github/InvokeAI
1
1
Fork 0
mirror of https://github.com/invoke-ai/InvokeAI.git synced 2026-04-23 03:00:31 -04:00
Code Issues Packages Projects Releases Wiki Activity

feat: Make SDXL work across the board + Custom VAE Support

Browse Source 
Also a major cleanup pass to the SDXL graphs to ensure there's no ID overlap
This commit is contained in:
blessedcoolant
2023-08-13 01:45:03 +12:00
parent 55d27f71a3
commit 500cd552bc
9 changed files with 132 additions and 139 deletions
Download Patch File Download Diff File Expand all files Collapse all files

46
invokeai/frontend/web/src/features/nodes/util/graphBuilders/buildLinearSDXLImageToImageGraph.ts
View File

@@ -10,9 +10,9 @@ import { addDynamicPromptsToGraph } from './addDynamicPromptsToGraph';
import { addNSFWCheckerToGraph } from './addNSFWCheckerToGraph';
import { addSDXLLoRAsToGraph } from './addSDXLLoRAstoGraph';
import { addSDXLRefinerToGraph } from './addSDXLRefinerToGraph';
import { addVAEToGraph } from './addVAEToGraph';
import { addWatermarkerToGraph } from './addWatermarkerToGraph';
import {
DENOISE_LATENTS,
IMAGE_TO_LATENTS,
LATENTS_TO_IMAGE,
METADATA_ACCUMULATOR,
@@ -20,6 +20,7 @@ import {
NOISE,
POSITIVE_CONDITIONING,
RESIZE,
SDXL_DENOISE_LATENTS,
SDXL_IMAGE_TO_IMAGE_GRAPH,
SDXL_MODEL_LOADER,
} from './constants';
@@ -115,9 +116,9 @@ export const buildLinearSDXLImageToImageGraph = (
id: LATENTS_TO_IMAGE,
fp32: vaePrecision === 'fp32' ? true : false,
},
[DENOISE_LATENTS]: {
[SDXL_DENOISE_LATENTS]: {
type: 'denoise_latents',
id: DENOISE_LATENTS,
id: SDXL_DENOISE_LATENTS,
cfg_scale,
scheduler,
steps,
@@ -144,30 +145,10 @@ export const buildLinearSDXLImageToImageGraph = (
field: 'unet',
},
destination: {
node_id: DENOISE_LATENTS,
node_id: SDXL_DENOISE_LATENTS,
field: 'unet',
},
},
{
source: {
node_id: SDXL_MODEL_LOADER,
field: 'vae',
},
destination: {
node_id: LATENTS_TO_IMAGE,
field: 'vae',
},
},
{
source: {
node_id: SDXL_MODEL_LOADER,
field: 'vae',
},
destination: {
node_id: IMAGE_TO_LATENTS,
field: 'vae',
},
},
{
source: {
node_id: SDXL_MODEL_LOADER,
@@ -215,7 +196,7 @@ export const buildLinearSDXLImageToImageGraph = (
field: 'conditioning',
},
destination: {
node_id: DENOISE_LATENTS,
node_id: SDXL_DENOISE_LATENTS,
field: 'positive_conditioning',
},
},
@@ -225,7 +206,7 @@ export const buildLinearSDXLImageToImageGraph = (
field: 'conditioning',
},
destination: {
node_id: DENOISE_LATENTS,
node_id: SDXL_DENOISE_LATENTS,
field: 'negative_conditioning',
},
},
@@ -235,7 +216,7 @@ export const buildLinearSDXLImageToImageGraph = (
field: 'noise',
},
destination: {
node_id: DENOISE_LATENTS,
node_id: SDXL_DENOISE_LATENTS,
field: 'noise',
},
},
@@ -245,14 +226,14 @@ export const buildLinearSDXLImageToImageGraph = (
field: 'latents',
},
destination: {
node_id: DENOISE_LATENTS,
node_id: SDXL_DENOISE_LATENTS,
field: 'latents',
},
},
// Decode Denoised Latents To Image
{
source: {
node_id: DENOISE_LATENTS,
node_id: SDXL_DENOISE_LATENTS,
field: 'latents',
},
destination: {
@@ -368,13 +349,16 @@ export const buildLinearSDXLImageToImageGraph = (
},
});
addSDXLLoRAsToGraph(state, graph, DENOISE_LATENTS, SDXL_MODEL_LOADER);
addSDXLLoRAsToGraph(state, graph, SDXL_DENOISE_LATENTS, SDXL_MODEL_LOADER);
// Add Refiner if enabled
if (shouldUseSDXLRefiner) {
addSDXLRefinerToGraph(state, graph, DENOISE_LATENTS);
addSDXLRefinerToGraph(state, graph, SDXL_DENOISE_LATENTS);
}
// optionally add custom VAE
addVAEToGraph(state, graph, SDXL_MODEL_LOADER);
// add dynamic prompts - also sets up core iteration and seed
addDynamicPromptsToGraph(state, graph);