feat(studio): editable context length with Apply/Reset for GGUF settings (#4592)
* feat(studio): editable context length with Apply/Reset for GGUF model settings Previously the Context Length field was read-only and the backend hardcoded `-c 0`, ignoring custom values entirely. KV Cache Dtype also triggered an immediate model reload with no way to cancel. Backend: - llama_cpp.py: pass the actual n_ctx value to `-c` instead of always 0 - models/inference.py: relax max_seq_length to 0..1048576 (0 = model default) so GGUF models with large context windows are supported Frontend: - chat-runtime-store: add customContextLength and loadedKvCacheDtype state fields for dirty tracking - chat-settings-sheet: make Context Length an editable number input, stop KV Cache Dtype from auto-reloading, show Apply/Reset buttons when either setting has been changed - use-chat-model-runtime: send customContextLength as max_seq_length in the load request, reset after successful load * fix: preserve maxSeqLength for non-GGUF models in load request customContextLength ?? 0 sent max_seq_length=0 for non-GGUF models, breaking the finetuning/inference path that needs the slider value. Now uses a three-way branch: - customContextLength set: use it (user edited GGUF context) - GGUF without custom: 0 (model's native context) - Non-GGUF: maxSeqLength from the sampling slider * fix: keep max_seq_length default at 4096 for non-GGUF callers Only relax the bounds (ge=0 for GGUF's "model default" mode, le=1048576 for large context windows). The default stays at 4096 so API callers that omit max_seq_length still get a sane value for non-GGUF models. * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * fix(studio): rename trust remote code toggle and hide when no model selected - Rename "Trust remote code" to "Enable custom code" - Shorten subtitle to "Only enable if sure" - Hide the toggle when no model is loaded (already hidden for GGUFs) * fix: restore ge=128 for max_seq_length validation Keep the minimum at 128 so the API rejects nonsensical values. GGUF path now sends the model's native context length (from ggufContextLength) instead of 0 when the user has not customized it. The upper bound stays at 1048576 for large-context GGUF models. * feat(studio): replace Context Length input with slider Use a ParamSlider (512 to model's native context, step 512) instead of a small number input. Shows "Max" when at the model's native context length. Consistent with the other slider controls in the settings panel. * feat(studio): add editable number input alongside Context Length slider The slider and number input stay synced -- dragging the slider updates the number, typing a number moves the slider. The input also accepts values beyond the slider range for power users who need custom context lengths larger than the model default. * fix(studio): widen context length input and use 1024 step for slider Make the number input wider (100px) so large values like 262144 are fully visible. Change slider step from 512 to 1024 and min from 512 to 1024. * fix(studio): context length number input increments by 1024 * fix(studio): cap context length input at model's native max Adds max attribute and clamps typed/incremented values so the context length cannot exceed the GGUF model's reported context window. * fix(studio): point "What's new" link to changelog page Changed from /blog to /docs/new/changelog. * fix(studio): preserve custom context length after Apply, remove stale subtitle - After a reload with a custom context length, keep the user's value in the UI instead of snapping back to the model's native max. ggufContextLength always reports the model's native metadata value regardless of what -c was passed, so we need to preserve customContextLength when it differs from native. - Remove "Reload to apply." from KV Cache Dtype subtitle since the Apply/Reset buttons now handle this. * feat(studio): auto-enable Search and Code tools when model supports them Previously toolsEnabled and codeToolsEnabled stayed false after loading a model even if it reported supports_tools=true. Now both toggles are automatically enabled when the loaded model supports tool calling, matching the existing behavior for reasoning. * fix(studio): auto-enable tools in autoLoadSmallestModel path The suggestion cards trigger autoLoadSmallestModel which bypasses selectModel entirely. It was hardcoding toolsEnabled: false and codeToolsEnabled: false even when the model supports tool calling. Now both are set from the load response, matching the selectModel behavior. Also sets kvCacheDtype/loadedKvCacheDtype for dirty tracking consistency. * fix(studio): re-read tool flags after auto-loading model The runtime state was captured once at the start of the chat adapter's run(), before autoLoadSmallestModel() executes. After auto-load enables tools in the store, the request was still built with the stale snapshot that had toolsEnabled=false. Now re-reads the store after auto-load so the first message includes tools. * fix(studio): re-read entire runtime state after auto-load, not just tools The runtime snapshot (including params.checkpoint, model id, and all tool/reasoning flags) was captured once before auto-load. After autoLoadSmallestModel sets the checkpoint and enables tools, the request was still built with stale params (empty checkpoint, tools disabled). Now re-reads the full store state after auto-load so the first message has the correct model, tools, and reasoning flags. * feat(studio): add Hugging Face token field in Preferences Adds a password input under Configuration > Preferences for users to enter their HF token. The token is persisted in localStorage and passed to all model validate/load/download calls, replacing the previously hardcoded null. This enables downloading gated and private models. * fix(studio): use model native context for GGUF auto-load, show friendly errors The auto-load paths and selectModel for GGUF were sending max_seq_length=4096 which now actually limits the context window (since we fixed the backend to respect n_ctx). Changed to send 0 for GGUF, which means "use model's native context size". Also replaced generic "An internal error occurred" messages with user-friendly descriptions for known errors like context size exceeded and lost connections. LoadRequest validation changed to ge=0 to allow the GGUF "model default" signal. The frontend slider still enforces min=128 for non-GGUF models. * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * fix(studio): filter out FP8 models from model search results Hide models matching *-FP8-* or *FP8-Dynamic* from both the recommended list and HF search results. These models are not yet supported in the inference UI. --------- Co-authored-by: Daniel Han <danielhanchen@users.noreply.github.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
This commit is contained in:
Daniel Han
GitHub
parent
6d6008a1ef
commit
55d24d7c49
10 changed files with 208 additions and 47 deletions
|
|
@ -848,7 +848,7 @@ class LlamaCppBackend:
|
|||
"--port",
|
||||
str(self._port),
|
||||
"-c",
|
||||
"0", # 0 = use model's native context size
|
||||
str(n_ctx) if n_ctx > 0 else "0", # 0 = model's native context size
|
||||
"--parallel",
|
||||
"1", # Single-user studio, saves VRAM
|
||||
"--flash-attn",
|
||||
|
|
|
|||
|
|
@ -22,7 +22,10 @@ class LoadRequest(BaseModel):
|
|||
None, description = "HuggingFace token for gated models"
|
||||
)
|
||||
max_seq_length: int = Field(
|
||||
4096, ge = 128, le = 32768, description = "Maximum sequence length"
|
||||
0,
|
||||
ge = 0,
|
||||
le = 1048576,
|
||||
description = "Maximum sequence length (0 = model default for GGUF)",
|
||||
)
|
||||
load_in_4bit: bool = Field(True, description = "Load model in 4-bit quantization")
|
||||
is_lora: bool = Field(False, description = "Whether this is a LoRA adapter")
|
||||
|
|
|
|||
|
|
@ -19,6 +19,27 @@ import asyncio
|
|||
import threading
|
||||
|
||||
|
||||
import re as _re
|
||||
|
||||
|
||||
def _friendly_error(exc: Exception) -> str:
|
||||
"""Extract a user-friendly message from known llama-server errors."""
|
||||
msg = str(exc)
|
||||
m = _re.search(
|
||||
r"request \((\d+) tokens?\) exceeds the available context size \((\d+) tokens?\)",
|
||||
msg,
|
||||
)
|
||||
if m:
|
||||
return (
|
||||
f"Message too long: {m.group(1)} tokens exceeds the {m.group(2)}-token "
|
||||
f"context window. Try increasing the Context Length in Model settings, "
|
||||
f"or shorten the conversation."
|
||||
)
|
||||
if "Lost connection to llama-server" in msg:
|
||||
return "Lost connection to the model server. It may have crashed -- try reloading the model."
|
||||
return "An internal error occurred"
|
||||
|
||||
|
||||
# Add backend directory to path
|
||||
backend_path = Path(__file__).parent.parent.parent
|
||||
if str(backend_path) not in sys.path:
|
||||
|
|
@ -550,7 +571,7 @@ async def generate_stream(
|
|||
except Exception as e:
|
||||
backend.reset_generation_state()
|
||||
logger.error(f"Error during generation: {e}", exc_info = True)
|
||||
yield f"data: {json.dumps({'error': 'An internal error occurred'})}\n\n"
|
||||
yield f"data: {json.dumps({'error': _friendly_error(e)})}\n\n"
|
||||
|
||||
return StreamingResponse(
|
||||
stream(),
|
||||
|
|
@ -944,7 +965,7 @@ async def openai_chat_completions(
|
|||
logger.error(
|
||||
f"Error during audio input streaming: {e}", exc_info = True
|
||||
)
|
||||
yield f"data: {json.dumps({'error': {'message': 'An internal error occurred', 'type': 'server_error'}})}\n\n"
|
||||
yield f"data: {json.dumps({'error': {'message': _friendly_error(e), 'type': 'server_error'}})}\n\n"
|
||||
|
||||
return StreamingResponse(
|
||||
audio_input_stream(),
|
||||
|
|
@ -1176,7 +1197,7 @@ async def openai_chat_completions(
|
|||
logger.error(f"Error during GGUF tool streaming: {e}\n{tb}")
|
||||
error_chunk = {
|
||||
"error": {
|
||||
"message": "An internal error occurred",
|
||||
"message": _friendly_error(e),
|
||||
"type": "server_error",
|
||||
},
|
||||
}
|
||||
|
|
@ -1314,7 +1335,7 @@ async def openai_chat_completions(
|
|||
logger.error(f"Error during GGUF streaming: {e}", exc_info = True)
|
||||
error_chunk = {
|
||||
"error": {
|
||||
"message": "An internal error occurred",
|
||||
"message": _friendly_error(e),
|
||||
"type": "server_error",
|
||||
},
|
||||
}
|
||||
|
|
@ -1495,7 +1516,7 @@ async def openai_chat_completions(
|
|||
logger.error(f"Error during OpenAI streaming: {e}", exc_info = True)
|
||||
error_chunk = {
|
||||
"error": {
|
||||
"message": "An internal error occurred",
|
||||
"message": _friendly_error(e),
|
||||
"type": "server_error",
|
||||
},
|
||||
}
|
||||
|
|
|
|||
|
|
@ -454,7 +454,8 @@ export function HubModelPicker({
|
|||
const recommendedIds = useMemo(() => {
|
||||
const all = dedupe([...models.map((model) => model.id), value ?? ""])
|
||||
.filter((id) => !downloadedSet.has(id.toLowerCase()))
|
||||
.filter((id) => !chatOnly || isGgufRepo(id));
|
||||
.filter((id) => !chatOnly || isGgufRepo(id))
|
||||
.filter((id) => !/-FP8[-.]|FP8-Dynamic/i.test(id));
|
||||
// Sort: GGUFs first, then hub models
|
||||
const gguf: string[] = [];
|
||||
const hub: string[] = [];
|
||||
|
|
@ -498,7 +499,8 @@ export function HubModelPicker({
|
|||
return results
|
||||
.map((result) => result.id)
|
||||
.filter((id) => !recommendedSet.has(id))
|
||||
.filter((id) => !chatOnly || isGgufRepo(id));
|
||||
.filter((id) => !chatOnly || isGgufRepo(id))
|
||||
.filter((id) => !/-FP8[-.]|FP8-Dynamic/i.test(id));
|
||||
}, [recommendedSet, results, showHfSection, chatOnly]);
|
||||
|
||||
const metricsById = useMemo(
|
||||
|
|
|
|||
|
|
@ -253,6 +253,7 @@ function waitForModelReady(abortSignal?: AbortSignal): Promise<void> {
|
|||
* falls back to smallest cached safetensors model.
|
||||
*/
|
||||
async function autoLoadSmallestModel(): Promise<boolean> {
|
||||
const hfToken = useChatRuntimeStore.getState().hfToken || null;
|
||||
const toastId = toast("Loading a model…", {
|
||||
description: "Auto-selecting the smallest downloaded model.",
|
||||
duration: 5000,
|
||||
|
|
@ -278,8 +279,8 @@ async function autoLoadSmallestModel(): Promise<boolean> {
|
|||
const variant = downloaded[0];
|
||||
const loadResp = await loadModel({
|
||||
model_path: repo.repo_id,
|
||||
hf_token: null,
|
||||
max_seq_length: 4096,
|
||||
hf_token: hfToken,
|
||||
max_seq_length: 0,
|
||||
load_in_4bit: true,
|
||||
is_lora: false,
|
||||
gguf_variant: variant.quant,
|
||||
|
|
@ -308,8 +309,10 @@ async function autoLoadSmallestModel(): Promise<boolean> {
|
|||
supportsReasoning: loadResp.supports_reasoning ?? false,
|
||||
reasoningEnabled: loadResp.supports_reasoning ?? false,
|
||||
supportsTools: loadResp.supports_tools ?? false,
|
||||
toolsEnabled: false,
|
||||
codeToolsEnabled: false,
|
||||
toolsEnabled: loadResp.supports_tools ?? false,
|
||||
codeToolsEnabled: loadResp.supports_tools ?? false,
|
||||
kvCacheDtype: loadResp.cache_type_kv ?? null,
|
||||
loadedKvCacheDtype: loadResp.cache_type_kv ?? null,
|
||||
defaultChatTemplate: loadResp.chat_template ?? null,
|
||||
chatTemplateOverride: null,
|
||||
});
|
||||
|
|
@ -329,7 +332,7 @@ async function autoLoadSmallestModel(): Promise<boolean> {
|
|||
try {
|
||||
const sfLoadResp = await loadModel({
|
||||
model_path: repo.repo_id,
|
||||
hf_token: null,
|
||||
hf_token: hfToken,
|
||||
max_seq_length: 4096,
|
||||
load_in_4bit: true,
|
||||
is_lora: false,
|
||||
|
|
@ -366,8 +369,8 @@ async function autoLoadSmallestModel(): Promise<boolean> {
|
|||
try {
|
||||
const loadResp = await loadModel({
|
||||
model_path: "unsloth/Qwen3.5-4B-GGUF",
|
||||
hf_token: null,
|
||||
max_seq_length: 4096,
|
||||
hf_token: hfToken,
|
||||
max_seq_length: 0,
|
||||
load_in_4bit: true,
|
||||
is_lora: false,
|
||||
gguf_variant: "UD-Q4_K_XL",
|
||||
|
|
@ -391,7 +394,10 @@ async function autoLoadSmallestModel(): Promise<boolean> {
|
|||
supportsReasoning: loadResp.supports_reasoning ?? false,
|
||||
reasoningEnabled: loadResp.supports_reasoning ?? false,
|
||||
supportsTools: loadResp.supports_tools ?? false,
|
||||
toolsEnabled: false,
|
||||
toolsEnabled: loadResp.supports_tools ?? false,
|
||||
codeToolsEnabled: loadResp.supports_tools ?? false,
|
||||
kvCacheDtype: loadResp.cache_type_kv ?? null,
|
||||
loadedKvCacheDtype: loadResp.cache_type_kv ?? null,
|
||||
defaultChatTemplate: loadResp.chat_template ?? null,
|
||||
chatTemplateOverride: null,
|
||||
});
|
||||
|
|
@ -410,8 +416,7 @@ async function autoLoadSmallestModel(): Promise<boolean> {
|
|||
export function createOpenAIStreamAdapter(): ChatModelAdapter {
|
||||
return {
|
||||
async *run({ messages, abortSignal, unstable_threadId }) {
|
||||
const runtime = useChatRuntimeStore.getState();
|
||||
const { params } = runtime;
|
||||
let runtime = useChatRuntimeStore.getState();
|
||||
|
||||
// Wait for in-progress model load to finish before inferring
|
||||
if (runtime.modelLoading) {
|
||||
|
|
@ -430,6 +435,9 @@ export function createOpenAIStreamAdapter(): ChatModelAdapter {
|
|||
}
|
||||
}
|
||||
|
||||
// Re-read store after potential auto-load / model ready wait
|
||||
runtime = useChatRuntimeStore.getState();
|
||||
const { params } = runtime;
|
||||
const {
|
||||
supportsTools,
|
||||
toolsEnabled,
|
||||
|
|
|
|||
|
|
@ -279,6 +279,14 @@ export function ChatSettingsPanel({
|
|||
const ggufContextLength = useChatRuntimeStore((s) => s.ggufContextLength);
|
||||
const kvCacheDtype = useChatRuntimeStore((s) => s.kvCacheDtype);
|
||||
const setKvCacheDtype = useChatRuntimeStore((s) => s.setKvCacheDtype);
|
||||
const loadedKvCacheDtype = useChatRuntimeStore((s) => s.loadedKvCacheDtype);
|
||||
const customContextLength = useChatRuntimeStore((s) => s.customContextLength);
|
||||
const setCustomContextLength = useChatRuntimeStore((s) => s.setCustomContextLength);
|
||||
|
||||
const ctxDisplayValue = customContextLength ?? ggufContextLength ?? "";
|
||||
const kvDirty = kvCacheDtype !== loadedKvCacheDtype;
|
||||
const ctxDirty = customContextLength !== null;
|
||||
const modelSettingsDirty = kvDirty || ctxDirty;
|
||||
const [customPresets, setCustomPresets] = useState<Preset[]>(() =>
|
||||
loadSavedCustomPresets(),
|
||||
);
|
||||
|
|
@ -467,32 +475,53 @@ export function ChatSettingsPanel({
|
|||
<div className="flex flex-col gap-3 py-1">
|
||||
{isGguf && (
|
||||
<>
|
||||
<div className="flex items-center justify-between gap-3">
|
||||
<div className="min-w-0">
|
||||
<div className="text-xs font-medium">Context Length</div>
|
||||
<div className="text-[11px] text-muted-foreground">
|
||||
Reported by the loaded GGUF model.
|
||||
</div>
|
||||
<div className="space-y-2">
|
||||
<div className="flex items-center justify-between">
|
||||
<span className="text-xs font-medium">Context Length</span>
|
||||
<Input
|
||||
type="number"
|
||||
value={typeof ctxDisplayValue === "number" ? ctxDisplayValue : (ggufContextLength ?? "")}
|
||||
placeholder="..."
|
||||
min={128}
|
||||
max={ggufContextLength ?? undefined}
|
||||
step={1024}
|
||||
className="h-6 w-[100px] text-right text-xs tabular-nums"
|
||||
onChange={(e) => {
|
||||
const raw = e.target.value;
|
||||
if (raw === "") {
|
||||
setCustomContextLength(null);
|
||||
return;
|
||||
}
|
||||
const v = parseInt(raw, 10);
|
||||
if (!Number.isNaN(v) && v >= 0) {
|
||||
const maxCtx = ggufContextLength ?? Infinity;
|
||||
const clamped = Math.min(v, maxCtx);
|
||||
setCustomContextLength(clamped === (ggufContextLength ?? 0) ? null : clamped);
|
||||
}
|
||||
}}
|
||||
/>
|
||||
</div>
|
||||
<Input
|
||||
value={ggufContextLength ?? ""}
|
||||
placeholder="Loading..."
|
||||
disabled={true}
|
||||
className="h-7 w-[90px] text-xs"
|
||||
<Slider
|
||||
min={1024}
|
||||
max={ggufContextLength ?? 4096}
|
||||
step={1024}
|
||||
value={[Math.min(typeof ctxDisplayValue === "number" ? ctxDisplayValue : (ggufContextLength ?? 4096), ggufContextLength ?? 4096)]}
|
||||
onValueChange={([v]) => {
|
||||
setCustomContextLength(v === (ggufContextLength ?? 0) ? null : v);
|
||||
}}
|
||||
/>
|
||||
</div>
|
||||
<div className="flex items-center justify-between gap-3">
|
||||
<div className="min-w-0">
|
||||
<div className="text-xs font-medium">KV Cache Dtype</div>
|
||||
<div className="text-[11px] text-muted-foreground">
|
||||
Quantize KV cache to reduce VRAM. Reload to apply.
|
||||
Quantize KV cache to reduce VRAM.
|
||||
</div>
|
||||
</div>
|
||||
<Select
|
||||
value={kvCacheDtype ?? "f16"}
|
||||
onValueChange={(v) => {
|
||||
setKvCacheDtype(v === "f16" ? null : v);
|
||||
onReloadModel?.();
|
||||
}}
|
||||
>
|
||||
<SelectTrigger className="h-7 w-[90px] text-xs">
|
||||
|
|
@ -507,14 +536,35 @@ export function ChatSettingsPanel({
|
|||
</SelectContent>
|
||||
</Select>
|
||||
</div>
|
||||
{modelSettingsDirty && (
|
||||
<div className="flex flex-wrap gap-1.5 pt-1">
|
||||
<button
|
||||
type="button"
|
||||
onClick={() => onReloadModel?.()}
|
||||
className="rounded-md bg-primary px-2.5 py-1 text-[11px] font-medium text-primary-foreground transition-colors hover:bg-primary/90"
|
||||
>
|
||||
Apply
|
||||
</button>
|
||||
<button
|
||||
type="button"
|
||||
onClick={() => {
|
||||
setCustomContextLength(null);
|
||||
setKvCacheDtype(loadedKvCacheDtype);
|
||||
}}
|
||||
className="rounded-md border px-2.5 py-1 text-[11px] font-medium text-muted-foreground transition-colors hover:bg-accent"
|
||||
>
|
||||
Reset
|
||||
</button>
|
||||
</div>
|
||||
)}
|
||||
</>
|
||||
)}
|
||||
{!isGguf && (
|
||||
{!isGguf && params.checkpoint && (
|
||||
<div className="flex items-center justify-between gap-3">
|
||||
<div className="min-w-0">
|
||||
<div className="text-xs font-medium">Trust remote code</div>
|
||||
<div className="text-xs font-medium">Enable custom code</div>
|
||||
<div className="text-[11px] text-muted-foreground">
|
||||
Allow models with custom code (e.g. Nemotron). Only enable for repos you trust.
|
||||
Allow models with custom code (e.g. Nemotron). Only enable if sure.
|
||||
</div>
|
||||
</div>
|
||||
<Switch
|
||||
|
|
@ -632,6 +682,7 @@ export function ChatSettingsPanel({
|
|||
onCheckedChange={onAutoTitleChange}
|
||||
/>
|
||||
</div>
|
||||
<HfTokenField />
|
||||
</div>
|
||||
</CollapsibleSection>
|
||||
|
||||
|
|
@ -775,6 +826,29 @@ function AutoHealToolCallsToggle() {
|
|||
);
|
||||
}
|
||||
|
||||
function HfTokenField() {
|
||||
const hfToken = useChatRuntimeStore((s) => s.hfToken);
|
||||
const setHfToken = useChatRuntimeStore((s) => s.setHfToken);
|
||||
|
||||
return (
|
||||
<div className="flex flex-col gap-1.5">
|
||||
<div className="min-w-0">
|
||||
<div className="text-xs font-medium">Hugging Face Token</div>
|
||||
<div className="text-[11px] text-muted-foreground">
|
||||
For downloading gated or private models.
|
||||
</div>
|
||||
</div>
|
||||
<Input
|
||||
type="password"
|
||||
value={hfToken}
|
||||
placeholder="hf_..."
|
||||
className="h-7 text-xs font-mono"
|
||||
onChange={(e) => setHfToken(e.target.value)}
|
||||
/>
|
||||
</div>
|
||||
);
|
||||
}
|
||||
|
||||
function ChatTemplateSection({
|
||||
onReloadModel,
|
||||
}: {
|
||||
|
|
|
|||
|
|
@ -354,12 +354,13 @@ export function useChatModelRuntime() {
|
|||
useChatRuntimeStore.getState().params.checkpoint;
|
||||
const paramsBeforeLoad = useChatRuntimeStore.getState().params;
|
||||
const maxSeqLength = paramsBeforeLoad.maxSeqLength;
|
||||
const hfToken = useChatRuntimeStore.getState().hfToken || null;
|
||||
try {
|
||||
// Lightweight pre-flight validation: avoid unloading a working model
|
||||
// if the new identifier is clearly invalid (e.g. bad HF id / path).
|
||||
await validateModel({
|
||||
model_path: modelId,
|
||||
hf_token: null,
|
||||
hf_token: hfToken,
|
||||
max_seq_length: maxSeqLength,
|
||||
load_in_4bit: true,
|
||||
is_lora: isLora,
|
||||
|
|
@ -371,11 +372,16 @@ export function useChatModelRuntime() {
|
|||
previousWasUnloaded = true;
|
||||
}
|
||||
|
||||
const { chatTemplateOverride, kvCacheDtype } = useChatRuntimeStore.getState();
|
||||
const { chatTemplateOverride, kvCacheDtype, customContextLength, ggufContextLength } = useChatRuntimeStore.getState();
|
||||
// GGUF: use custom context length, or 0 = model's native context
|
||||
// Non-GGUF: use the Max Seq Length slider value
|
||||
const effectiveMaxSeqLength = customContextLength != null
|
||||
? customContextLength
|
||||
: ggufVariant != null ? (ggufContextLength ?? 0) : maxSeqLength;
|
||||
const loadResponse = await loadModel({
|
||||
model_path: modelId,
|
||||
hf_token: null,
|
||||
max_seq_length: maxSeqLength,
|
||||
hf_token: hfToken,
|
||||
max_seq_length: effectiveMaxSeqLength,
|
||||
load_in_4bit: true,
|
||||
is_lora: isLora,
|
||||
gguf_variant: ggufVariant ?? null,
|
||||
|
|
@ -403,15 +409,27 @@ export function useChatModelRuntime() {
|
|||
}
|
||||
}
|
||||
}
|
||||
const loadedKv = loadResponse.cache_type_kv ?? null;
|
||||
const nativeCtx = loadResponse.is_gguf
|
||||
? (loadResponse.context_length ?? 131072)
|
||||
: null;
|
||||
// Keep customContextLength if the user set one and it differs
|
||||
// from the model's native context; otherwise clear it so the
|
||||
// display shows the native value without a dirty marker.
|
||||
const keepCustomCtx = customContextLength != null
|
||||
&& customContextLength !== nativeCtx
|
||||
? customContextLength
|
||||
: null;
|
||||
useChatRuntimeStore.setState({
|
||||
ggufContextLength: loadResponse.is_gguf
|
||||
? (loadResponse.context_length ?? 131072)
|
||||
: null,
|
||||
ggufContextLength: nativeCtx,
|
||||
supportsReasoning: loadResponse.supports_reasoning ?? false,
|
||||
reasoningEnabled: reasoningDefault,
|
||||
supportsTools: loadResponse.supports_tools ?? false,
|
||||
toolsEnabled: false,
|
||||
kvCacheDtype: loadResponse.cache_type_kv ?? null,
|
||||
toolsEnabled: loadResponse.supports_tools ?? false,
|
||||
codeToolsEnabled: loadResponse.supports_tools ?? false,
|
||||
kvCacheDtype: loadedKv,
|
||||
loadedKvCacheDtype: loadedKv,
|
||||
customContextLength: keepCustomCtx,
|
||||
defaultChatTemplate: loadResponse.chat_template ?? null,
|
||||
chatTemplateOverride: null,
|
||||
});
|
||||
|
|
@ -432,7 +450,7 @@ export function useChatModelRuntime() {
|
|||
try {
|
||||
await loadModel({
|
||||
model_path: previousCheckpoint,
|
||||
hf_token: null,
|
||||
hf_token: hfToken,
|
||||
max_seq_length: maxSeqLength,
|
||||
load_in_4bit: true,
|
||||
is_lora: previousIsLora,
|
||||
|
|
|
|||
|
|
@ -337,7 +337,7 @@ export function SharedComposer({
|
|||
async function ensureModelLoaded(sel: CompareModelSelection): Promise<string> {
|
||||
const resp = await loadModel({
|
||||
model_path: sel.id,
|
||||
hf_token: null,
|
||||
hf_token: useChatRuntimeStore.getState().hfToken || null,
|
||||
max_seq_length: maxSeqLength,
|
||||
load_in_4bit: true,
|
||||
is_lora: sel.isLora,
|
||||
|
|
|
|||
|
|
@ -14,6 +14,7 @@ const AUTO_TITLE_KEY = "unsloth_chat_auto_title";
|
|||
const AUTO_HEAL_TOOL_CALLS_KEY = "unsloth_auto_heal_tool_calls";
|
||||
const MAX_TOOL_CALLS_KEY = "unsloth_max_tool_calls_per_message";
|
||||
const TOOL_CALL_TIMEOUT_KEY = "unsloth_tool_call_timeout";
|
||||
const HF_TOKEN_KEY = "unsloth_hf_token";
|
||||
const INFERENCE_PARAMS_KEY = "unsloth_chat_inference_params";
|
||||
let hasShownInferencePersistenceWarning = false;
|
||||
|
||||
|
|
@ -62,6 +63,24 @@ function saveInt(key: string, value: number): void {
|
|||
}
|
||||
}
|
||||
|
||||
function loadString(key: string, fallback: string): string {
|
||||
if (!canUseStorage()) return fallback;
|
||||
try {
|
||||
return localStorage.getItem(key) ?? fallback;
|
||||
} catch {
|
||||
return fallback;
|
||||
}
|
||||
}
|
||||
|
||||
function saveString(key: string, value: string): void {
|
||||
if (!canUseStorage()) return;
|
||||
try {
|
||||
localStorage.setItem(key, value);
|
||||
} catch {
|
||||
// ignore
|
||||
}
|
||||
}
|
||||
|
||||
function asFiniteNumber(value: unknown, fallback: number): number {
|
||||
return typeof value === "number" && Number.isFinite(value) ? value : fallback;
|
||||
}
|
||||
|
|
@ -127,6 +146,7 @@ type ChatRuntimeStore = {
|
|||
loras: ChatLoraSummary[];
|
||||
runningByThreadId: Record<string, boolean>;
|
||||
autoTitle: boolean;
|
||||
hfToken: string;
|
||||
modelsError: string | null;
|
||||
activeGgufVariant: string | null;
|
||||
ggufContextLength: number | null;
|
||||
|
|
@ -141,6 +161,8 @@ type ChatRuntimeStore = {
|
|||
maxToolCallsPerMessage: number;
|
||||
toolCallTimeout: number;
|
||||
kvCacheDtype: string | null;
|
||||
loadedKvCacheDtype: string | null;
|
||||
customContextLength: number | null;
|
||||
defaultChatTemplate: string | null;
|
||||
chatTemplateOverride: string | null;
|
||||
activeThreadId: string | null;
|
||||
|
|
@ -159,6 +181,7 @@ type ChatRuntimeStore = {
|
|||
setLoras: (loras: ChatLoraSummary[]) => void;
|
||||
setThreadRunning: (threadId: string, running: boolean) => void;
|
||||
setAutoTitle: (enabled: boolean) => void;
|
||||
setHfToken: (token: string) => void;
|
||||
setModelsError: (error: string | null) => void;
|
||||
setCheckpoint: (modelId: string, ggufVariant?: string | null) => void;
|
||||
setActiveThreadId: (threadId: string | null) => void;
|
||||
|
|
@ -172,6 +195,7 @@ type ChatRuntimeStore = {
|
|||
setMaxToolCallsPerMessage: (value: number) => void;
|
||||
setToolCallTimeout: (value: number) => void;
|
||||
setKvCacheDtype: (dtype: string | null) => void;
|
||||
setCustomContextLength: (v: number | null) => void;
|
||||
setChatTemplateOverride: (template: string | null) => void;
|
||||
setPendingAudio: (base64: string, name: string) => void;
|
||||
clearPendingAudio: () => void;
|
||||
|
|
@ -184,6 +208,7 @@ export const useChatRuntimeStore = create<ChatRuntimeStore>((set) => ({
|
|||
loras: [],
|
||||
runningByThreadId: {},
|
||||
autoTitle: loadBool(AUTO_TITLE_KEY, false),
|
||||
hfToken: loadString(HF_TOKEN_KEY, ""),
|
||||
modelsError: null,
|
||||
activeGgufVariant: null,
|
||||
ggufContextLength: null,
|
||||
|
|
@ -198,6 +223,8 @@ export const useChatRuntimeStore = create<ChatRuntimeStore>((set) => ({
|
|||
maxToolCallsPerMessage: loadInt(MAX_TOOL_CALLS_KEY, 10),
|
||||
toolCallTimeout: loadInt(TOOL_CALL_TIMEOUT_KEY, 5),
|
||||
kvCacheDtype: null,
|
||||
loadedKvCacheDtype: null,
|
||||
customContextLength: null,
|
||||
defaultChatTemplate: null,
|
||||
chatTemplateOverride: null,
|
||||
activeThreadId: null,
|
||||
|
|
@ -235,6 +262,11 @@ export const useChatRuntimeStore = create<ChatRuntimeStore>((set) => ({
|
|||
saveBool(AUTO_TITLE_KEY, autoTitle);
|
||||
return { autoTitle };
|
||||
}),
|
||||
setHfToken: (hfToken) =>
|
||||
set(() => {
|
||||
saveString(HF_TOKEN_KEY, hfToken);
|
||||
return { hfToken };
|
||||
}),
|
||||
setModelsError: (modelsError) => set({ modelsError }),
|
||||
setCheckpoint: (modelId, ggufVariant) =>
|
||||
set((state) => ({
|
||||
|
|
@ -261,6 +293,8 @@ export const useChatRuntimeStore = create<ChatRuntimeStore>((set) => ({
|
|||
codeToolsEnabled: false,
|
||||
toolStatus: null,
|
||||
kvCacheDtype: null,
|
||||
loadedKvCacheDtype: null,
|
||||
customContextLength: null,
|
||||
defaultChatTemplate: null,
|
||||
chatTemplateOverride: null,
|
||||
})),
|
||||
|
|
@ -285,6 +319,7 @@ export const useChatRuntimeStore = create<ChatRuntimeStore>((set) => ({
|
|||
return { toolCallTimeout };
|
||||
}),
|
||||
setKvCacheDtype: (kvCacheDtype) => set({ kvCacheDtype }),
|
||||
setCustomContextLength: (customContextLength) => set({ customContextLength }),
|
||||
setChatTemplateOverride: (chatTemplateOverride) => set({ chatTemplateOverride }),
|
||||
setPendingAudio: (base64, name) =>
|
||||
set({ pendingAudioBase64: base64, pendingAudioName: name }),
|
||||
|
|
|
|||
|
|
@ -172,7 +172,7 @@ export function ThreadSidebar({
|
|||
<span>Learn more in docs</span>
|
||||
</a>
|
||||
<a
|
||||
href="https://unsloth.ai/blog"
|
||||
href="https://unsloth.ai/docs/new/changelog"
|
||||
target="_blank"
|
||||
rel="noopener noreferrer"
|
||||
className="flex items-center gap-2 rounded-md px-2 py-1.5 text-xs text-muted-foreground transition-colors hover:bg-accent hover:text-foreground"
|
||||
|
|
|
|||
Loading…
Add table
Add a link
Reference in a new issue