fix(ui): missing vae precision in graph builders

invokeai/frontend/web/src/features/nodes/util/graph/generation/addImageToImage.ts

@@ -14,7 +14,8 @@ export const addImageToImage = async (
   originalSize: Dimensions,
   scaledSize: Dimensions,
   bbox: CanvasV2State['bbox'],
-  denoising_start: number
+  denoising_start: number,
+  fp32: boolean
 ): Promise<Invocation<'img_resize' | 'l2i'>> => {
   denoise.denoising_start = denoising_start;
 
@@ -28,7 +29,7 @@ export const addImageToImage = async (
       image: { image_name },
       ...scaledSize,
     });
-    const i2l = g.addNode({ id: 'i2l', type: 'i2l' });
+    const i2l = g.addNode({ id: 'i2l', type: 'i2l', fp32 });
     const resizeImageToOriginalSize = g.addNode({
       type: 'img_resize',
       id: getPrefixedId('initial_image_resize_out'),
@@ -43,8 +44,8 @@ export const addImageToImage = async (
     // This is the new output node
     return resizeImageToOriginalSize;
   } else {
-    // No need to resize, just denoise
-    const i2l = g.addNode({ id: 'i2l', type: 'i2l', image: { image_name } });
+    // No need to resize, just decode
+    const i2l = g.addNode({ id: 'i2l', type: 'i2l', image: { image_name }, fp32 });
     g.addEdge(vaeSource, 'vae', i2l, 'vae');
     g.addEdge(i2l, 'latents', denoise, 'latents');
     return l2i;

invokeai/frontend/web/src/features/nodes/util/graph/generation/addInpaint.ts

@@ -3,7 +3,6 @@ import type { CanvasManager } from 'features/controlLayers/konva/CanvasManager';
 import { getPrefixedId } from 'features/controlLayers/konva/util';
 import type { CanvasV2State, Dimensions } from 'features/controlLayers/store/types';
 import type { Graph } from 'features/nodes/util/graph/generation/Graph';
-import type { ParameterPrecision } from 'features/parameters/types/parameterSchemas';
 import { isEqual } from 'lodash-es';
 import type { Invocation } from 'services/api/types';
 
@@ -20,7 +19,7 @@ export const addInpaint = async (
   bbox: CanvasV2State['bbox'],
   compositing: CanvasV2State['compositing'],
   denoising_start: number,
-  vaePrecision: ParameterPrecision
+  fp32: boolean
 ): Promise<Invocation<'canvas_v2_mask_and_crop'>> => {
   denoise.denoising_start = denoising_start;
 
@@ -30,7 +29,7 @@ export const addInpaint = async (
 
   if (!isEqual(scaledSize, originalSize)) {
     // Scale before processing requires some resizing
-    const i2l = g.addNode({ id: getPrefixedId('i2l'), type: 'i2l' });
+    const i2l = g.addNode({ id: getPrefixedId('i2l'), type: 'i2l', fp32 });
     const resizeImageToScaledSize = g.addNode({
       type: 'img_resize',
       id: getPrefixedId('resize_image_to_scaled_size'),
@@ -64,7 +63,7 @@ export const addInpaint = async (
       coherence_mode: compositing.canvasCoherenceMode,
       minimum_denoise: compositing.canvasCoherenceMinDenoise,
       edge_radius: compositing.canvasCoherenceEdgeSize,
-      fp32: vaePrecision === 'fp32',
+      fp32,
     });
     const canvasPasteBack = g.addNode({
       id: getPrefixedId('canvas_v2_mask_and_crop'),
@@ -100,7 +99,12 @@ export const addInpaint = async (
     return canvasPasteBack;
   } else {
     // No scale before processing, much simpler
-    const i2l = g.addNode({ id: getPrefixedId('i2l'), type: 'i2l', image: { image_name: initialImage.image_name } });
+    const i2l = g.addNode({
+      id: getPrefixedId('i2l'),
+      type: 'i2l',
+      image: { image_name: initialImage.image_name },
+      fp32,
+    });
     const alphaToMask = g.addNode({
       id: getPrefixedId('alpha_to_mask'),
       type: 'tomask',
@@ -113,7 +117,7 @@ export const addInpaint = async (
       coherence_mode: compositing.canvasCoherenceMode,
       minimum_denoise: compositing.canvasCoherenceMinDenoise,
       edge_radius: compositing.canvasCoherenceEdgeSize,
-      fp32: vaePrecision === 'fp32',
+      fp32,
       image: { image_name: initialImage.image_name },
     });
     const canvasPasteBack = g.addNode({

invokeai/frontend/web/src/features/nodes/util/graph/generation/addOutpaint.ts

@@ -4,7 +4,6 @@ import { getPrefixedId } from 'features/controlLayers/konva/util';
 import type { CanvasV2State, Dimensions } from 'features/controlLayers/store/types';
 import type { Graph } from 'features/nodes/util/graph/generation/Graph';
 import { getInfill } from 'features/nodes/util/graph/graphBuilderUtils';
-import type { ParameterPrecision } from 'features/parameters/types/parameterSchemas';
 import { isEqual } from 'lodash-es';
 import type { Invocation } from 'services/api/types';
 
@@ -21,7 +20,7 @@ export const addOutpaint = async (
   bbox: CanvasV2State['bbox'],
   compositing: CanvasV2State['compositing'],
   denoising_start: number,
-  vaePrecision: ParameterPrecision
+  fp32: boolean
 ): Promise<Invocation<'canvas_v2_mask_and_crop'>> => {
   denoise.denoising_start = denoising_start;
 
@@ -76,7 +75,7 @@ export const addOutpaint = async (
       coherence_mode: compositing.canvasCoherenceMode,
       minimum_denoise: compositing.canvasCoherenceMinDenoise,
       edge_radius: compositing.canvasCoherenceEdgeSize,
-      fp32: vaePrecision === 'fp32',
+      fp32,
     });
     g.addEdge(infill, 'image', createGradientMask, 'image');
     g.addEdge(resizeInputMaskToScaledSize, 'image', createGradientMask, 'mask');
@@ -85,7 +84,7 @@ export const addOutpaint = async (
     g.addEdge(createGradientMask, 'denoise_mask', denoise, 'denoise_mask');
 
     // Decode infilled image and connect to denoise
-    const i2l = g.addNode({ id: getPrefixedId('i2l'), type: 'i2l' });
+    const i2l = g.addNode({ id: getPrefixedId('i2l'), type: 'i2l', fp32 });
     g.addEdge(infill, 'image', i2l, 'image');
     g.addEdge(vaeSource, 'vae', i2l, 'vae');
     g.addEdge(i2l, 'latents', denoise, 'latents');
@@ -125,7 +124,7 @@ export const addOutpaint = async (
   } else {
     infill.image = { image_name: initialImage.image_name };
     // No scale before processing, much simpler
-    const i2l = g.addNode({ id: getPrefixedId('i2l'), type: 'i2l' });
+    const i2l = g.addNode({ id: getPrefixedId('i2l'), type: 'i2l', fp32 });
     const maskAlphaToMask = g.addNode({
       id: getPrefixedId('mask_alpha_to_mask'),
       type: 'tomask',
@@ -147,7 +146,7 @@ export const addOutpaint = async (
       coherence_mode: compositing.canvasCoherenceMode,
       minimum_denoise: compositing.canvasCoherenceMinDenoise,
       edge_radius: compositing.canvasCoherenceEdgeSize,
-      fp32: vaePrecision === 'fp32',
+      fp32,
       image: { image_name: initialImage.image_name },
     });
     const canvasPasteBack = g.addNode({

invokeai/frontend/web/src/features/nodes/util/graph/generation/buildSD1Graph.ts

@@ -48,8 +48,8 @@ export const buildSD1Graph = async (
 
   assert(model, 'No model found in state');
 
+  const fp32 = vaePrecision === 'fp32';
   const { positivePrompt, negativePrompt } = getPresetModifiedPrompts(state);
-
   const { originalSize, scaledSize } = getSizes(bbox);
 
   const g = new Graph(getPrefixedId('sd1_graph'));
@@ -102,7 +102,7 @@ export const buildSD1Graph = async (
   const l2i = g.addNode({
     type: 'l2i',
     id: getPrefixedId('l2i'),
-    fp32: vaePrecision === 'fp32',
+    fp32,
   });
   const vaeLoader =
     vae?.base === model.base
@@ -168,7 +168,8 @@ export const buildSD1Graph = async (
       originalSize,
       scaledSize,
       bbox,
-      1 - params.img2imgStrength
+      1 - params.img2imgStrength,
+      vaePrecision === 'fp32'
     );
   } else if (generationMode === 'inpaint') {
     const { compositing } = state.canvasV2;
@@ -185,7 +186,7 @@ export const buildSD1Graph = async (
       bbox,
       compositing,
       1 - params.img2imgStrength,
-      vaePrecision
+      vaePrecision === 'fp32'
     );
   } else if (generationMode === 'outpaint') {
     const { compositing } = state.canvasV2;
@@ -202,7 +203,7 @@ export const buildSD1Graph = async (
       bbox,
       compositing,
       1 - params.img2imgStrength,
-      vaePrecision
+      fp32
     );
   }
 

invokeai/frontend/web/src/features/nodes/util/graph/generation/buildSDXLGraph.ts

@@ -49,8 +49,8 @@ export const buildSDXLGraph = async (
 
   assert(model, 'No model found in state');
 
+  const fp32 = vaePrecision === 'fp32';
   const { originalSize, scaledSize } = getSizes(bbox);
-
   const { positivePrompt, negativePrompt, positiveStylePrompt, negativeStylePrompt } = getPresetModifiedPrompts(state);
 
   const g = new Graph(getPrefixedId('sdxl_graph'));
@@ -100,7 +100,7 @@ export const buildSDXLGraph = async (
   const l2i = g.addNode({
     type: 'l2i',
     id: getPrefixedId('l2i'),
-    fp32: vaePrecision === 'fp32',
+    fp32,
   });
   const vaeLoader =
     vae?.base === model.base
@@ -171,7 +171,8 @@ export const buildSDXLGraph = async (
       originalSize,
       scaledSize,
       bbox,
-      refinerModel ? Math.min(refinerStart, 1 - params.img2imgStrength) : 1 - params.img2imgStrength
+      refinerModel ? Math.min(refinerStart, 1 - params.img2imgStrength) : 1 - params.img2imgStrength,
+      fp32
     );
   } else if (generationMode === 'inpaint') {
     const { compositing } = state.canvasV2;
@@ -188,7 +189,7 @@ export const buildSDXLGraph = async (
       bbox,
       compositing,
       refinerModel ? Math.min(refinerStart, 1 - params.img2imgStrength) : 1 - params.img2imgStrength,
-      vaePrecision
+      fp32
     );
   } else if (generationMode === 'outpaint') {
     const { compositing } = state.canvasV2;
@@ -205,7 +206,7 @@ export const buildSDXLGraph = async (
       bbox,
       compositing,
       refinerModel ? Math.min(refinerStart, 1 - params.img2imgStrength) : 1 - params.img2imgStrength,
-      vaePrecision
+      fp32
     );
   }