ? width : scaledBoundingBoxDimensions.width,
height: !isUsingScaledDimensions
? height
: scaledBoundingBoxDimensions.height,
positive_prompt: '', // set in addDynamicPromptsToGraph
negative_prompt: negativePrompt,
model,

2
invokeai/frontend/web/src/features/nodes/util/graphBuilders/constants.ts
View File

@ -17,6 +17,7 @@ export const CLIP_SKIP = 'clip_skip';
export const IMAGE_TO_LATENTS = 'image_to_latents';
export const LATENTS_TO_LATENTS = 'latents_to_latents';
export const RESIZE = 'resize_image';
export const IMG2IMG_RESIZE = 'img2img_resize';
export const CANVAS_OUTPUT = 'canvas_output';
export const INPAINT_IMAGE = 'inpaint_image';
export const SCALED_INPAINT_IMAGE = 'scaled_inpaint_image';
@ -25,6 +26,7 @@ export const INPAINT_IMAGE_RESIZE_DOWN = 'inpaint_image_resize_down';
export const INPAINT_INFILL = 'inpaint_infill';
export const INPAINT_INFILL_RESIZE_DOWN = 'inpaint_infill_resize_down';
export const INPAINT_FINAL_IMAGE = 'inpaint_final_image';
export const INPAINT_CREATE_MASK = 'inpaint_create_mask';
export const CANVAS_COHERENCE_DENOISE_LATENTS =
'canvas_coherence_denoise_latents';
export const CANVAS_COHERENCE_NOISE = 'canvas_coherence_noise';

120
invokeai/frontend/web/src/services/api/schema.d.ts vendored
View File

File diff suppressed because one or more lines are too long

1
invokeai/frontend/web/src/services/api/types.ts
View File

@ -111,6 +111,7 @@ export type ImageBlurInvocation = s['ImageBlurInvocation'];
export type ImageScaleInvocation = s['ImageScaleInvocation'];
export type InfillPatchMatchInvocation = s['InfillPatchMatchInvocation'];
export type InfillTileInvocation = s['InfillTileInvocation'];
export type CreateDenoiseMaskInvocation = s['CreateDenoiseMaskInvocation'];
export type RandomIntInvocation = s['RandomIntInvocation'];
export type CompelInvocation = s['CompelInvocation'];
export type DynamicPromptInvocation = s['DynamicPromptInvocation'];