ceos/unsloth
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Images: LoRA free-text Hugging Face entry + recipe round-trip

Browse source 
The backend has always accepted a bare Hugging Face repo id (owner/name, or
owner/name:weight-file.safetensors) as a LoRA, downloading and applying it. But the
picker only rendered when the curated catalog had entries, and the catalog is empty,
so there was no UI path to apply any LoRA. Show the LoRA section whenever the loaded
model supports LoRA, and replace the curated-only dropdown with a text input: type a
Hub repo id, or pick a discovered adapter from a datalist of suggestions when the
catalog is populated.

Also restore LoRAs when loading a recipe. restoreSettings now parses the recipe's
"id:weight" strings (splitting on the last colon, since the id itself may contain one
for a specific weight file) back into the selection, so replaying a saved image
reproduces its adapters. The generate payload trims hand-typed ids and drops empty /
zero-weight rows, and a model swap clears the selection (a LoRA is family-specific)
without discarding a free-text pick that is not in the curated list.
This commit is contained in:
Daniel Han 2026-07-01 14:05:16 +00:00
commit 3d187f1bab
1 changed files with 59 additions and 38 deletions
Download patch file Download diff file Expand all files Collapse all files

97
studio/frontend/src/features/images/images-page.tsx
View file

@ -984,8 +984,11 @@ export function ImagesPage({ active = true }: { active?: boolean }) {
galleryCache.quant = quant;
}, [images, hasMore, selectedId, quant]);
// Refresh the LoRA picker's options when the loaded model (family) changes, and drop any
// selected adapters the new model can't use so a stale/incompatible LoRA is never sent.
// Refresh the LoRA picker's suggestions when the loaded model (family) changes. A LoRA is
// trained for a specific base family, so a model swap invalidates the current selection --
// clear it (the user re-adds a suggestion or types a Hub repo id for the new family). We do
// NOT filter the selection against the discovered catalog: a valid pick can be a free-text
// Hugging Face repo id that is not in the (often empty) curated list.
const loraCapable = Boolean(status?.loaded && status?.supports_lora);
useEffect(() => {
if (!loraCapable) {
@ -993,13 +996,11 @@ export function ImagesPage({ active = true }: { active?: boolean }) {
setLoras([]);
return;
}
setLoras([]);
let cancelled = false;
listDiffusionLoras(status?.family ?? undefined)
.then((list) => {
if (cancelled) return;
setAvailableLoras(list);
const ids = new Set(list.map((l) => l.id));
setLoras((prev) => prev.filter((s) => ids.has(s.id)));
if (!cancelled) setAvailableLoras(list);
})
.catch(() => {
if (!cancelled) setAvailableLoras([]);
@ -1127,6 +1128,16 @@ export function ImagesPage({ active = true }: { active?: boolean }) {
// The batch shared one seed, so image batch_index>0 only reproduces by replaying the
// whole batch: restore the batch size too (older recipes without it default to 1).
setBatchSize(image.batch_size ?? 1);
// Restore the LoRA selection. The recipe stores each adapter as an "id:weight" string;
// the id itself may contain a colon (owner/name:weight-file.safetensors), so split on
// the LAST colon to recover the weight. A malformed entry falls back to weight 1.
setLoras(
(image.loras ?? []).map((s) => {
const idx = s.lastIndexOf(":");
const w = idx > 0 ? Number.parseFloat(s.slice(idx + 1)) : NaN;
return Number.isFinite(w) ? { id: s.slice(0, idx), weight: w } : { id: s, weight: 1 };
}),
);
const m = matchAspect(image.width, image.height);
setAspect(m.key);
setPortrait(m.portrait);
@ -1551,8 +1562,14 @@ export function ImagesPage({ active = true }: { active?: boolean }) {
strength: condStrength,
upscale: condUpscale,
reference_images: condRefImages,
// Drop zero-weight rows so the recipe records only adapters that actually applied.
loras: loras.length ? loras.filter((l) => l.weight > 0) : undefined,
// Drop empty (no id typed yet) and zero-weight rows, and trim hand-typed repo ids,
// so the recipe records only adapters that actually applied. Empty -> omit entirely.
loras: (() => {
const active = loras
.map((l) => ({ id: l.id.trim(), weight: l.weight }))
.filter((l) => l.id && l.weight > 0);
return active.length ? active : undefined;
})(),
// ControlNet: sent only when a model + control image are chosen; v1 conditions plain
// text-to-image only, so skip it for image-conditioned workflows.
controlnet:
@ -2019,43 +2036,46 @@ export function ImagesPage({ active = true }: { active?: boolean }) {
onChange={(e) => setPrompt(e.target.value)}
/>
</Field>
{/* LoRA adapters: shown only when the loaded model + quant can apply them and at
least one adapter is discoverable. Stack multiple, each with a 0-2 weight. The
backend owns how they apply (native prompt tags / diffusers set_adapters); the
{/* LoRA adapters: shown whenever the loaded model + quant can apply them. Type a
Hugging Face repo id (owner/name, or owner/name:weight-file.safetensors) or pick
a discovered adapter from the suggestions. Stack multiple, each with a 0-2 weight.
The backend owns how they apply (native prompt tags / diffusers set_adapters); the
UI only sends {id, weight}. */}
{loraCapable && availableLoras.length > 0 && (
{loraCapable && (
<Field
label="LoRAs"
hint="Style or character adapters applied on top of the model. Stack several; each has its own strength (1.0 = full effect, 0 disables)."
hint="Style or character adapters applied on top of the model. Enter a Hugging Face repo id (or pick a suggestion) and set the strength (1.0 = full effect, 0 disables). Stack several."
>
<div className="space-y-2">
{availableLoras.length > 0 && (
<datalist id="diffusion-lora-suggestions">
{availableLoras.map((a) => (
<option key={a.id} value={a.id}>
{a.display_name}
</option>
))}
</datalist>
)}
{loras.map((sel, i) => (
<div
key={i}
className="space-y-1.5 rounded-lg border border-border bg-muted/30 p-2"
>
<div className="flex items-center gap-2">
<Select
<Input
value={sel.id}
onValueChange={(v) =>
setLoras((prev) => prev.map((p, j) => (j === i ? { ...p, id: v } : p)))
list={availableLoras.length > 0 ? "diffusion-lora-suggestions" : undefined}
placeholder="owner/name or owner/name:file.safetensors"
spellCheck={false}
autoCapitalize="none"
autoCorrect="off"
className="h-8 flex-1 text-xs"
onChange={(e) =>
setLoras((prev) =>
prev.map((p, j) => (j === i ? { ...p, id: e.target.value } : p)),
)
}
>
<SelectTrigger className="h-8 flex-1 text-xs">
<SelectValue placeholder="Select a LoRA" />
</SelectTrigger>
<SelectContent>
{availableLoras.map((a) => (
<SelectItem
key={a.id}
value={a.id}
disabled={a.id !== sel.id && loras.some((l) => l.id === a.id)}
>
{a.display_name}
</SelectItem>
))}
</SelectContent>
</Select>
/>
<Button
type="button"
variant="ghost"
@ -2079,20 +2099,21 @@ export function ImagesPage({ active = true }: { active?: boolean }) {
/>
</div>
))}
{loras.length < Math.min(availableLoras.length, 8) && (
{loras.length < 8 && (
<Button
type="button"
variant="secondary"
size="sm"
className="w-full"
onClick={() => {
// Prefill with the first unused suggestion when a curated catalog exists,
// else an empty row the user fills with a Hugging Face repo id.
const taken = new Set(loras.map((l) => l.id));
const next = availableLoras.find((a) => !taken.has(a.id));
if (next)
setLoras((prev) => [
...prev,
{ id: next.id, weight: next.weight_default || 1 },
]);
setLoras((prev) => [
...prev,
next ? { id: next.id, weight: next.weight_default || 1 } : { id: "", weight: 1 },
]);
}}
>
<HugeiconsIcon icon={ImageAdd02Icon} className="size-3.5" />