Skip to content

Local UI to run and train LLMs and diffusion models. Supports GGUF, MLX, Qwen3.8, DeepSeek-V4.1, MiniMax-H3, Gemma 4, FLUX and more.

Studio: translate the live generation figures, the reply timing tooltip and the weight placement row (#16452)

oobaboogadcfc3924ee62
diff --git a/studio/frontend/src/components/assistant-ui/live-generation-line.tsx b/studio/frontend/src/components/assistant-ui/live-generation-line.tsx
index 094a3e24ecb..47debf5bb39 100644
--- a/studio/frontend/src/components/assistant-ui/live-generation-line.tsx
+++ b/studio/frontend/src/components/assistant-ui/live-generation-line.tsx
@@ -6,7 +6,7 @@ import {
useChatPreferencesStore,
useChatRuntimeStore,
} from "@/features/chat";
-import { useLocale } from "@/i18n";
+import { useLocale, useT } from "@/i18n";
import { useAuiState } from "@assistant-ui/react";
import type { FC } from "react";
@@ -24,6 +24,7 @@ const RunningGate: FC = () => {
const LiveDecodeFigures: FC = () => {
const locale = useLocale();
+ const t = useT();
// Keyed like the prefill indicator: a first turn reads "__default" until the thread persists.
const threadKey =
useAuiState(({ threadListItem }) => threadListItem.remoteId) ?? "__default";
@@ -38,7 +39,7 @@ const LiveDecodeFigures: FC = () => {
data-slot="live-generation-line"
className="aui-live-generation-line ml-[var(--icon-btn-inset)] mr-2 min-w-0 select-none font-mono text-ui-11p5 text-muted-foreground tabular-nums"
>
- {formatLiveDecode(decode, locale)}
+ {formatLiveDecode(decode, t, locale)}
</span>
);
};
diff --git a/studio/frontend/src/components/assistant-ui/message-timing.tsx b/studio/frontend/src/components/assistant-ui/message-timing.tsx
index 87711aabf5a..4f971b4c691 100644
--- a/studio/frontend/src/components/assistant-ui/message-timing.tsx
+++ b/studio/frontend/src/components/assistant-ui/message-timing.tsx
@@ -6,9 +6,12 @@ import {
TooltipContent,
TooltipTrigger,
} from "@/components/ui/tooltip";
+import { useT } from "@/i18n";
import { cn } from "@/lib/utils";
import type { FC } from "react";
+type Translate = ReturnType<typeof useT>;
+
const formatTimingMs = (ms: number | undefined): string => {
if (ms === undefined) return "—";
if (ms < 1000) return `${Math.round(ms)}ms`;
@@ -21,25 +24,29 @@ const formatNumber = (n: number): string => {
/** A user stop wins over the stream's last finish_reason. */
function stopReasonLabel(
+ t: Translate,
finishReason: unknown,
incompleteReason?: unknown,
): string | null {
- if (incompleteReason === "cancelled") return "Stopped by you";
+ if (incompleteReason === "cancelled") return t("generationStats.timing.stoppedByYou");
switch (finishReason) {
case "stop":
- return "End of turn";
+ return t("generationStats.timing.endOfTurn");
case "length":
- return "Length limit";
+ return t("generationStats.timing.lengthLimit");
case "tool_calls":
case "function_call":
- return "Tool call";
+ return t("generationStats.timing.toolCall");
case "content_filter":
- return "Content filter";
+ return t("generationStats.timing.contentFilter");
default:
return typeof finishReason === "string" && finishReason ? finishReason : null;
}
}
+const tokens = (t: Translate, n: number): string =>
+ t(n === 1 ? "generationStats.tokensOne" : "generationStats.tokens", { count: formatNumber(n) });
+
const formatRate = (r: number | undefined): string => {
if (r === undefined || !Number.isFinite(r)) return "—";
return `${Math.round(r).toLocaleString()} tok/s`;
@@ -51,6 +58,7 @@ export const MessageTiming: FC<{
}> = ({ className, side = "right" }) => {
const timing = useMessageTiming();
const message = useMessage();
+ const t = useT();
if (timing?.totalStreamTime === undefined) return null;
@@ -68,7 +76,7 @@ export const MessageTiming: FC<{
}
| undefined;
const st = custom?.serverTimings;
- const stopReason = stopReasonLabel(custom?.stopReason, custom?.incomplete?.reason);
+ const stopReason = stopReasonLabel(t, custom?.stopReason, custom?.incomplete?.reason);
const draftN = st?.draft_n ?? 0;
const draftAccepted = st?.draft_n_accepted;
const draftText =
@@ -102,7 +110,7 @@ export const MessageTiming: FC<{
<button
type="button"
data-slot="message-timing-trigger"
- aria-label="Message timing"
+ aria-label={t("generationStats.timing.trigger")}
className={cn(
"flex items-center rounded-[10px] p-1 font-mono text-chat-icon-fg text-ui-13 tabular-nums transition-colors hover:bg-chat-icon-bg-hover hover:text-chat-icon-fg-hover",
className,
@@ -124,7 +132,7 @@ export const MessageTiming: FC<{
<>
{timing.firstTokenTime !== undefined && (
<div className="flex items-center justify-between gap-4">
- <span className="text-muted-foreground">First token</span>
+ <span className="text-muted-foreground">{t("generationStats.timing.firstToken")}</span>
<span className="font-mono tabular-nums">
{formatTimingMs(timing.firstTokenTime)}
</span>
@@ -132,7 +140,7 @@ export const MessageTiming: FC<{
)}
{st?.diffusion_parallel_tok_s != null && (
<div className="flex items-center justify-between gap-4">
- <span className="text-muted-foreground">Speed (in-step)</span>
+ <span className="text-muted-foreground">{t("generationStats.timing.speedInStep")}</span>
<span className="font-mono tabular-nums">
{formatRate(st.diffusion_parallel_tok_s)}
</span>
@@ -140,7 +148,7 @@ export const MessageTiming: FC<{
)}
{st?.diffusion_effective_tok_s != null && (
<div className="flex items-center justify-between gap-4">
- <span className="text-muted-foreground">Effective</span>
+ <span className="text-muted-foreground">{t("generationStats.timing.effective")}</span>
<span className="font-mono tabular-nums">
{formatRate(st.diffusion_effective_tok_s)}
</span>
@@ -148,7 +156,7 @@ export const MessageTiming: FC<{
)}
{st?.diffusion_output_tok_s != null && (
<div className="flex items-center justify-between gap-4">
- <span className="text-muted-foreground">Output</span>
+ <span className="text-muted-foreground">{t("generationStats.timing.output")}</span>
<span className="font-mono tabular-nums">
{formatRate(st.diffusion_output_tok_s)}
</span>
@@ -156,26 +164,36 @@ export const MessageTiming: FC<{
)}
{st?.diffusion_steps != null && (
<div className="flex items-center justify-between gap-4">
- <span className="text-muted-foreground">Denoising</span>
+ <span className="text-muted-foreground">{t("generationStats.timing.denoising")}</span>
<span className="font-mono tabular-nums">
- {formatNumber(st.diffusion_steps)} steps
+ {t(
+ st.diffusion_steps === 1
+ ? "generationStats.timing.stepsOne"
+ : "generationStats.timing.steps",
+ { count: formatNumber(st.diffusion_steps) },
+ )}
{st?.diffusion_blocks != null
- ? `, ${formatNumber(st.diffusion_blocks)} block${st.diffusion_blocks === 1 ? "" : "s"}`
+ ? `, ${t(
+ st.diffusion_blocks === 1
+ ? "generationStats.timing.blocksOne"
+ : "generationStats.timing.blocks",
+ { count: formatNumber(st.diffusion_blocks) },
+ )}`
: ""}
</span>
</div>
)}
{st?.diffusion_canvas != null && (
<div className="flex items-center justify-between gap-4">
- <span className="text-muted-foreground">Canvas</span>
+ <span className="text-muted-foreground">{t("generationStats.timing.canvas")}</span>
<span className="font-mono tabular-nums">
- {formatNumber(st.diffusion_canvas)} tokens
+ {tokens(t, st.diffusion_canvas)}
</span>
</div>
)}
{(st?.diffusion_wall_ms ?? st?.predicted_ms) != null && (
<div className="flex items-center justify-between gap-4">
- <span className="text-muted-foreground">Generation</span>
+ <span className="text-muted-foreground">{t("generationStats.timing.generation")}</span>
<span className="font-mono tabular-nums">
{formatTimingMs(st.diffusion_wall_ms ?? st.predicted_ms)}
</span>
@@ -183,7 +201,7 @@ export const MessageTiming: FC<{
)}
{timing.tokenCount !== undefined && (
<div className="flex items-center justify-between gap-4">
- <span className="text-muted-foreground">Answer tokens</span>
+ <span className="text-muted-foreground">{t("generationStats.timing.answerTokens")}</span>
<span className="font-mono tabular-nums">
{formatNumber(timing.tokenCount)}
</span>
@@ -191,21 +209,21 @@ export const MessageTiming: FC<{
)}
{(st?.diffusion_prompt_n ?? st?.prompt_n) != null && (
<div className="flex items-center justify-between gap-4">
- <span className="text-muted-foreground">Prompt</span>
+ <span className="text-muted-foreground">{t("generationStats.timing.prompt")}</span>
<span className="font-mono tabular-nums">
- {formatNumber(st.diffusion_prompt_n ?? st.prompt_n)} tokens
+ {tokens(t, st.diffusion_prompt_n ?? st.prompt_n)}
</span>
</div>
)}
<div className="my-0.5 border-t border-border/40" />
<div className="flex items-center justify-between gap-4">
- <span className="text-muted-foreground">Total</span>
+ <span className="text-muted-foreground">{t("generationStats.timing.total")}</span>
<span className="font-mono tabular-nums">
{formatTimingMs(timing.totalStreamTime)}
</span>
</div>
<div className="flex items-center justify-between gap-4">
- <span className="text-muted-foreground">Chunks</span>
+ <span className="text-muted-foreground">{t("generationStats.timing.chunks")}</span>
<span className="font-mono tabular-nums">
{timing.totalChunks}
</span>
@@ -215,7 +233,7 @@ export const MessageTiming: FC<{
<>
{st?.prompt_ms != null && (
<div className="flex items-center justify-between gap-4">
- <span className="text-muted-foreground">Prompt eval</span>
+ <span className="text-muted-foreground">{t("generationStats.timing.promptEval")}</span>
<span className="font-mono tabular-nums">
{formatTimingMs(st.prompt_ms)}
</span>
@@ -225,7 +243,7 @@ export const MessageTiming: FC<{
st?.prompt_per_second != null &&
st.prompt_per_second > 0 && (
<div className="flex items-center justify-between gap-4">
- <span className="text-muted-foreground">Prompt speed</span>
+ <span className="text-muted-foreground">{t("generationStats.timing.promptSpeed")}</span>
<span className="font-mono tabular-nums">
{st.prompt_per_second.toFixed(1)} tok/s
</span>
@@ -233,7 +251,7 @@ export const MessageTiming: FC<{
)}
{hasPredicted && st?.predicted_ms != null && (
<div className="flex items-center justify-between gap-4">
- <span className="text-muted-foreground">Generation</span>
+ <span className="text-muted-foreground">{t("generationStats.timing.generation")}</span>
<span className="font-mono tabular-nums">
{formatTimingMs(st.predicted_ms)}
</span>
@@ -241,7 +259,7 @@ export const MessageTiming: FC<{
)}
{predictedRate != null && (
<div className="flex items-center justify-between gap-4">
- <span className="text-muted-foreground">Speed</span>
+ <span className="text-muted-foreground">{t("generationStats.timing.speed")}</span>
<span className="font-mono tabular-nums">
{predictedRate.toFixed(1)} tok/s
</span>
@@ -249,7 +267,7 @@ export const MessageTiming: FC<{
)}
{timing.tokenCount !== undefined && (
<div className="flex items-center justify-between gap-4">
- <span className="text-muted-foreground">Tokens</span>
+ <span className="text-muted-foreground">{t("generationStats.timing.tokens")}</span>
<span className="font-mono tabular-nums">
{formatNumber(timing.tokenCount)}
</span>
@@ -257,13 +275,13 @@ export const MessageTiming: FC<{
)}
{draftText !== null && (
<div className="flex items-center justify-between gap-4">
- <span className="text-muted-foreground">Draft accepted</span>
+ <span className="text-muted-foreground">{t("generationStats.timing.draftAccepted")}</span>
<span className="font-mono tabular-nums">{draftText}</span>
</div>
)}
{timing.firstTokenTime !== undefined && (
<div className="flex items-center justify-between gap-4">
- <span className="text-muted-foreground">First token</span>
+ <span className="text-muted-foreground">{t("generationStats.timing.firstToken")}</span>
<span className="font-mono tabular-nums">
{formatTimingMs(timing.firstTokenTime)}
</span>
@@ -271,7 +289,7 @@ export const MessageTiming: FC<{
)}
{st?.diffusion_steps != null && (
<div className="flex items-center justify-between gap-4">
- <span className="text-muted-foreground">Denoising steps</span>
+ <span className="text-muted-foreground">{t("generationStats.timing.denoisingSteps")}</span>
<span className="font-mono tabular-nums">
{formatNumber(st.diffusion_steps)}
</span>
@@ -279,7 +297,7 @@ export const MessageTiming: FC<{
)}
{st?.diffusion_blocks != null && (
<div className="flex items-center justify-between gap-4">
- <span className="text-muted-foreground">Blocks</span>
+ <span className="text-muted-foreground">{t("generationStats.timing.blocksLabel")}</span>
<span className="font-mono tabular-nums">
{formatNumber(st.diffusion_blocks)}
</span>
@@ -287,7 +305,7 @@ export const MessageTiming: FC<{
)}
{cacheHits > 0 && (
<div className="flex items-center justify-between gap-4">
- <span className="text-muted-foreground">Cache hits</span>
+ <span className="text-muted-foreground">{t("generationStats.timing.cacheHits")}</span>
<span className="font-mono tabular-nums">
{formatNumber(cacheHits)}
</span>
@@ -295,7 +313,7 @@ export const MessageTiming: FC<{
)}
{cacheWrites > 0 && (
<div className="flex items-center justify-between gap-4">
- <span className="text-muted-foreground">Cache writes</span>
+ <span className="text-muted-foreground">{t("generationStats.timing.cacheWrites")}</span>
<span className="font-mono tabular-nums">
{formatNumber(cacheWrites)}
</span>
@@ -303,20 +321,20 @@ export const MessageTiming: FC<{
)}
<div className="my-0.5 border-t border-border/40" />
<div className="flex items-center justify-between gap-4">
- <span className="text-muted-foreground">Total</span>
+ <span className="text-muted-foreground">{t("generationStats.timing.total")}</span>
<span className="font-mono tabular-nums">
{formatTimingMs(timing.totalStreamTime)}
</span>
</div>
<div className="flex items-center justify-between gap-4">
- <span className="text-muted-foreground">Chunks</span>
+ <span className="text-muted-foreground">{t("generationStats.timing.chunks")}</span>
<span className="font-mono tabular-nums">
{timing.totalChunks}
</span>
</div>
{stopReason !== null && (
<div className="flex items-center justify-between gap-4">
- <span className="text-muted-foreground">Stop reason</span>
+ <span className="text-muted-foreground">{t("generationStats.timing.stopReason")}</span>
<span>{stopReason}</span>
</div>
)}
@@ -326,7 +344,7 @@ export const MessageTiming: FC<{
<>
{timing.firstTokenTime !== undefined && (
<div className="flex items-center justify-between gap-4">
- <span className="text-muted-foreground">First token</span>
+ <span className="text-muted-foreground">{t("generationStats.timing.firstToken")}</span>
<span className="font-mono tabular-nums">
{formatTimingMs(timing.firstTokenTime)}
</span>
@@ -334,7 +352,7 @@ export const MessageTiming: FC<{
)}
{cacheHits > 0 && (
<div className="flex items-center justify-between gap-4">
- <span className="text-muted-foreground">Cache hits</span>
+ <span className="text-muted-foreground">{t("generationStats.timing.cacheHits")}</span>
<span className="font-mono tabular-nums">
{formatNumber(cacheHits)}
</span>
@@ -342,27 +360,27 @@ export const MessageTiming: FC<{
)}
{cacheWrites > 0 && (
<div className="flex items-center justify-between gap-4">
- <span className="text-muted-foreground">Cache writes</span>
+ <span className="text-muted-foreground">{t("generationStats.timing.cacheWrites")}</span>
<span className="font-mono tabular-nums">
{formatNumber(cacheWrites)}
</span>
</div>
)}
<div className="flex items-center justify-between gap-4">
- <span className="text-muted-foreground">Total</span>
+ <span className="text-muted-foreground">{t("generationStats.timing.total")}</span>
<span className="font-mono tabular-nums">
{formatTimingMs(timing.totalStreamTime)}
</span>
</div>
<div className="flex items-center justify-between gap-4">
- <span className="text-muted-foreground">Chunks</span>
+ <span className="text-muted-foreground">{t("generationStats.timing.chunks")}</span>
<span className="font-mono tabular-nums">
{timing.totalChunks}
</span>
</div>
{stopReason !== null && (
<div className="flex items-center justify-between gap-4">
- <span className="text-muted-foreground">Stop reason</span>
+ <span className="text-muted-foreground">{t("generationStats.timing.stopReason")}</span>
<span>{stopReason}</span>
</div>
)}
diff --git a/studio/frontend/src/components/assistant-ui/thread.tsx b/studio/frontend/src/components/assistant-ui/thread.tsx
index 4d2a0a86999..025e62fb61a 100644
--- a/studio/frontend/src/components/assistant-ui/thread.tsx
+++ b/studio/frontend/src/components/assistant-ui/thread.tsx
@@ -7704,24 +7704,25 @@ const GeneratingStatus: FC = () => {
);
const showDetails = useChatPreferencesStore((s) => s.showGenerationDetails);
const locale = useLocale();
+ const t = useT();
const percent = prefill ? prefillPercent(prefill) : null;
// Held at 100% until the first token: going back to Working... for that moment would flash too.
if (prefill === null || !labelShown || percent === null || decoding) {
return (
<span className="generating-glint text-sm text-muted-foreground">
- Working...
+ {t("generationStats.working")}
</span>
);
}
- const timeLeft = formatPrefillTimeLeft(prefill);
+ const timeLeft = formatPrefillTimeLeft(prefill, t);
return (
<span
data-slot="prefill-progress"
className="generating-glint text-sm text-muted-foreground tabular-nums whitespace-normal!"
>
- Processing prompt {percent}%
+ {t("generationStats.processingPrompt", { percent })}
{timeLeft ? ` · ${timeLeft}` : null}
- {showDetails ? ` · ${formatPrefillDetails(prefill, locale)}` : null}
+ {showDetails ? ` · ${formatPrefillDetails(prefill, t, locale)}` : null}
</span>
);
};
diff --git a/studio/frontend/src/features/chat/index.ts b/studio/frontend/src/features/chat/index.ts
index 92628eeb04a..fe4a0b15051 100644
--- a/studio/frontend/src/features/chat/index.ts
+++ b/studio/frontend/src/features/chat/index.ts
@@ -91,6 +91,7 @@ export {
} from "./stores/chat-runtime-store";
export { resolveStagedDiffusionClassification } from "./lib/gpu-placement";
export { describePlacement, offloadCountsFrom, offloadWarning } from "./lib/partial-offload";
+export type { PlacementPool } from "./lib/partial-offload";
export {
formatLiveDecode,
formatPrefillDetails,
diff --git a/studio/frontend/src/features/chat/lib/generation-progress.ts b/studio/frontend/src/features/chat/lib/generation-progress.ts
index 1ccaf5b86b2..a9d5ab734dd 100644
--- a/studio/frontend/src/features/chat/lib/generation-progress.ts
+++ b/studio/frontend/src/features/chat/lib/generation-progress.ts
@@ -8,6 +8,15 @@ import type {
LivePrefillFit,
} from "../stores/chat-runtime-store";
import { formatEta } from "../utils/format-transfer";
+import type { InterpolationValues, TranslationKey } from "@/i18n";
+
+export type Translate = (key: TranslationKey, values?: InterpolationValues) => string;
+
+function tokenCount(count: number, t: Translate, locale?: string): string {
+ return t(count === 1 ? "generationStats.tokensOne" : "generationStats.tokens", {
+ count: count.toLocaleString(locale),
+ });
+}
export type GenerationProgressUpdate =
| { phase: "prefill"; prefill: LivePrefillProgress }
@@ -70,14 +79,18 @@ export function readGenerationProgress(
* cached prefix like the percent. */
export function formatPrefillDetails(
prefill: LivePrefillProgress,
+ t: Translate,
locale?: string,
): string {
const cached = prefill.cached ?? 0;
- const done = Math.max(0, prefill.processed - cached).toLocaleString(locale);
+ const done = Math.max(0, prefill.processed - cached);
const parts = [
prefill.total
- ? `${done} / ${Math.max(0, prefill.total - cached).toLocaleString(locale)} tokens`
- : `${done} tokens`,
+ ? t("generationStats.tokensOf", {
+ done: done.toLocaleString(locale),
+ total: Math.max(0, prefill.total - cached).toLocaleString(locale),
+ })
+ : tokenCount(done, t, locale),
];
// The label shows only once prefillLabelShown has a rate it trusts, so the rate is past the first sample.
if (prefill.tokPerS !== null) {
@@ -192,17 +205,18 @@ export function prefillLabelShown(
return left !== null && left >= PREFILL_LABEL_MIN_SECONDS;
}
-/** "12s left", like the load toast; null without a trusted rate or under a second. */
-export function formatPrefillTimeLeft(prefill: LivePrefillProgress): string | null {
+export function formatPrefillTimeLeft(prefill: LivePrefillProgress, t: Translate): string | null {
const left = prefillSecondsLeft(prefill);
- return left !== null && left >= 1 ? `${formatEta(left)} left` : null;
+ return left !== null && left >= 1 ? t("generationStats.timeLeft", { eta: formatEta(left) }) : null;
}
/** "4,306 tokens · 58.3 tok/s · draft 1,791/5,030 accepted (36%)", every number in the reader's locale. */
-export function formatLiveDecode(decode: LiveDecodeProgress, locale?: string): string {
- const parts = [
- `${decode.tokens.toLocaleString(locale)} ${decode.tokens === 1 ? "token" : "tokens"}`,
- ];
+export function formatLiveDecode(
+ decode: LiveDecodeProgress,
+ t: Translate,
+ locale?: string,
+): string {
+ const parts = [tokenCount(decode.tokens, t, locale)];
if (decode.tokPerS !== null) {
const rate = decode.tokPerS.toLocaleString(locale, {
minimumFractionDigits: 1,
@@ -211,9 +225,12 @@ export function formatLiveDecode(decode: LiveDecodeProgress, locale?: string): s
parts.push(`${rate} tok/s`);
}
if (decode.draftN && decode.draftAccepted !== null) {
- const share = Math.round((decode.draftAccepted / decode.draftN) * 100);
parts.push(
- `draft ${decode.draftAccepted.toLocaleString(locale)}/${decode.draftN.toLocaleString(locale)} accepted (${share}%)`,
+ t("generationStats.draftAccepted", {
+ accepted: decode.draftAccepted.toLocaleString(locale),
+ drafted: decode.draftN.toLocaleString(locale),
+ percent: Math.round((decode.draftAccepted / decode.draftN) * 100),
+ }),
);
}
return parts.join(" · ");
diff --git a/studio/frontend/src/features/chat/lib/partial-offload.ts b/studio/frontend/src/features/chat/lib/partial-offload.ts
index 62b2b17147d..a3cd2aacb82 100644
--- a/studio/frontend/src/features/chat/lib/partial-offload.ts
+++ b/studio/frontend/src/features/chat/lib/partial-offload.ts
@@ -2,6 +2,7 @@
// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
import { formatGiB } from "../../../lib/memory/format";
+import type { Translate } from "./generation-progress";
import type { LoadedPlacement } from "../stores/chat-runtime-store";
import { cpuFallbackMessage } from "../utils/mmproj-fallback";
@@ -110,10 +111,6 @@ function weightAmount(mib: number): string {
: formatGiB(mib / 1024);
}
-function layerCount(n: number): string {
- return `${n} ${n === 1 ? "layer" : "layers"}`;
-}
-
/** --fit reports its own expert spill; a Manual load says it through n_cpu_moe instead. */
export function loadedPlacementFrom(response: {
is_gguf?: boolean | null;
@@ -147,47 +144,61 @@ export function loadedPlacementFrom(response: {
/** An overflowing layer has expert weights in RAM: all its routed experts, except --fit's first partial layer
* per device, which keeps only part of its FFN there, so the text says expert weights, not experts. */
-function describeLayers(p: LoadedPlacement): string | null {
+function describeLayers(p: LoadedPlacement, t: Translate): string | null {
const { offloaded, total, overflowLayers, expertCacheMib } = p;
if (offloaded === null || total === null || total <= 0) return null;
// A GPU build whose backend failed to start still logs "offloaded N/N"; its buffers are all host ones.
const noGpuBuffers = p.gpuWeightsMib === 0 && (p.hostWeightsMib ?? 0) > 0;
- if (offloaded <= 0 || noGpuBuffers) return `All ${total} layers on the CPU`;
+ if (offloaded <= 0 || noGpuBuffers) return t("generationStats.placement.allOnCpu", { total });
if (offloaded < total)
- return `${offloaded} of ${total} layers on the GPU, ${total - offloaded} on the CPU`;
+ return t("generationStats.placement.split", { gpu: offloaded, total, cpu: total - offloaded });
if (expertCacheMib !== null && expertCacheMib > 0) {
- return `All ${total} layers on the GPU, all experts in RAM (${weightAmount(expertCacheMib)} expert cache on the GPU)`;
+ return t("generationStats.placement.expertCache", { total, amount: weightAmount(expertCacheMib) });
}
if (overflowLayers !== null && overflowLayers > 0) {
- return `All ${total} layers on the GPU, expert weights of ${layerCount(overflowLayers)} in RAM`;
+ return t(
+ overflowLayers === 1
+ ? "generationStats.placement.expertSpillOne"
+ : "generationStats.placement.expertSpill",
+ { total, count: overflowLayers },
+ );
}
- return `All ${total} layers on the GPU`;
+ return t("generationStats.placement.allOnGpu", { total });
}
+export type PlacementPool = "unified" | "shared";
+
/** Both figures are shown: token embeddings stay in RAM even when every layer is on the GPU. On a machine whose
- * GPU and CPU share one pool, `pool` names it ("unified memory") and the weights are one figure. */
+ * GPU and CPU share one pool, `pool` names it and the weights are one figure. */
function describeWeights(
gpuWeightsMib: number | null,
hostWeightsMib: number | null,
- pool: string | null,
+ pool: PlacementPool | null,
+ t: Translate,
): string | null {
const gpu = gpuWeightsMib !== null && gpuWeightsMib > 0 ? gpuWeightsMib : 0;
const host =
hostWeightsMib !== null && hostWeightsMib > 0 ? hostWeightsMib : 0;
if (pool !== null && gpu + host > 0)
- return `${weightAmount(gpu + host)} in ${pool}`;
+ return t(
+ pool === "unified"
+ ? "generationStats.placement.inUnifiedMemory"
+ : "generationStats.placement.inSharedMemory",
+ { amount: weightAmount(gpu + host) },
+ );
const parts: string[] = [];
- if (gpu > 0) parts.push(`${weightAmount(gpu)} in VRAM`);
- if (host > 0) parts.push(`${weightAmount(host)} in RAM`);
+ if (gpu > 0) parts.push(t("generationStats.placement.inVram", { amount: weightAmount(gpu) }));
+ if (host > 0) parts.push(t("generationStats.placement.inRam", { amount: weightAmount(host) }));
return parts.length > 0 ? parts.join(" · ") : null;
}
export function describePlacement(
p: LoadedPlacement,
- pool: string | null = null,
+ t: Translate,
+ pool: PlacementPool | null = null,
): { layers: string | null; memory: string | null } {
return {
- layers: describeLayers(p),
- memory: describeWeights(p.gpuWeightsMib, p.hostWeightsMib, pool),
+ layers: describeLayers(p, t),
+ memory: describeWeights(p.gpuWeightsMib, p.hostWeightsMib, pool, t),
};
}
diff --git a/studio/frontend/src/features/model-picker/components/loaded-placement-row.tsx b/studio/frontend/src/features/model-picker/components/loaded-placement-row.tsx
index 150970cc84f..d975ce34a4a 100644
--- a/studio/frontend/src/features/model-picker/components/loaded-placement-row.tsx
+++ b/studio/frontend/src/features/model-picker/components/loaded-placement-row.tsx
@@ -3,17 +3,20 @@
import {
describePlacement,
+ type PlacementPool,
useChatPreferencesStore,
useChatRuntimeStore,
} from "@/features/chat";
+import { useT } from "@/i18n";
/** Where the resident GGUF's weights landed. Under Auto, --fit decides, and the layer count alone hides a
* MoE spill, so the experts and the RAM side are shown too. `pool` is the panel's single memory pool, if any. */
-export function LoadedPlacementRow({ pool }: { pool: string | null }) {
+export function LoadedPlacementRow({ pool }: { pool: PlacementPool | null }) {
+ const t = useT();
const enabled = useChatPreferencesStore((s) => s.showGenerationDetails);
const placement = useChatRuntimeStore((s) => s.loadedPlacement);
if (!enabled || placement === null) return null;
- const { layers, memory } = describePlacement(placement, pool);
+ const { layers, memory } = describePlacement(placement, t, pool);
if (layers === null && memory === null) return null;
return (
<div
@@ -21,7 +24,7 @@ export function LoadedPlacementRow({ pool }: { pool: string | null }) {
className="flex flex-col gap-1 border-b border-border pb-5"
>
<span className="flex min-h-8 items-center text-ui-13 font-medium leading-[1.25] tracking-nav text-foreground">
- Weight placement
+ {t("generationStats.placement.title")}
</span>
{layers !== null && (
<span className="text-ui-12 leading-snug text-muted-foreground">
diff --git a/studio/frontend/src/features/model-picker/components/model-config-page.tsx b/studio/frontend/src/features/model-picker/components/model-config-page.tsx
index 8380a037066..e0b8361da87 100644
--- a/studio/frontend/src/features/model-picker/components/model-config-page.tsx
+++ b/studio/frontend/src/features/model-picker/components/model-config-page.tsx
@@ -4230,7 +4230,7 @@ export function ModelConfigPage({
{isActiveModel && target.isGguf && !audioRuntimeGguf ? (
// The same pool the estimate above names: "Unified" on Apple Silicon, "Shared" on other one-pool machines.
<LoadedPlacementRow
- pool={singleMemoryPool ? (isAppleUnifiedMemory ? "unified memory" : "shared memory") : null}
+ pool={singleMemoryPool ? (isAppleUnifiedMemory ? "unified" : "shared") : null}
/>
) : null}
{target.isGguf && !audioRuntimeGguf && (
diff --git a/studio/frontend/src/i18n/locales/ar.ts b/studio/frontend/src/i18n/locales/ar.ts
index 400a52ea0a0..ad9b233717b 100644
--- a/studio/frontend/src/i18n/locales/ar.ts
+++ b/studio/frontend/src/i18n/locales/ar.ts
@@ -5454,4 +5454,59 @@ export const ar = {
update: "تحديث",
undo: "تراجع",
},
+ generationStats: {
+ working: "جارٍ التنفيذ...",
+ processingPrompt: "جارٍ معالجة الطلب {percent}%",
+ timeLeft: "يتبقى {eta}",
+ tokensOne: "التوكنات: {count}",
+ tokens: "التوكنات: {count}",
+ tokensOf: "التوكنات: {done} / {total}",
+ draftAccepted: "المسودة: قُبل {accepted}/{drafted} ({percent}%)",
+ timing: {
+ trigger: "توقيت الرسالة",
+ firstToken: "أول توكن",
+ speedInStep: "السرعة (داخل الخطوة)",
+ effective: "الفعلية",
+ output: "المخرجات",
+ denoising: "إزالة الضوضاء",
+ stepsOne: "الخطوات: {count}",
+ steps: "الخطوات: {count}",
+ blocksOne: "الكتل: {count}",
+ blocks: "الكتل: {count}",
+ canvas: "اللوحة",
+ generation: "التوليد",
+ answerTokens: "توكنات الإجابة",
+ prompt: "الطلب",
+ total: "الإجمالي",
+ chunks: "الأجزاء",
+ promptEval: "تقييم الطلب",
+ promptSpeed: "سرعة الطلب",
+ speed: "السرعة",
+ tokens: "التوكنات",
+ draftAccepted: "المقبول من المسودة",
+ denoisingSteps: "خطوات إزالة الضوضاء",
+ blocksLabel: "الكتل",
+ cacheHits: "إصابات الذاكرة المؤقتة",
+ cacheWrites: "كتابات الذاكرة المؤقتة",
+ stopReason: "سبب التوقف",
+ stoppedByYou: "أوقفته أنت",
+ endOfTurn: "نهاية الدور",
+ lengthLimit: "حد الطول",
+ toolCall: "استدعاء أداة",
+ contentFilter: "مرشّح المحتوى",
+ },
+ placement: {
+ title: "توزيع الأوزان",
+ allOnCpu: "جميع الطبقات ({total}) على CPU",
+ split: "الطبقات: {gpu} من {total} على GPU، و{cpu} على CPU",
+ allOnGpu: "جميع الطبقات ({total}) على GPU",
+ expertCache: "جميع الطبقات ({total}) على GPU، وكل الخبراء في RAM (ذاكرة مؤقتة للخبراء بحجم {amount} على GPU)",
+ expertSpillOne: "جميع الطبقات ({total}) على GPU، وأوزان خبراء {count} طبقة في RAM",
+ expertSpill: "جميع الطبقات ({total}) على GPU، وأوزان الخبراء في RAM، الطبقات: {count}",
+ inUnifiedMemory: "{amount} في الذاكرة الموحّدة",
+ inSharedMemory: "{amount} في الذاكرة المشتركة",
+ inVram: "{amount} في VRAM",
+ inRam: "{amount} في RAM",
+ },
+ },
} satisfies DeepPartialMessageTree<typeof en>;
diff --git a/studio/frontend/src/i18n/locales/cs.ts b/studio/frontend/src/i18n/locales/cs.ts
index 94c76c58863..21e608c73c9 100644
--- a/studio/frontend/src/i18n/locales/cs.ts
+++ b/studio/frontend/src/i18n/locales/cs.ts
@@ -5880,4 +5880,59 @@ export const cs = {
update: "Aktualizovat",
undo: "Vrátit zpět",
},
+ generationStats: {
+ working: "Probíhá zpracování...",
+ processingPrompt: "Zpracování promptu {percent} %",
+ timeLeft: "zbývá {eta}",
+ tokensOne: "tokeny: {count}",
+ tokens: "tokeny: {count}",
+ tokensOf: "tokeny: {done} / {total}",
+ draftAccepted: "návrh: přijato {accepted}/{drafted} ({percent} %)",
+ timing: {
+ trigger: "Časování zprávy",
+ firstToken: "První token",
+ speedInStep: "Rychlost (v kroku)",
+ effective: "Efektivní",
+ output: "Výstup",
+ denoising: "Odšumování",
+ stepsOne: "kroky: {count}",
+ steps: "kroky: {count}",
+ blocksOne: "bloky: {count}",
+ blocks: "bloky: {count}",
+ canvas: "Plátno",
+ generation: "Generování",
+ answerTokens: "Tokeny odpovědi",
+ prompt: "Prompt",
+ total: "Celkem",
+ chunks: "Části",
+ promptEval: "Vyhodnocení promptu",
+ promptSpeed: "Rychlost promptu",
+ speed: "Rychlost",
+ tokens: "Tokeny",
+ draftAccepted: "Přijatý návrh",
+ denoisingSteps: "Kroky odšumování",
+ blocksLabel: "Bloky",
+ cacheHits: "Zásahy mezipaměti",
+ cacheWrites: "Zápisy do mezipaměti",
+ stopReason: "Důvod zastavení",
+ stoppedByYou: "Zastaveno vámi",
+ endOfTurn: "Konec odpovědi",
+ lengthLimit: "Limit délky",
+ toolCall: "Volání nástroje",
+ contentFilter: "Filtr obsahu",
+ },
+ placement: {
+ title: "Umístění vah",
+ allOnCpu: "Všechny vrstvy ({total}) na CPU",
+ split: "{gpu} z {total} vrstev na GPU, {cpu} na CPU",
+ allOnGpu: "Všechny vrstvy ({total}) na GPU",
+ expertCache: "Všechny vrstvy ({total}) na GPU, všichni experti v RAM ({amount} mezipaměti expertů na GPU)",
+ expertSpillOne: "Všechny vrstvy ({total}) na GPU, váhy expertů {count} vrstvy v RAM",
+ expertSpill: "Všechny vrstvy ({total}) na GPU, váhy expertů v RAM, vrstev: {count}",
+ inUnifiedMemory: "{amount} ve sjednocené paměti",
+ inSharedMemory: "{amount} ve sdílené paměti",
+ inVram: "{amount} ve VRAM",
+ inRam: "{amount} v RAM",
+ },
+ },
} satisfies DeepPartialMessageTree<typeof en>;
diff --git a/studio/frontend/src/i18n/locales/de.ts b/studio/frontend/src/i18n/locales/de.ts
index 0dee35a8643..56ffdbecdb7 100644
--- a/studio/frontend/src/i18n/locales/de.ts
+++ b/studio/frontend/src/i18n/locales/de.ts
@@ -5528,4 +5528,59 @@ export const de = {
update: "Aktualisieren",
undo: "Rückgängig",
},
+ generationStats: {
+ working: "Wird bearbeitet...",
+ processingPrompt: "Prompt-Verarbeitung {percent}%",
+ timeLeft: "noch {eta}",
+ tokensOne: "{count} Token",
+ tokens: "{count} Tokens",
+ tokensOf: "{done} / {total} Tokens",
+ draftAccepted: "Draft {accepted}/{drafted} akzeptiert ({percent}%)",
+ timing: {
+ trigger: "Zeitstatistik der Nachricht",
+ firstToken: "Erstes Token",
+ speedInStep: "Geschwindigkeit (pro Schritt)",
+ effective: "Effektiv",
+ output: "Ausgabe",
+ denoising: "Denoising",
+ stepsOne: "{count} Schritt",
+ steps: "{count} Schritte",
+ blocksOne: "{count} Block",
+ blocks: "{count} Blöcke",
+ canvas: "Canvas",
+ generation: "Generierung",
+ answerTokens: "Antwort-Tokens",
+ prompt: "Prompt",
+ total: "Gesamt",
+ chunks: "Chunks",
+ promptEval: "Prompt-Auswertung",
+ promptSpeed: "Prompt-Geschwindigkeit",
+ speed: "Geschwindigkeit",
+ tokens: "Tokens",
+ draftAccepted: "Draft akzeptiert",
+ denoisingSteps: "Denoising-Schritte",
+ blocksLabel: "Blöcke",
+ cacheHits: "Cache-Treffer",
+ cacheWrites: "Cache-Schreibvorgänge",
+ stopReason: "Stoppgrund",
+ stoppedByYou: "Von dir gestoppt",
+ endOfTurn: "Ende der Antwort",
+ lengthLimit: "Längenlimit",
+ toolCall: "Tool-Aufruf",
+ contentFilter: "Inhaltsfilter",
+ },
+ placement: {
+ title: "Platzierung der Gewichte",
+ allOnCpu: "Alle {total} Schichten auf der CPU",
+ split: "{gpu} von {total} Schichten auf der GPU, {cpu} auf der CPU",
+ allOnGpu: "Alle {total} Schichten auf der GPU",
+ expertCache: "Alle {total} Schichten auf der GPU, alle Experten im RAM ({amount} Experten-Cache auf der GPU)",
+ expertSpillOne: "Alle {total} Schichten auf der GPU, Expertengewichte von {count} Schicht im RAM",
+ expertSpill: "Alle {total} Schichten auf der GPU, Expertengewichte von {count} Schichten im RAM",
+ inUnifiedMemory: "{amount} im Unified Memory",
+ inSharedMemory: "{amount} im gemeinsamen Speicher",
+ inVram: "{amount} im VRAM",
+ inRam: "{amount} im RAM",
+ },
+ },
} satisfies DeepPartialMessageTree<typeof en>;
diff --git a/studio/frontend/src/i18n/locales/en.ts b/studio/frontend/src/i18n/locales/en.ts
index 0c66d8b73d5..3a720f2b5b7 100644
--- a/studio/frontend/src/i18n/locales/en.ts
+++ b/studio/frontend/src/i18n/locales/en.ts
@@ -5493,4 +5493,64 @@ export const en = {
update: "Update",
undo: "Undo",
},
+ // The live prompt and reply figures (Settings > Chat > Display > Generation details), the reply's
+ // timing tooltip and the Run settings "Weight placement" row. Keep GPU, CPU, RAM and VRAM.
+ generationStats: {
+ working: "Working...",
+ processingPrompt: "Processing prompt {percent}%",
+ // {eta}: a short duration such as "12s" or "2m 10s".
+ timeLeft: "{eta} left",
+ tokensOne: "{count} token",
+ tokens: "{count} tokens",
+ tokensOf: "{done} / {total} tokens",
+ // Speculative decoding: of {drafted} drafted tokens, the model kept {accepted}.
+ draftAccepted: "draft {accepted}/{drafted} accepted ({percent}%)",
+ timing: {
+ trigger: "Message timing",
+ firstToken: "First token",
+ speedInStep: "Speed (in-step)",
+ effective: "Effective",
+ output: "Output",
+ denoising: "Denoising",
+ stepsOne: "{count} step",
+ steps: "{count} steps",
+ blocksOne: "{count} block",
+ blocks: "{count} blocks",
+ canvas: "Canvas",
+ generation: "Generation",
+ answerTokens: "Answer tokens",
+ prompt: "Prompt",
+ total: "Total",
+ chunks: "Chunks",
+ promptEval: "Prompt eval",
+ promptSpeed: "Prompt speed",
+ speed: "Speed",
+ tokens: "Tokens",
+ draftAccepted: "Draft accepted",
+ denoisingSteps: "Denoising steps",
+ blocksLabel: "Blocks",
+ cacheHits: "Cache hits",
+ cacheWrites: "Cache writes",
+ stopReason: "Stop reason",
+ stoppedByYou: "Stopped by you",
+ endOfTurn: "End of turn",
+ lengthLimit: "Length limit",
+ toolCall: "Tool call",
+ contentFilter: "Content filter",
+ },
+ placement: {
+ title: "Weight placement",
+ allOnCpu: "All {total} layers on the CPU",
+ split: "{gpu} of {total} layers on the GPU, {cpu} on the CPU",
+ allOnGpu: "All {total} layers on the GPU",
+ // A MoE with every routed expert in RAM and a GPU cache of the busiest ones; {amount} is like "18 GiB".
+ expertCache: "All {total} layers on the GPU, all experts in RAM ({amount} expert cache on the GPU)",
+ expertSpillOne: "All {total} layers on the GPU, expert weights of {count} layer in RAM",
+ expertSpill: "All {total} layers on the GPU, expert weights of {count} layers in RAM",
+ inUnifiedMemory: "{amount} in unified memory",
+ inSharedMemory: "{amount} in shared memory",
+ inVram: "{amount} in VRAM",
+ inRam: "{amount} in RAM",
+ },
+ },
} as const;
diff --git a/studio/frontend/src/i18n/locales/es-419.ts b/studio/frontend/src/i18n/locales/es-419.ts
index 1b55f349508..ecd42837ad6 100644
--- a/studio/frontend/src/i18n/locales/es-419.ts
+++ b/studio/frontend/src/i18n/locales/es-419.ts
@@ -5951,4 +5951,59 @@ export const es419 = {
update: "Actualizar",
undo: "Deshacer",
},
+ generationStats: {
+ working: "Procesando...",
+ processingPrompt: "Procesando el prompt {percent}%",
+ timeLeft: "quedan {eta}",
+ tokensOne: "{count} token",
+ tokens: "{count} tokens",
+ tokensOf: "{done} / {total} tokens",
+ draftAccepted: "borrador {accepted}/{drafted} aceptados ({percent}%)",
+ timing: {
+ trigger: "Tiempos del mensaje",
+ firstToken: "Primer token",
+ speedInStep: "Velocidad (por paso)",
+ effective: "Efectiva",
+ output: "Salida",
+ denoising: "Eliminación de ruido",
+ stepsOne: "{count} paso",
+ steps: "{count} pasos",
+ blocksOne: "{count} bloque",
+ blocks: "{count} bloques",
+ canvas: "Lienzo",
+ generation: "Generación",
+ answerTokens: "Tokens de respuesta",
+ prompt: "Prompt",
+ total: "Total",
+ chunks: "Fragmentos",
+ promptEval: "Evaluación del prompt",
+ promptSpeed: "Velocidad del prompt",
+ speed: "Velocidad",
+ tokens: "Tokens",
+ draftAccepted: "Borrador aceptado",
+ denoisingSteps: "Pasos de eliminación de ruido",
+ blocksLabel: "Bloques",
+ cacheHits: "Aciertos de caché",
+ cacheWrites: "Escrituras en caché",
+ stopReason: "Motivo de parada",
+ stoppedByYou: "Detenido por ti",
+ endOfTurn: "Fin del turno",
+ lengthLimit: "Límite de longitud",
+ toolCall: "Llamada a herramienta",
+ contentFilter: "Filtro de contenido",
+ },
+ placement: {
+ title: "Ubicación de los pesos",
+ allOnCpu: "Las {total} capas en la CPU",
+ split: "{gpu} de {total} capas en la GPU, {cpu} en la CPU",
+ allOnGpu: "Las {total} capas en la GPU",
+ expertCache: "Las {total} capas en la GPU, todos los expertos en RAM ({amount} de caché de expertos en la GPU)",
+ expertSpillOne: "Las {total} capas en la GPU, pesos de expertos de {count} capa en RAM",
+ expertSpill: "Las {total} capas en la GPU, pesos de expertos de {count} capas en RAM",
+ inUnifiedMemory: "{amount} en memoria unificada",
+ inSharedMemory: "{amount} en memoria compartida",
+ inVram: "{amount} en VRAM",
+ inRam: "{amount} en RAM",
+ },
+ },
} satisfies DeepPartialMessageTree<typeof en>;
diff --git a/studio/frontend/src/i18n/locales/es.ts b/studio/frontend/src/i18n/locales/es.ts
index 1d22876674f..d512e666ccd 100644
--- a/studio/frontend/src/i18n/locales/es.ts
+++ b/studio/frontend/src/i18n/locales/es.ts
@@ -5527,4 +5527,59 @@ export const es = {
update: "Actualizar",
undo: "Deshacer",
},
+ generationStats: {
+ working: "Procesando...",
+ processingPrompt: "Procesando el prompt {percent}%",
+ timeLeft: "quedan {eta}",
+ tokensOne: "{count} token",
+ tokens: "{count} tokens",
+ tokensOf: "{done} / {total} tokens",
+ draftAccepted: "borrador {accepted}/{drafted} aceptados ({percent}%)",
+ timing: {
+ trigger: "Tiempos del mensaje",
+ firstToken: "Primer token",
+ speedInStep: "Velocidad (por paso)",
+ effective: "Efectiva",
+ output: "Salida",
+ denoising: "Eliminación de ruido",
+ stepsOne: "{count} paso",
+ steps: "{count} pasos",
+ blocksOne: "{count} bloque",
+ blocks: "{count} bloques",
+ canvas: "Lienzo",
+ generation: "Generación",
+ answerTokens: "Tokens de respuesta",
+ prompt: "Prompt",
+ total: "Total",
+ chunks: "Fragmentos",
+ promptEval: "Evaluación del prompt",
+ promptSpeed: "Velocidad del prompt",
+ speed: "Velocidad",
+ tokens: "Tokens",
+ draftAccepted: "Borrador aceptado",
+ denoisingSteps: "Pasos de eliminación de ruido",
+ blocksLabel: "Bloques",
+ cacheHits: "Aciertos de caché",
+ cacheWrites: "Escrituras en caché",
+ stopReason: "Motivo de parada",
+ stoppedByYou: "Detenido por ti",
+ endOfTurn: "Fin del turno",
+ lengthLimit: "Límite de longitud",
+ toolCall: "Llamada a herramienta",
+ contentFilter: "Filtro de contenido",
+ },
+ placement: {
+ title: "Ubicación de los pesos",
+ allOnCpu: "Las {total} capas en la CPU",
+ split: "{gpu} de {total} capas en la GPU, {cpu} en la CPU",
+ allOnGpu: "Las {total} capas en la GPU",
+ expertCache: "Las {total} capas en la GPU, todos los expertos en RAM ({amount} de caché de expertos en la GPU)",
+ expertSpillOne: "Las {total} capas en la GPU, pesos de expertos de {count} capa en RAM",
+ expertSpill: "Las {total} capas en la GPU, pesos de expertos de {count} capas en RAM",
+ inUnifiedMemory: "{amount} en memoria unificada",
+ inSharedMemory: "{amount} en memoria compartida",
+ inVram: "{amount} en VRAM",
+ inRam: "{amount} en RAM",
+ },
+ },
} satisfies DeepPartialMessageTree<typeof en>;
diff --git a/studio/frontend/src/i18n/locales/fa.ts b/studio/frontend/src/i18n/locales/fa.ts
index 758050b0d36..21385c5de8c 100644
--- a/studio/frontend/src/i18n/locales/fa.ts
+++ b/studio/frontend/src/i18n/locales/fa.ts
@@ -5832,4 +5832,59 @@ export const fa = {
update: "به‌روزرسانی",
undo: "واگرد",
},
+ generationStats: {
+ working: "در حال انجام...",
+ processingPrompt: "پردازش پرامپت {percent}%",
+ timeLeft: "{eta} مانده",
+ tokensOne: "{count} توکن",
+ tokens: "{count} توکن",
+ tokensOf: "{done} / {total} توکن",
+ draftAccepted: "پیش‌نویس {accepted}/{drafted} پذیرفته شد ({percent}%)",
+ timing: {
+ trigger: "زمان‌بندی پیام",
+ firstToken: "اولین توکن",
+ speedInStep: "سرعت (درون گام)",
+ effective: "مؤثر",
+ output: "خروجی",
+ denoising: "نویززدایی",
+ stepsOne: "{count} گام",
+ steps: "{count} گام",
+ blocksOne: "{count} بلوک",
+ blocks: "{count} بلوک",
+ canvas: "بوم",
+ generation: "تولید",
+ answerTokens: "توکن‌های پاسخ",
+ prompt: "پرامپت",
+ total: "کل",
+ chunks: "تکه‌ها",
+ promptEval: "ارزیابی پرامپت",
+ promptSpeed: "سرعت پرامپت",
+ speed: "سرعت",
+ tokens: "توکن‌ها",
+ draftAccepted: "پیش‌نویس پذیرفته‌شده",
+ denoisingSteps: "گام‌های نویززدایی",
+ blocksLabel: "بلوک‌ها",
+ cacheHits: "برخورد کش",
+ cacheWrites: "نوشتن در کش",
+ stopReason: "دلیل توقف",
+ stoppedByYou: "توسط شما متوقف شد",
+ endOfTurn: "پایان نوبت",
+ lengthLimit: "محدودیت طول",
+ toolCall: "فراخوانی ابزار",
+ contentFilter: "فیلتر محتوا",
+ },
+ placement: {
+ title: "محل قرارگیری وزن‌ها",
+ allOnCpu: "هر {total} لایه روی CPU",
+ split: "{gpu} از {total} لایه روی GPU، {cpu} روی CPU",
+ allOnGpu: "هر {total} لایه روی GPU",
+ expertCache: "هر {total} لایه روی GPU، همهٔ متخصص‌ها در RAM ({amount} کش متخصص روی GPU)",
+ expertSpillOne: "هر {total} لایه روی GPU، وزن‌های متخصص {count} لایه در RAM",
+ expertSpill: "هر {total} لایه روی GPU، وزن‌های متخصص {count} لایه در RAM",
+ inUnifiedMemory: "{amount} در حافظهٔ یکپارچه",
+ inSharedMemory: "{amount} در حافظهٔ مشترک",
+ inVram: "{amount} در VRAM",
+ inRam: "{amount} در RAM",
+ },
+ },
} satisfies DeepPartialMessageTree<typeof en>;
diff --git a/studio/frontend/src/i18n/locales/fr.ts b/studio/frontend/src/i18n/locales/fr.ts
index ea0651c97c4..3a19215fe2e 100644
--- a/studio/frontend/src/i18n/locales/fr.ts
+++ b/studio/frontend/src/i18n/locales/fr.ts
@@ -5538,4 +5538,59 @@ export const fr = {
update: "Mettre à jour",
undo: "Annuler",
},
+ generationStats: {
+ working: "Traitement en cours...",
+ processingPrompt: "Traitement du prompt {percent} %",
+ timeLeft: "reste {eta}",
+ tokensOne: "{count} token",
+ tokens: "{count} tokens",
+ tokensOf: "{done} / {total} tokens",
+ draftAccepted: "brouillon {accepted}/{drafted} acceptés ({percent} %)",
+ timing: {
+ trigger: "Temps du message",
+ firstToken: "Premier token",
+ speedInStep: "Vitesse (par étape)",
+ effective: "Effective",
+ output: "Sortie",
+ denoising: "Débruitage",
+ stepsOne: "{count} étape",
+ steps: "{count} étapes",
+ blocksOne: "{count} bloc",
+ blocks: "{count} blocs",
+ canvas: "Canevas",
+ generation: "Génération",
+ answerTokens: "Tokens de réponse",
+ prompt: "Prompt",
+ total: "Total",
+ chunks: "Fragments",
+ promptEval: "Évaluation du prompt",
+ promptSpeed: "Vitesse du prompt",
+ speed: "Vitesse",
+ tokens: "Tokens",
+ draftAccepted: "Brouillon accepté",
+ denoisingSteps: "Étapes de débruitage",
+ blocksLabel: "Blocs",
+ cacheHits: "Lectures du cache",
+ cacheWrites: "Écritures du cache",
+ stopReason: "Motif d’arrêt",
+ stoppedByYou: "Arrêté par vous",
+ endOfTurn: "Fin de la réponse",
+ lengthLimit: "Limite de longueur",
+ toolCall: "Appel d’outil",
+ contentFilter: "Filtre de contenu",
+ },
+ placement: {
+ title: "Placement des poids",
+ allOnCpu: "Toutes les {total} couches sur le CPU",
+ split: "{gpu} couches sur {total} sur le GPU, {cpu} sur le CPU",
+ allOnGpu: "Toutes les {total} couches sur le GPU",
+ expertCache: "Toutes les {total} couches sur le GPU, tous les experts en RAM ({amount} de cache d’experts sur le GPU)",
+ expertSpillOne: "Toutes les {total} couches sur le GPU, poids des experts de {count} couche en RAM",
+ expertSpill: "Toutes les {total} couches sur le GPU, poids des experts de {count} couches en RAM",
+ inUnifiedMemory: "{amount} en mémoire unifiée",
+ inSharedMemory: "{amount} en mémoire partagée",
+ inVram: "{amount} en VRAM",
+ inRam: "{amount} en RAM",
+ },
+ },
} satisfies DeepPartialMessageTree<typeof en>;
diff --git a/studio/frontend/src/i18n/locales/he.ts b/studio/frontend/src/i18n/locales/he.ts
index 7999f7b850c..5d3d46296f8 100644
--- a/studio/frontend/src/i18n/locales/he.ts
+++ b/studio/frontend/src/i18n/locales/he.ts
@@ -5552,4 +5552,59 @@ export const he = {
update: "עדכן",
undo: "ביטול פעולה",
},
+ generationStats: {
+ working: "מעבד...",
+ processingPrompt: "עיבוד הפרומפט {percent}%",
+ timeLeft: "נותרו {eta}",
+ tokensOne: "{count} טוקן",
+ tokens: "{count} טוקנים",
+ tokensOf: "{done} / {total} טוקנים",
+ draftAccepted: "טיוטה: {accepted}/{drafted} התקבלו ({percent}%)",
+ timing: {
+ trigger: "תזמון ההודעה",
+ firstToken: "טוקן ראשון",
+ speedInStep: "מהירות (בתוך שלב)",
+ effective: "אפקטיבית",
+ output: "פלט",
+ denoising: "הסרת רעש",
+ stepsOne: "{count} שלב",
+ steps: "{count} שלבים",
+ blocksOne: "{count} בלוק",
+ blocks: "{count} בלוקים",
+ canvas: "קנבס",
+ generation: "יצירה",
+ answerTokens: "טוקני תשובה",
+ prompt: "פרומפט",
+ total: "סה״כ",
+ chunks: "מקטעים",
+ promptEval: "הערכת פרומפט",
+ promptSpeed: "מהירות פרומפט",
+ speed: "מהירות",
+ tokens: "טוקנים",
+ draftAccepted: "טיוטה שהתקבלה",
+ denoisingSteps: "שלבי הסרת רעש",
+ blocksLabel: "בלוקים",
+ cacheHits: "פגיעות במטמון",
+ cacheWrites: "כתיבות למטמון",
+ stopReason: "סיבת עצירה",
+ stoppedByYou: "נעצר על ידך",
+ endOfTurn: "סוף התור",
+ lengthLimit: "מגבלת אורך",
+ toolCall: "קריאה לכלי",
+ contentFilter: "מסנן תוכן",
+ },
+ placement: {
+ title: "מיקום המשקולות",
+ allOnCpu: "כל {total} השכבות על ה-CPU",
+ split: "{gpu} מתוך {total} שכבות על ה-GPU, {cpu} על ה-CPU",
+ allOnGpu: "כל {total} השכבות על ה-GPU",
+ expertCache: "כל {total} השכבות על ה-GPU, כל המומחים ב-RAM (מטמון מומחים של {amount} על ה-GPU)",
+ expertSpillOne: "כל {total} השכבות על ה-GPU, משקולות המומחים של {count} שכבה ב-RAM",
+ expertSpill: "כל {total} השכבות על ה-GPU, משקולות המומחים של {count} שכבות ב-RAM",
+ inUnifiedMemory: "{amount} בזיכרון מאוחד",
+ inSharedMemory: "{amount} בזיכרון משותף",
+ inVram: "{amount} ב-VRAM",
+ inRam: "{amount} ב-RAM",
+ },
+ },
} satisfies DeepPartialMessageTree<typeof en>;
diff --git a/studio/frontend/src/i18n/locales/hi.ts b/studio/frontend/src/i18n/locales/hi.ts
index 1f295bc15e3..8ada02c706e 100644
--- a/studio/frontend/src/i18n/locales/hi.ts
+++ b/studio/frontend/src/i18n/locales/hi.ts
@@ -5465,4 +5465,59 @@ export const hi = {
update: "अपडेट करें",
undo: "पूर्ववत करें",
},
+ generationStats: {
+ working: "प्रक्रिया जारी है...",
+ processingPrompt: "प्रॉम्प्ट प्रोसेस हो रहा है {percent}%",
+ timeLeft: "{eta} बाकी",
+ tokensOne: "{count} टोकन",
+ tokens: "{count} टोकन",
+ tokensOf: "{done} / {total} टोकन",
+ draftAccepted: "ड्राफ़्ट {accepted}/{drafted} स्वीकृत ({percent}%)",
+ timing: {
+ trigger: "संदेश का समय",
+ firstToken: "पहला टोकन",
+ speedInStep: "गति (स्टेप में)",
+ effective: "प्रभावी",
+ output: "आउटपुट",
+ denoising: "डीनॉइज़िंग",
+ stepsOne: "{count} स्टेप",
+ steps: "{count} स्टेप्स",
+ blocksOne: "{count} ब्लॉक",
+ blocks: "{count} ब्लॉक",
+ canvas: "कैनवास",
+ generation: "जनरेशन",
+ answerTokens: "उत्तर टोकन",
+ prompt: "प्रॉम्प्ट",
+ total: "कुल",
+ chunks: "चंक",
+ promptEval: "प्रॉम्प्ट मूल्यांकन",
+ promptSpeed: "प्रॉम्प्ट गति",
+ speed: "गति",
+ tokens: "टोकन",
+ draftAccepted: "स्वीकृत ड्राफ़्ट",
+ denoisingSteps: "डीनॉइज़िंग स्टेप्स",
+ blocksLabel: "ब्लॉक",
+ cacheHits: "कैश हिट",
+ cacheWrites: "कैश राइट",
+ stopReason: "रुकने का कारण",
+ stoppedByYou: "आपने रोका",
+ endOfTurn: "टर्न समाप्त",
+ lengthLimit: "लंबाई सीमा",
+ toolCall: "टूल कॉल",
+ contentFilter: "कंटेंट फ़िल्टर",
+ },
+ placement: {
+ title: "वेट्स का स्थान",
+ allOnCpu: "सभी {total} परतें CPU पर",
+ split: "{total} में से {gpu} परतें GPU पर, {cpu} CPU पर",
+ allOnGpu: "सभी {total} परतें GPU पर",
+ expertCache: "सभी {total} परतें GPU पर, सभी एक्सपर्ट RAM में (GPU पर {amount} एक्सपर्ट कैश)",
+ expertSpillOne: "सभी {total} परतें GPU पर, {count} परत के एक्सपर्ट वेट्स RAM में",
+ expertSpill: "सभी {total} परतें GPU पर, {count} परतों के एक्सपर्ट वेट्स RAM में",
+ inUnifiedMemory: "{amount} यूनिफाइड मेमोरी में",
+ inSharedMemory: "{amount} साझा मेमोरी में",
+ inVram: "{amount} VRAM में",
+ inRam: "{amount} RAM में",
+ },
+ },
} satisfies DeepPartialMessageTree<typeof en>;
diff --git a/studio/frontend/src/i18n/locales/id.ts b/studio/frontend/src/i18n/locales/id.ts
index 49f71d4f9a0..551a2410f41 100644
--- a/studio/frontend/src/i18n/locales/id.ts
+++ b/studio/frontend/src/i18n/locales/id.ts
@@ -5864,4 +5864,59 @@ export const id = {
update: "Perbarui",
undo: "Urungkan",
},
+ generationStats: {
+ working: "Memproses...",
+ processingPrompt: "Memproses prompt {percent}%",
+ timeLeft: "{eta} lagi",
+ tokensOne: "{count} token",
+ tokens: "{count} token",
+ tokensOf: "{done} / {total} token",
+ draftAccepted: "draf {accepted}/{drafted} diterima ({percent}%)",
+ timing: {
+ trigger: "Waktu pesan",
+ firstToken: "Token pertama",
+ speedInStep: "Kecepatan (dalam langkah)",
+ effective: "Efektif",
+ output: "Output",
+ denoising: "Denoising",
+ stepsOne: "{count} langkah",
+ steps: "{count} langkah",
+ blocksOne: "{count} blok",
+ blocks: "{count} blok",
+ canvas: "Kanvas",
+ generation: "Pembuatan",
+ answerTokens: "Token jawaban",
+ prompt: "Prompt",
+ total: "Total",
+ chunks: "Potongan",
+ promptEval: "Evaluasi prompt",
+ promptSpeed: "Kecepatan prompt",
+ speed: "Kecepatan",
+ tokens: "Token",
+ draftAccepted: "Draf diterima",
+ denoisingSteps: "Langkah denoising",
+ blocksLabel: "Blok",
+ cacheHits: "Cache hit",
+ cacheWrites: "Penulisan cache",
+ stopReason: "Alasan berhenti",
+ stoppedByYou: "Dihentikan oleh Anda",
+ endOfTurn: "Akhir giliran",
+ lengthLimit: "Batas panjang",
+ toolCall: "Panggilan alat",
+ contentFilter: "Filter konten",
+ },
+ placement: {
+ title: "Penempatan bobot",
+ allOnCpu: "Semua {total} lapisan di CPU",
+ split: "{gpu} dari {total} lapisan di GPU, {cpu} di CPU",
+ allOnGpu: "Semua {total} lapisan di GPU",
+ expertCache: "Semua {total} lapisan di GPU, semua expert di RAM (cache expert {amount} di GPU)",
+ expertSpillOne: "Semua {total} lapisan di GPU, bobot expert dari {count} lapisan di RAM",
+ expertSpill: "Semua {total} lapisan di GPU, bobot expert dari {count} lapisan di RAM",
+ inUnifiedMemory: "{amount} di memori terpadu",
+ inSharedMemory: "{amount} di memori bersama",
+ inVram: "{amount} di VRAM",
+ inRam: "{amount} di RAM",
+ },
+ },
} satisfies DeepPartialMessageTree<typeof en>;
diff --git a/studio/frontend/src/i18n/locales/it.ts b/studio/frontend/src/i18n/locales/it.ts
index 8aa98f74755..ca2abea82df 100644
--- a/studio/frontend/src/i18n/locales/it.ts
+++ b/studio/frontend/src/i18n/locales/it.ts
@@ -5523,4 +5523,59 @@ export const it = {
update: "Aggiorna",
undo: "Annulla",
},
+ generationStats: {
+ working: "Elaborazione...",
+ processingPrompt: "Elaborazione del prompt {percent}%",
+ timeLeft: "{eta} rimanenti",
+ tokensOne: "{count} token",
+ tokens: "{count} token",
+ tokensOf: "{done} / {total} token",
+ draftAccepted: "draft {accepted}/{drafted} accettati ({percent}%)",
+ timing: {
+ trigger: "Tempi del messaggio",
+ firstToken: "Primo token",
+ speedInStep: "Velocità (nello step)",
+ effective: "Effettiva",
+ output: "Output",
+ denoising: "Denoising",
+ stepsOne: "{count} step",
+ steps: "{count} step",
+ blocksOne: "{count} blocco",
+ blocks: "{count} blocchi",
+ canvas: "Canvas",
+ generation: "Generazione",
+ answerTokens: "Token di risposta",
+ prompt: "Prompt",
+ total: "Totale",
+ chunks: "Chunk",
+ promptEval: "Valutazione prompt",
+ promptSpeed: "Velocità prompt",
+ speed: "Velocità",
+ tokens: "Token",
+ draftAccepted: "Draft accettati",
+ denoisingSteps: "Step di denoising",
+ blocksLabel: "Blocchi",
+ cacheHits: "Hit della cache",
+ cacheWrites: "Scritture in cache",
+ stopReason: "Motivo dell'arresto",
+ stoppedByYou: "Interrotto da te",
+ endOfTurn: "Fine del turno",
+ lengthLimit: "Limite di lunghezza",
+ toolCall: "Chiamata a strumento",
+ contentFilter: "Filtro dei contenuti",
+ },
+ placement: {
+ title: "Posizione dei pesi",
+ allOnCpu: "Tutti i {total} livelli sulla CPU",
+ split: "{gpu} livelli su {total} sulla GPU, {cpu} sulla CPU",
+ allOnGpu: "Tutti i {total} livelli sulla GPU",
+ expertCache: "Tutti i {total} livelli sulla GPU, tutti gli esperti in RAM ({amount} di cache degli esperti sulla GPU)",
+ expertSpillOne: "Tutti i {total} livelli sulla GPU, pesi degli esperti di {count} livello in RAM",
+ expertSpill: "Tutti i {total} livelli sulla GPU, pesi degli esperti di {count} livelli in RAM",
+ inUnifiedMemory: "{amount} in memoria unificata",
+ inSharedMemory: "{amount} in memoria condivisa",
+ inVram: "{amount} in VRAM",
+ inRam: "{amount} in RAM",
+ },
+ },
} as const;
diff --git a/studio/frontend/src/i18n/locales/ja.ts b/studio/frontend/src/i18n/locales/ja.ts
index c742ad8a293..554f2318b7e 100644
--- a/studio/frontend/src/i18n/locales/ja.ts
+++ b/studio/frontend/src/i18n/locales/ja.ts
@@ -5405,4 +5405,59 @@ export const ja = {
update: "更新",
undo: "元に戻す",
},
+ generationStats: {
+ working: "処理中...",
+ processingPrompt: "プロンプトを処理中 {percent}%",
+ timeLeft: "残り {eta}",
+ tokensOne: "{count} トークン",
+ tokens: "{count} トークン",
+ tokensOf: "{done} / {total} トークン",
+ draftAccepted: "ドラフト {accepted}/{drafted} 採用 ({percent}%)",
+ timing: {
+ trigger: "メッセージのタイミング",
+ firstToken: "最初のトークン",
+ speedInStep: "速度(ステップ内)",
+ effective: "実効",
+ output: "出力",
+ denoising: "ノイズ除去",
+ stepsOne: "{count} ステップ",
+ steps: "{count} ステップ",
+ blocksOne: "{count} ブロック",
+ blocks: "{count} ブロック",
+ canvas: "キャンバス",
+ generation: "生成",
+ answerTokens: "回答トークン",
+ prompt: "プロンプト",
+ total: "合計",
+ chunks: "チャンク",
+ promptEval: "プロンプト評価",
+ promptSpeed: "プロンプト速度",
+ speed: "速度",
+ tokens: "トークン",
+ draftAccepted: "採用されたドラフト",
+ denoisingSteps: "ノイズ除去ステップ",
+ blocksLabel: "ブロック",
+ cacheHits: "キャッシュヒット",
+ cacheWrites: "キャッシュ書き込み",
+ stopReason: "停止理由",
+ stoppedByYou: "ユーザーが停止",
+ endOfTurn: "ターン終了",
+ lengthLimit: "長さの上限",
+ toolCall: "ツール呼び出し",
+ contentFilter: "コンテンツフィルター",
+ },
+ placement: {
+ title: "重みの配置",
+ allOnCpu: "全 {total} レイヤーが CPU 上",
+ split: "{total} レイヤー中 {gpu} が GPU、{cpu} が CPU",
+ allOnGpu: "全 {total} レイヤーが GPU 上",
+ expertCache: "全 {total} レイヤーが GPU 上、全エキスパートは RAM(GPU 上に {amount} のエキスパートキャッシュ)",
+ expertSpillOne: "全 {total} レイヤーが GPU 上、{count} レイヤーのエキスパートの重みは RAM",
+ expertSpill: "全 {total} レイヤーが GPU 上、{count} レイヤーのエキスパートの重みは RAM",
+ inUnifiedMemory: "ユニファイドメモリに {amount}",
+ inSharedMemory: "共有メモリに {amount}",
+ inVram: "VRAM に {amount}",
+ inRam: "RAM に {amount}",
+ },
+ },
} satisfies DeepPartialMessageTree<typeof en>;
diff --git a/studio/frontend/src/i18n/locales/ko.ts b/studio/frontend/src/i18n/locales/ko.ts
index 8338bd8ddc4..a5368d31cf8 100644
--- a/studio/frontend/src/i18n/locales/ko.ts
+++ b/studio/frontend/src/i18n/locales/ko.ts
@@ -5445,4 +5445,59 @@ export const ko = {
update: "업데이트",
undo: "실행 취소",
},
+ generationStats: {
+ working: "처리 중...",
+ processingPrompt: "프롬프트 처리 중 {percent}%",
+ timeLeft: "{eta} 남음",
+ tokensOne: "토큰 {count}개",
+ tokens: "토큰 {count}개",
+ tokensOf: "{done} / {total} 토큰",
+ draftAccepted: "초안 {accepted}/{drafted} 수락 ({percent}%)",
+ timing: {
+ trigger: "메시지 타이밍",
+ firstToken: "첫 토큰",
+ speedInStep: "속도(스텝 내)",
+ effective: "실효",
+ output: "출력",
+ denoising: "노이즈 제거",
+ stepsOne: "{count}스텝",
+ steps: "{count}스텝",
+ blocksOne: "{count}블록",
+ blocks: "{count}블록",
+ canvas: "캔버스",
+ generation: "생성",
+ answerTokens: "답변 토큰",
+ prompt: "프롬프트",
+ total: "합계",
+ chunks: "청크",
+ promptEval: "프롬프트 평가",
+ promptSpeed: "프롬프트 속도",
+ speed: "속도",
+ tokens: "토큰",
+ draftAccepted: "수락된 초안",
+ denoisingSteps: "노이즈 제거 스텝",
+ blocksLabel: "블록",
+ cacheHits: "캐시 적중",
+ cacheWrites: "캐시 쓰기",
+ stopReason: "중지 이유",
+ stoppedByYou: "사용자가 중지함",
+ endOfTurn: "턴 종료",
+ lengthLimit: "길이 제한",
+ toolCall: "도구 호출",
+ contentFilter: "콘텐츠 필터",
+ },
+ placement: {
+ title: "가중치 배치",
+ allOnCpu: "레이어 {total}개 모두 CPU에",
+ split: "레이어 {total}개 중 {gpu}개는 GPU, {cpu}개는 CPU에",
+ allOnGpu: "레이어 {total}개 모두 GPU에",
+ expertCache: "레이어 {total}개 모두 GPU에, 모든 전문가는 RAM에 (GPU에 전문가 캐시 {amount})",
+ expertSpillOne: "레이어 {total}개 모두 GPU에, 레이어 {count}개의 전문가 가중치는 RAM에",
+ expertSpill: "레이어 {total}개 모두 GPU에, 레이어 {count}개의 전문가 가중치는 RAM에",
+ inUnifiedMemory: "통합 메모리에 {amount}",
+ inSharedMemory: "공유 메모리에 {amount}",
+ inVram: "VRAM에 {amount}",
+ inRam: "RAM에 {amount}",
+ },
+ },
} satisfies DeepPartialMessageTree<typeof en>;
diff --git a/studio/frontend/src/i18n/locales/nl.ts b/studio/frontend/src/i18n/locales/nl.ts
index 49be4475690..754d7423b5a 100644
--- a/studio/frontend/src/i18n/locales/nl.ts
+++ b/studio/frontend/src/i18n/locales/nl.ts
@@ -5921,4 +5921,59 @@ export const nl = {
update: "Bijwerken",
undo: "Ongedaan maken",
},
+ generationStats: {
+ working: "Bezig...",
+ processingPrompt: "Prompt verwerken {percent}%",
+ timeLeft: "nog {eta}",
+ tokensOne: "{count} token",
+ tokens: "{count} tokens",
+ tokensOf: "{done} / {total} tokens",
+ draftAccepted: "draft {accepted}/{drafted} geaccepteerd ({percent}%)",
+ timing: {
+ trigger: "Berichttiming",
+ firstToken: "Eerste token",
+ speedInStep: "Snelheid (binnen stap)",
+ effective: "Effectief",
+ output: "Uitvoer",
+ denoising: "Denoising",
+ stepsOne: "{count} stap",
+ steps: "{count} stappen",
+ blocksOne: "{count} blok",
+ blocks: "{count} blokken",
+ canvas: "Canvas",
+ generation: "Generatie",
+ answerTokens: "Antwoordtokens",
+ prompt: "Prompt",
+ total: "Totaal",
+ chunks: "Chunks",
+ promptEval: "Prompt-evaluatie",
+ promptSpeed: "Promptsnelheid",
+ speed: "Snelheid",
+ tokens: "Tokens",
+ draftAccepted: "Draft geaccepteerd",
+ denoisingSteps: "Denoisingstappen",
+ blocksLabel: "Blokken",
+ cacheHits: "Cachetreffers",
+ cacheWrites: "Cacheschrijfacties",
+ stopReason: "Stopreden",
+ stoppedByYou: "Door jou gestopt",
+ endOfTurn: "Einde van beurt",
+ lengthLimit: "Lengtelimiet",
+ toolCall: "Toolaanroep",
+ contentFilter: "Inhoudsfilter",
+ },
+ placement: {
+ title: "Plaatsing van gewichten",
+ allOnCpu: "Alle {total} lagen op de CPU",
+ split: "{gpu} van {total} lagen op de GPU, {cpu} op de CPU",
+ allOnGpu: "Alle {total} lagen op de GPU",
+ expertCache: "Alle {total} lagen op de GPU, alle experts in RAM ({amount} expertcache op de GPU)",
+ expertSpillOne: "Alle {total} lagen op de GPU, expertgewichten van {count} laag in RAM",
+ expertSpill: "Alle {total} lagen op de GPU, expertgewichten van {count} lagen in RAM",
+ inUnifiedMemory: "{amount} in unified memory",
+ inSharedMemory: "{amount} in gedeeld geheugen",
+ inVram: "{amount} in VRAM",
+ inRam: "{amount} in RAM",
+ },
+ },
} satisfies DeepPartialMessageTree<typeof en>;
diff --git a/studio/frontend/src/i18n/locales/pl.ts b/studio/frontend/src/i18n/locales/pl.ts
index 9577f39d686..369a1b9997e 100644
--- a/studio/frontend/src/i18n/locales/pl.ts
+++ b/studio/frontend/src/i18n/locales/pl.ts
@@ -5916,4 +5916,59 @@ export const pl = {
update: "Aktualizuj",
undo: "Cofnij",
},
+ generationStats: {
+ working: "Przetwarzanie...",
+ processingPrompt: "Przetwarzanie promptu {percent}%",
+ timeLeft: "zostało {eta}",
+ tokensOne: "tokeny: {count}",
+ tokens: "tokeny: {count}",
+ tokensOf: "tokeny: {done} / {total}",
+ draftAccepted: "szkic: zaakceptowano {accepted}/{drafted} ({percent}%)",
+ timing: {
+ trigger: "Czas wiadomości",
+ firstToken: "Pierwszy token",
+ speedInStep: "Szybkość (w kroku)",
+ effective: "Efektywna",
+ output: "Wyjście",
+ denoising: "Odszumianie",
+ stepsOne: "kroki: {count}",
+ steps: "kroki: {count}",
+ blocksOne: "bloki: {count}",
+ blocks: "bloki: {count}",
+ canvas: "Płótno",
+ generation: "Generowanie",
+ answerTokens: "Tokeny odpowiedzi",
+ prompt: "Prompt",
+ total: "Łącznie",
+ chunks: "Fragmenty",
+ promptEval: "Ewaluacja promptu",
+ promptSpeed: "Szybkość promptu",
+ speed: "Szybkość",
+ tokens: "Tokeny",
+ draftAccepted: "Zaakceptowane szkice",
+ denoisingSteps: "Kroki odszumiania",
+ blocksLabel: "Bloki",
+ cacheHits: "Trafienia w pamięci podręcznej",
+ cacheWrites: "Zapisy do pamięci podręcznej",
+ stopReason: "Powód zatrzymania",
+ stoppedByYou: "Zatrzymane przez Ciebie",
+ endOfTurn: "Koniec tury",
+ lengthLimit: "Limit długości",
+ toolCall: "Wywołanie narzędzia",
+ contentFilter: "Filtr treści",
+ },
+ placement: {
+ title: "Rozmieszczenie wag",
+ allOnCpu: "Wszystkie warstwy ({total}) na CPU",
+ split: "{gpu} z {total} warstw na GPU, {cpu} na CPU",
+ allOnGpu: "Wszystkie warstwy ({total}) na GPU",
+ expertCache: "Wszystkie warstwy ({total}) na GPU, wszyscy eksperci w RAM ({amount} pamięci podręcznej ekspertów na GPU)",
+ expertSpillOne: "Wszystkie warstwy ({total}) na GPU, wagi ekspertów z {count} warstwy w RAM",
+ expertSpill: "Wszystkie warstwy ({total}) na GPU, wagi ekspertów w RAM, warstwy: {count}",
+ inUnifiedMemory: "{amount} w pamięci zunifikowanej",
+ inSharedMemory: "{amount} w pamięci współdzielonej",
+ inVram: "{amount} w VRAM",
+ inRam: "{amount} w RAM",
+ },
+ },
} satisfies DeepPartialMessageTree<typeof en>;
diff --git a/studio/frontend/src/i18n/locales/pt-br.ts b/studio/frontend/src/i18n/locales/pt-br.ts
index f9a555d1e93..28869b668ee 100644
--- a/studio/frontend/src/i18n/locales/pt-br.ts
+++ b/studio/frontend/src/i18n/locales/pt-br.ts
@@ -5483,4 +5483,59 @@ export const ptBR = {
update: "Atualizar",
undo: "Desfazer",
},
+ generationStats: {
+ working: "Processando...",
+ processingPrompt: "Processando prompt {percent}%",
+ timeLeft: "faltam {eta}",
+ tokensOne: "{count} token",
+ tokens: "{count} tokens",
+ tokensOf: "{done} / {total} tokens",
+ draftAccepted: "rascunho {accepted}/{drafted} aceitos ({percent}%)",
+ timing: {
+ trigger: "Tempo da mensagem",
+ firstToken: "Primeiro token",
+ speedInStep: "Velocidade (no passo)",
+ effective: "Efetiva",
+ output: "Saída",
+ denoising: "Remoção de ruído",
+ stepsOne: "{count} passo",
+ steps: "{count} passos",
+ blocksOne: "{count} bloco",
+ blocks: "{count} blocos",
+ canvas: "Canvas",
+ generation: "Geração",
+ answerTokens: "Tokens da resposta",
+ prompt: "Prompt",
+ total: "Total",
+ chunks: "Chunks",
+ promptEval: "Avaliação do prompt",
+ promptSpeed: "Velocidade do prompt",
+ speed: "Velocidade",
+ tokens: "Tokens",
+ draftAccepted: "Rascunho aceito",
+ denoisingSteps: "Passos de remoção de ruído",
+ blocksLabel: "Blocos",
+ cacheHits: "Acertos de cache",
+ cacheWrites: "Gravações em cache",
+ stopReason: "Motivo da parada",
+ stoppedByYou: "Interrompido por você",
+ endOfTurn: "Fim do turno",
+ lengthLimit: "Limite de tamanho",
+ toolCall: "Chamada de ferramenta",
+ contentFilter: "Filtro de conteúdo",
+ },
+ placement: {
+ title: "Posicionamento dos pesos",
+ allOnCpu: "Todas as {total} camadas na CPU",
+ split: "{gpu} de {total} camadas na GPU, {cpu} na CPU",
+ allOnGpu: "Todas as {total} camadas na GPU",
+ expertCache: "Todas as {total} camadas na GPU, todos os especialistas na RAM ({amount} de cache de especialistas na GPU)",
+ expertSpillOne: "Todas as {total} camadas na GPU, pesos de especialistas de {count} camada na RAM",
+ expertSpill: "Todas as {total} camadas na GPU, pesos de especialistas de {count} camadas na RAM",
+ inUnifiedMemory: "{amount} na memória unificada",
+ inSharedMemory: "{amount} na memória compartilhada",
+ inVram: "{amount} na VRAM",
+ inRam: "{amount} na RAM",
+ },
+ },
} satisfies DeepPartialMessageTree<typeof en>;
diff --git a/studio/frontend/src/i18n/locales/ru.ts b/studio/frontend/src/i18n/locales/ru.ts
index 39344b33e29..89d69e20f31 100644
--- a/studio/frontend/src/i18n/locales/ru.ts
+++ b/studio/frontend/src/i18n/locales/ru.ts
@@ -5482,4 +5482,59 @@ export const ru = {
update: "Обновить",
undo: "Отменить",
},
+ generationStats: {
+ working: "Обработка...",
+ processingPrompt: "Обработка промпта {percent}%",
+ timeLeft: "осталось {eta}",
+ tokensOne: "токены: {count}",
+ tokens: "токены: {count}",
+ tokensOf: "токены: {done} / {total}",
+ draftAccepted: "черновик: принято {accepted}/{drafted} ({percent}%)",
+ timing: {
+ trigger: "Время генерации сообщения",
+ firstToken: "Первый токен",
+ speedInStep: "Скорость (в шаге)",
+ effective: "Эффективная",
+ output: "Вывод",
+ denoising: "Удаление шума",
+ stepsOne: "шаги: {count}",
+ steps: "шаги: {count}",
+ blocksOne: "блоки: {count}",
+ blocks: "блоки: {count}",
+ canvas: "Холст",
+ generation: "Генерация",
+ answerTokens: "Токены ответа",
+ prompt: "Промпт",
+ total: "Всего",
+ chunks: "Фрагменты",
+ promptEval: "Обработка промпта",
+ promptSpeed: "Скорость промпта",
+ speed: "Скорость",
+ tokens: "Токены",
+ draftAccepted: "Принято из черновика",
+ denoisingSteps: "Шаги удаления шума",
+ blocksLabel: "Блоки",
+ cacheHits: "Попадания в кеш",
+ cacheWrites: "Записи в кеш",
+ stopReason: "Причина остановки",
+ stoppedByYou: "Остановлено вами",
+ endOfTurn: "Конец ответа",
+ lengthLimit: "Лимит длины",
+ toolCall: "Вызов инструмента",
+ contentFilter: "Фильтр контента",
+ },
+ placement: {
+ title: "Размещение весов",
+ allOnCpu: "Все слои ({total}) на CPU",
+ split: "Слои: {gpu} из {total} на GPU, {cpu} на CPU",
+ allOnGpu: "Все слои ({total}) на GPU",
+ expertCache: "Все слои ({total}) на GPU, все эксперты в RAM (кеш экспертов {amount} на GPU)",
+ expertSpillOne: "Все слои ({total}) на GPU, веса экспертов {count} слоя в RAM",
+ expertSpill: "Все слои ({total}) на GPU, веса экспертов в RAM, слоёв: {count}",
+ inUnifiedMemory: "{amount} в единой памяти",
+ inSharedMemory: "{amount} в общей памяти",
+ inVram: "{amount} в VRAM",
+ inRam: "{amount} в RAM",
+ },
+ },
} satisfies DeepPartialMessageTree<typeof en>;
diff --git a/studio/frontend/src/i18n/locales/sv.ts b/studio/frontend/src/i18n/locales/sv.ts
index af7647fa7d9..a288847cbeb 100644
--- a/studio/frontend/src/i18n/locales/sv.ts
+++ b/studio/frontend/src/i18n/locales/sv.ts
@@ -5697,4 +5697,59 @@ export const sv = {
update: "Uppdatera",
undo: "Ångra",
},
+ generationStats: {
+ working: "Arbetar...",
+ processingPrompt: "Bearbetar prompt {percent}%",
+ timeLeft: "{eta} kvar",
+ tokensOne: "{count} token",
+ tokens: "{count} token",
+ tokensOf: "{done} / {total} token",
+ draftAccepted: "utkast {accepted}/{drafted} accepterade ({percent}%)",
+ timing: {
+ trigger: "Meddelandetider",
+ firstToken: "Första token",
+ speedInStep: "Hastighet (inom steg)",
+ effective: "Effektiv",
+ output: "Utdata",
+ denoising: "Avbrusning",
+ stepsOne: "{count} steg",
+ steps: "{count} steg",
+ blocksOne: "{count} block",
+ blocks: "{count} block",
+ canvas: "Arbetsyta",
+ generation: "Generering",
+ answerTokens: "Svarstoken",
+ prompt: "Prompt",
+ total: "Totalt",
+ chunks: "Delar",
+ promptEval: "Promptbearbetning",
+ promptSpeed: "Prompthastighet",
+ speed: "Hastighet",
+ tokens: "Token",
+ draftAccepted: "Accepterade utkast",
+ denoisingSteps: "Avbrusningssteg",
+ blocksLabel: "Block",
+ cacheHits: "Cacheträffar",
+ cacheWrites: "Cacheskrivningar",
+ stopReason: "Stopporsak",
+ stoppedByYou: "Stoppad av dig",
+ endOfTurn: "Turen avslutad",
+ lengthLimit: "Längdgräns",
+ toolCall: "Verktygsanrop",
+ contentFilter: "Innehållsfilter",
+ },
+ placement: {
+ title: "Viktplacering",
+ allOnCpu: "Alla {total} lager på CPU:n",
+ split: "{gpu} av {total} lager på GPU:n, {cpu} på CPU:n",
+ allOnGpu: "Alla {total} lager på GPU:n",
+ expertCache: "Alla {total} lager på GPU:n, alla experter i RAM ({amount} expertcache på GPU:n)",
+ expertSpillOne: "Alla {total} lager på GPU:n, expertvikter för {count} lager i RAM",
+ expertSpill: "Alla {total} lager på GPU:n, expertvikter för {count} lager i RAM",
+ inUnifiedMemory: "{amount} i enhetligt minne",
+ inSharedMemory: "{amount} i delat minne",
+ inVram: "{amount} i VRAM",
+ inRam: "{amount} i RAM",
+ },
+ },
} as const;
diff --git a/studio/frontend/src/i18n/locales/th.ts b/studio/frontend/src/i18n/locales/th.ts
index 814036b1d06..c20e4bb60ee 100644
--- a/studio/frontend/src/i18n/locales/th.ts
+++ b/studio/frontend/src/i18n/locales/th.ts
@@ -5584,4 +5584,59 @@ export const th = {
update: "อัปเดต",
undo: "เลิกทำ",
},
+ generationStats: {
+ working: "กำลังดำเนินการ...",
+ processingPrompt: "กำลังประมวลผลพรอมต์ {percent}%",
+ timeLeft: "เหลืออีก {eta}",
+ tokensOne: "{count} โทเค็น",
+ tokens: "{count} โทเค็น",
+ tokensOf: "{done} / {total} โทเค็น",
+ draftAccepted: "ยอมรับดราฟต์ {accepted}/{drafted} ({percent}%)",
+ timing: {
+ trigger: "เวลาของข้อความ",
+ firstToken: "โทเค็นแรก",
+ speedInStep: "ความเร็ว (ในขั้นตอน)",
+ effective: "ความเร็วจริง",
+ output: "เอาต์พุต",
+ denoising: "การลดสัญญาณรบกวน",
+ stepsOne: "{count} ขั้นตอน",
+ steps: "{count} ขั้นตอน",
+ blocksOne: "{count} บล็อก",
+ blocks: "{count} บล็อก",
+ canvas: "แคนวาส",
+ generation: "การสร้าง",
+ answerTokens: "โทเค็นคำตอบ",
+ prompt: "พรอมต์",
+ total: "ทั้งหมด",
+ chunks: "ส่วนย่อย",
+ promptEval: "ประมวลผลพรอมต์",
+ promptSpeed: "ความเร็วพรอมต์",
+ speed: "ความเร็ว",
+ tokens: "โทเค็น",
+ draftAccepted: "ดราฟต์ที่ยอมรับ",
+ denoisingSteps: "ขั้นตอนลดสัญญาณรบกวน",
+ blocksLabel: "บล็อก",
+ cacheHits: "ใช้แคชซ้ำ",
+ cacheWrites: "เขียนแคช",
+ stopReason: "เหตุผลที่หยุด",
+ stoppedByYou: "คุณหยุดเอง",
+ endOfTurn: "จบเทิร์น",
+ lengthLimit: "ขีดจำกัดความยาว",
+ toolCall: "การเรียกใช้เครื่องมือ",
+ contentFilter: "ตัวกรองเนื้อหา",
+ },
+ placement: {
+ title: "ตำแหน่งน้ำหนักโมเดล",
+ allOnCpu: "ทั้ง {total} เลเยอร์อยู่บน CPU",
+ split: "{gpu} จาก {total} เลเยอร์อยู่บน GPU และ {cpu} อยู่บน CPU",
+ allOnGpu: "ทั้ง {total} เลเยอร์อยู่บน GPU",
+ expertCache: "ทั้ง {total} เลเยอร์อยู่บน GPU และ expert ทั้งหมดอยู่ใน RAM (แคช expert {amount} บน GPU)",
+ expertSpillOne: "ทั้ง {total} เลเยอร์อยู่บน GPU และน้ำหนัก expert ของ {count} เลเยอร์อยู่ใน RAM",
+ expertSpill: "ทั้ง {total} เลเยอร์อยู่บน GPU และน้ำหนัก expert ของ {count} เลเยอร์อยู่ใน RAM",
+ inUnifiedMemory: "{amount} ในหน่วยความจำแบบรวม",
+ inSharedMemory: "{amount} ในหน่วยความจำที่ใช้ร่วมกัน",
+ inVram: "{amount} ใน VRAM",
+ inRam: "{amount} ใน RAM",
+ },
+ },
} satisfies DeepPartialMessageTree<typeof en>;
diff --git a/studio/frontend/src/i18n/locales/tr.ts b/studio/frontend/src/i18n/locales/tr.ts
index da2d17ec068..74bdf62799e 100644
--- a/studio/frontend/src/i18n/locales/tr.ts
+++ b/studio/frontend/src/i18n/locales/tr.ts
@@ -5860,4 +5860,59 @@ export const tr = {
update: "Güncelle",
undo: "Geri al",
},
+ generationStats: {
+ working: "İşleniyor...",
+ processingPrompt: "İstem işleniyor %{percent}",
+ timeLeft: "{eta} kaldı",
+ tokensOne: "{count} token",
+ tokens: "{count} token",
+ tokensOf: "{done} / {total} token",
+ draftAccepted: "taslak {accepted}/{drafted} kabul edildi (%{percent})",
+ timing: {
+ trigger: "Mesaj zamanlaması",
+ firstToken: "İlk token",
+ speedInStep: "Hız (adım içi)",
+ effective: "Etkin",
+ output: "Çıktı",
+ denoising: "Gürültü giderme",
+ stepsOne: "{count} adım",
+ steps: "{count} adım",
+ blocksOne: "{count} blok",
+ blocks: "{count} blok",
+ canvas: "Tuval",
+ generation: "Oluşturma",
+ answerTokens: "Yanıt tokenları",
+ prompt: "İstem",
+ total: "Toplam",
+ chunks: "Parçalar",
+ promptEval: "İstem değerlendirme",
+ promptSpeed: "İstem hızı",
+ speed: "Hız",
+ tokens: "Token",
+ draftAccepted: "Kabul edilen taslak",
+ denoisingSteps: "Gürültü giderme adımları",
+ blocksLabel: "Bloklar",
+ cacheHits: "Önbellek isabetleri",
+ cacheWrites: "Önbellek yazmaları",
+ stopReason: "Durma nedeni",
+ stoppedByYou: "Siz durdurdunuz",
+ endOfTurn: "Tur sonu",
+ lengthLimit: "Uzunluk sınırı",
+ toolCall: "Araç çağrısı",
+ contentFilter: "İçerik filtresi",
+ },
+ placement: {
+ title: "Ağırlık yerleşimi",
+ allOnCpu: "{total} katmanın tümü CPU'da",
+ split: "{total} katmandan {gpu} tanesi GPU'da, {cpu} tanesi CPU'da",
+ allOnGpu: "{total} katmanın tümü GPU'da",
+ expertCache: "{total} katmanın tümü GPU'da, tüm uzmanlar RAM'de (GPU'da {amount} uzman önbelleği)",
+ expertSpillOne: "{total} katmanın tümü GPU'da, {count} katmanın uzman ağırlıkları RAM'de",
+ expertSpill: "{total} katmanın tümü GPU'da, {count} katmanın uzman ağırlıkları RAM'de",
+ inUnifiedMemory: "{amount} birleşik bellekte",
+ inSharedMemory: "{amount} paylaşılan bellekte",
+ inVram: "{amount} VRAM'de",
+ inRam: "{amount} RAM'de",
+ },
+ },
} satisfies DeepPartialMessageTree<typeof en>;
diff --git a/studio/frontend/src/i18n/locales/uk.ts b/studio/frontend/src/i18n/locales/uk.ts
index 6218cff3ab6..0db36668d53 100644
--- a/studio/frontend/src/i18n/locales/uk.ts
+++ b/studio/frontend/src/i18n/locales/uk.ts
@@ -5907,4 +5907,59 @@ export const uk = {
update: "Оновити",
undo: "Скасувати",
},
+ generationStats: {
+ working: "Виконання...",
+ processingPrompt: "Обробка запиту {percent}%",
+ timeLeft: "залишилось {eta}",
+ tokensOne: "токени: {count}",
+ tokens: "токени: {count}",
+ tokensOf: "токени: {done} / {total}",
+ draftAccepted: "чернетка: прийнято {accepted}/{drafted} ({percent}%)",
+ timing: {
+ trigger: "Час генерування повідомлення",
+ firstToken: "Перший токен",
+ speedInStep: "Швидкість (у кроці)",
+ effective: "Ефективна",
+ output: "Вивід",
+ denoising: "Знешумлення",
+ stepsOne: "кроки: {count}",
+ steps: "кроки: {count}",
+ blocksOne: "блоки: {count}",
+ blocks: "блоки: {count}",
+ canvas: "Полотно",
+ generation: "Генерування",
+ answerTokens: "Токени відповіді",
+ prompt: "Запит",
+ total: "Усього",
+ chunks: "Фрагменти",
+ promptEval: "Обробка запиту",
+ promptSpeed: "Швидкість запиту",
+ speed: "Швидкість",
+ tokens: "Токени",
+ draftAccepted: "Прийнято з чернетки",
+ denoisingSteps: "Кроки знешумлення",
+ blocksLabel: "Блоки",
+ cacheHits: "Влучання в кеш",
+ cacheWrites: "Записи в кеш",
+ stopReason: "Причина зупинки",
+ stoppedByYou: "Зупинено вами",
+ endOfTurn: "Кінець відповіді",
+ lengthLimit: "Ліміт довжини",
+ toolCall: "Виклик інструмента",
+ contentFilter: "Фільтр вмісту",
+ },
+ placement: {
+ title: "Розміщення ваг",
+ allOnCpu: "Усі шари ({total}) на CPU",
+ split: "Шари: {gpu} з {total} на GPU, {cpu} на CPU",
+ allOnGpu: "Усі шари ({total}) на GPU",
+ expertCache: "Усі шари ({total}) на GPU, усі експерти в RAM (кеш експертів {amount} на GPU)",
+ expertSpillOne: "Усі шари ({total}) на GPU, ваги експертів {count} шару в RAM",
+ expertSpill: "Усі шари ({total}) на GPU, ваги експертів у RAM, шарів: {count}",
+ inUnifiedMemory: "{amount} в уніфікованій пам’яті",
+ inSharedMemory: "{amount} у спільній пам’яті",
+ inVram: "{amount} у VRAM",
+ inRam: "{amount} у RAM",
+ },
+ },
} satisfies DeepPartialMessageTree<typeof en>;
diff --git a/studio/frontend/src/i18n/locales/vi.ts b/studio/frontend/src/i18n/locales/vi.ts
index 6e30d2c1f08..d39be0785fe 100644
--- a/studio/frontend/src/i18n/locales/vi.ts
+++ b/studio/frontend/src/i18n/locales/vi.ts
@@ -5863,4 +5863,59 @@ export const vi = {
update: "Cập nhật",
undo: "Hoàn tác",
},
+ generationStats: {
+ working: "Đang xử lý...",
+ processingPrompt: "Đang xử lý lời nhắc {percent}%",
+ timeLeft: "còn {eta}",
+ tokensOne: "{count} token",
+ tokens: "{count} token",
+ tokensOf: "{done} / {total} token",
+ draftAccepted: "bản nháp được chấp nhận {accepted}/{drafted} ({percent}%)",
+ timing: {
+ trigger: "Thời gian của tin nhắn",
+ firstToken: "Token đầu tiên",
+ speedInStep: "Tốc độ (trong bước)",
+ effective: "Hiệu dụng",
+ output: "Đầu ra",
+ denoising: "Khử nhiễu",
+ stepsOne: "{count} bước",
+ steps: "{count} bước",
+ blocksOne: "{count} khối",
+ blocks: "{count} khối",
+ canvas: "Khung vẽ",
+ generation: "Quá trình tạo",
+ answerTokens: "Token câu trả lời",
+ prompt: "Lời nhắc",
+ total: "Tổng",
+ chunks: "Đoạn",
+ promptEval: "Xử lý lời nhắc",
+ promptSpeed: "Tốc độ lời nhắc",
+ speed: "Tốc độ",
+ tokens: "Token",
+ draftAccepted: "Bản nháp được chấp nhận",
+ denoisingSteps: "Bước khử nhiễu",
+ blocksLabel: "Khối",
+ cacheHits: "Lượt dùng bộ nhớ đệm",
+ cacheWrites: "Lượt ghi bộ nhớ đệm",
+ stopReason: "Lý do dừng",
+ stoppedByYou: "Bạn đã dừng",
+ endOfTurn: "Kết thúc lượt",
+ lengthLimit: "Giới hạn độ dài",
+ toolCall: "Lệnh gọi công cụ",
+ contentFilter: "Bộ lọc nội dung",
+ },
+ placement: {
+ title: "Vị trí trọng số",
+ allOnCpu: "Cả {total} lớp trên CPU",
+ split: "{gpu} trong {total} lớp trên GPU, {cpu} trên CPU",
+ allOnGpu: "Cả {total} lớp trên GPU",
+ expertCache: "Cả {total} lớp trên GPU, mọi expert trong RAM ({amount} bộ nhớ đệm expert trên GPU)",
+ expertSpillOne: "Cả {total} lớp trên GPU, trọng số expert của {count} lớp trong RAM",
+ expertSpill: "Cả {total} lớp trên GPU, trọng số expert của {count} lớp trong RAM",
+ inUnifiedMemory: "{amount} trong bộ nhớ hợp nhất",
+ inSharedMemory: "{amount} trong bộ nhớ dùng chung",
+ inVram: "{amount} trong VRAM",
+ inRam: "{amount} trong RAM",
+ },
+ },
} satisfies DeepPartialMessageTree<typeof en>;
diff --git a/studio/frontend/src/i18n/locales/zh-CN.ts b/studio/frontend/src/i18n/locales/zh-CN.ts
index 7d31c2628bd..c06d7f1163a 100644
--- a/studio/frontend/src/i18n/locales/zh-CN.ts
+++ b/studio/frontend/src/i18n/locales/zh-CN.ts
@@ -5356,4 +5356,59 @@ export const zhCN = {
update: "更新",
undo: "撤销",
},
+ generationStats: {
+ working: "正在处理...",
+ processingPrompt: "正在处理提示词 {percent}%",
+ timeLeft: "剩余 {eta}",
+ tokensOne: "{count} 个 token",
+ tokens: "{count} 个 token",
+ tokensOf: "{done} / {total} 个 token",
+ draftAccepted: "草稿接受 {accepted}/{drafted}({percent}%)",
+ timing: {
+ trigger: "消息耗时",
+ firstToken: "首个 token",
+ speedInStep: "速度(步内)",
+ effective: "有效速度",
+ output: "输出",
+ denoising: "去噪",
+ stepsOne: "{count} 步",
+ steps: "{count} 步",
+ blocksOne: "{count} 个块",
+ blocks: "{count} 个块",
+ canvas: "画布",
+ generation: "生成",
+ answerTokens: "回答 token",
+ prompt: "提示词",
+ total: "总计",
+ chunks: "分块数",
+ promptEval: "提示词处理",
+ promptSpeed: "提示词速度",
+ speed: "速度",
+ tokens: "Token",
+ draftAccepted: "草稿接受",
+ denoisingSteps: "去噪步数",
+ blocksLabel: "块数",
+ cacheHits: "缓存命中",
+ cacheWrites: "缓存写入",
+ stopReason: "停止原因",
+ stoppedByYou: "由你停止",
+ endOfTurn: "回合结束",
+ lengthLimit: "长度上限",
+ toolCall: "工具调用",
+ contentFilter: "内容过滤",
+ },
+ placement: {
+ title: "权重放置",
+ allOnCpu: "全部 {total} 层位于 CPU",
+ split: "{total} 层中 {gpu} 层位于 GPU,{cpu} 层位于 CPU",
+ allOnGpu: "全部 {total} 层位于 GPU",
+ expertCache: "全部 {total} 层位于 GPU,所有专家位于 RAM(GPU 上有 {amount} 专家缓存)",
+ expertSpillOne: "全部 {total} 层位于 GPU,{count} 层的专家权重位于 RAM",
+ expertSpill: "全部 {total} 层位于 GPU,{count} 层的专家权重位于 RAM",
+ inUnifiedMemory: "{amount} 位于统一内存",
+ inSharedMemory: "{amount} 位于共享内存",
+ inVram: "{amount} 位于 VRAM",
+ inRam: "{amount} 位于 RAM",
+ },
+ },
} satisfies DeepPartialMessageTree<typeof en>;
diff --git a/studio/frontend/src/i18n/locales/zh-TW.ts b/studio/frontend/src/i18n/locales/zh-TW.ts
index 802bf6c1b34..cee2eae9e98 100644
--- a/studio/frontend/src/i18n/locales/zh-TW.ts
+++ b/studio/frontend/src/i18n/locales/zh-TW.ts
@@ -5535,4 +5535,59 @@ export const zhTW = {
update: "更新",
undo: "復原",
},
+ generationStats: {
+ working: "處理中...",
+ processingPrompt: "正在處理提示詞 {percent}%",
+ timeLeft: "剩餘 {eta}",
+ tokensOne: "{count} 個 token",
+ tokens: "{count} 個 token",
+ tokensOf: "{done} / {total} 個 token",
+ draftAccepted: "草稿接受 {accepted}/{drafted}({percent}%)",
+ timing: {
+ trigger: "訊息耗時",
+ firstToken: "首個 token",
+ speedInStep: "速度(步內)",
+ effective: "有效速度",
+ output: "輸出",
+ denoising: "去噪",
+ stepsOne: "{count} 步",
+ steps: "{count} 步",
+ blocksOne: "{count} 個區塊",
+ blocks: "{count} 個區塊",
+ canvas: "畫布",
+ generation: "生成",
+ answerTokens: "回答 token",
+ prompt: "提示詞",
+ total: "總計",
+ chunks: "分塊數",
+ promptEval: "提示詞處理",
+ promptSpeed: "提示詞速度",
+ speed: "速度",
+ tokens: "token",
+ draftAccepted: "草稿接受",
+ denoisingSteps: "去噪步數",
+ blocksLabel: "區塊數",
+ cacheHits: "快取命中",
+ cacheWrites: "快取寫入",
+ stopReason: "停止原因",
+ stoppedByYou: "由您停止",
+ endOfTurn: "回合結束",
+ lengthLimit: "長度上限",
+ toolCall: "工具呼叫",
+ contentFilter: "內容篩選",
+ },
+ placement: {
+ title: "權重配置",
+ allOnCpu: "全部 {total} 層位於 CPU",
+ split: "{total} 層中有 {gpu} 層位於 GPU,{cpu} 層位於 CPU",
+ allOnGpu: "全部 {total} 層位於 GPU",
+ expertCache: "全部 {total} 層位於 GPU,所有專家位於 RAM(GPU 上有 {amount} 專家快取)",
+ expertSpillOne: "全部 {total} 層位於 GPU,{count} 層的專家權重位於 RAM",
+ expertSpill: "全部 {total} 層位於 GPU,{count} 層的專家權重位於 RAM",
+ inUnifiedMemory: "{amount} 位於統一記憶體",
+ inSharedMemory: "{amount} 位於共用記憶體",
+ inVram: "{amount} 位於 VRAM",
+ inRam: "{amount} 位於 RAM",
+ },
+ },
} satisfies DeepPartialMessageTree<typeof en>;
diff --git a/studio/frontend/tests/generation-details.test.ts b/studio/frontend/tests/generation-details.test.ts
index 76ac5e8e9f5..cd90b9e2691 100644
--- a/studio/frontend/tests/generation-details.test.ts
+++ b/studio/frontend/tests/generation-details.test.ts
@@ -32,6 +32,19 @@ const { normalizeChatGenerationChunkPayload } = await import(
const { generationChunkCountsTowardTiming } = await import(
"../src/features/chat/utils/chat-generation-recovery.ts"
);
+const { en } = await import("../src/i18n/locales/en.ts");
+const { de } = await import("../src/i18n/locales/de.ts");
+
+/** useT() over a real catalog, falling back to English like translate(), so a reworded key fails here. */
+function translator(catalog: object) {
+ const read = (tree: object, key: string) =>
+ key.split(".").reduce<unknown>((node, part) => (node as Record<string, unknown> | undefined)?.[part], tree);
+ return (key: string, values: Record<string, unknown> = {}): string => {
+ const message = (read(catalog, key) ?? read(en, key)) as string;
+ return message.replace(/\{([^}]+)\}/g, (_match, name: string) => String(values[name] ?? ""));
+ };
+}
+const t = translator(en);
const prefill = (
fields: Omit<LivePrefillProgress, "elapsedMs" | "fit"> & Partial<LivePrefillProgress>,
@@ -58,17 +71,17 @@ test("a follow-up in a long chat counts only the part the cache did not cover",
// 29,000 of 30,000 tokens reused: halfway through the new 1,000, not 98% of the whole prompt.
const followUp = prefill({ processed: 29500, total: 30000, cached: 29000, tokPerS: 1000 });
assert.equal(prefillPercent(followUp), 50);
- assert.equal(formatPrefillDetails(followUp, "en-US"), "500 / 1,000 tokens · 1,000 tok/s");
+ assert.equal(formatPrefillDetails(followUp, t, "en-US"), "500 / 1,000 tokens · 1,000 tok/s");
// llama.cpp evaluates at least one token, but a cache that covers it all is done, not NaN.
assert.equal(prefillPercent(prefill({ processed: 30000, total: 30000, cached: 30000, tokPerS: null })), 100);
});
test("the time left follows the rate so far and drops out under a second", () => {
- assert.equal(formatPrefillTimeLeft(prefill({ processed: 4118, total: 18880, cached: 0, tokPerS: 1401.2 })), "11s left");
- assert.equal(formatPrefillTimeLeft(prefill({ processed: 5000, total: 200000, cached: 0, tokPerS: 1500 })), "2m 10s left");
+ assert.equal(formatPrefillTimeLeft(prefill({ processed: 4118, total: 18880, cached: 0, tokPerS: 1401.2 }), t), "11s left");
+ assert.equal(formatPrefillTimeLeft(prefill({ processed: 5000, total: 200000, cached: 0, tokPerS: 1500 }), t), "2m 10s left");
// The first sample's rate is noise, and at the end there is nothing worth counting down.
- assert.equal(formatPrefillTimeLeft(prefill({ processed: 22, total: 18844, cached: 0, tokPerS: 242 })), null);
- assert.equal(formatPrefillTimeLeft(prefill({ processed: 18500, total: 18880, cached: 0, tokPerS: 1400 })), null);
+ assert.equal(formatPrefillTimeLeft(prefill({ processed: 22, total: 18844, cached: 0, tokPerS: 242 }), t), null);
+ assert.equal(formatPrefillTimeLeft(prefill({ processed: 18500, total: 18880, cached: 0, tokPerS: 1400 }), t), null);
});
// llama-server's own [processed, time_ms] samples for two prompts. The first, 30,382 tokens through
@@ -184,10 +197,10 @@ test("decode frames keep draft acceptance only when both counts are there", () =
test("the prefill details give the rate when there is one", () => {
assert.equal(
- formatPrefillDetails(prefill({ processed: 4118, total: 18880, cached: 0, tokPerS: 1401.2 }), "en-US"),
+ formatPrefillDetails(prefill({ processed: 4118, total: 18880, cached: 0, tokPerS: 1401.2 }), t, "en-US"),
"4,118 / 18,880 tokens · 1,401 tok/s",
);
- assert.equal(formatPrefillDetails(prefill({ processed: 900, total: null, cached: null, tokPerS: null }), "en-US"), "900 tokens");
+ assert.equal(formatPrefillDetails(prefill({ processed: 900, total: null, cached: null, tokPerS: null }), t, "en-US"), "900 tokens");
});
test("the prefill label shows only for a pass expected to keep it up, then stays for that pass", () => {
@@ -219,16 +232,80 @@ test("the prefill label shows only for a pass expected to keep it up, then stays
test("the live line writes every number in the reader's locale", () => {
const decode = { tokens: 3060, tokPerS: 58.25, draftN: 5030, draftAccepted: 1791 };
assert.equal(
- formatLiveDecode(decode, "en-US"),
+ formatLiveDecode(decode, t, "en-US"),
"3,060 tokens · 58.3 tok/s · draft 1,791/5,030 accepted (36%)",
);
// A locale that groups with "." must not get "." as the rate's decimal point too.
assert.equal(
- formatLiveDecode(decode, "de-DE"),
+ formatLiveDecode(decode, t, "de-DE"),
"3.060 tokens · 58,3 tok/s · draft 1.791/5.030 accepted (36%)",
);
});
+test("the live figures and the placement read in the reader's language", () => {
+ const tDe = translator(de);
+ let previous: { prefill: LivePrefillProgress; prefillShown: boolean; decode: null } | undefined;
+ for (const [processed, elapsedMs] of LONG_PROMPTS[0].samples) {
+ if (processed > 16385) break;
+ const current = withPrefillFit(previous, prefill({
+ processed, total: 30382, cached: 0,
+ tokPerS: elapsedMs > 0 ? processed * 1000 / elapsedMs : null,
+ elapsedMs,
+ }));
+ previous = { prefill: current, prefillShown: false, decode: null };
+ }
+ assert.equal(formatPrefillTimeLeft(previous!.prefill, tDe), "noch 20s");
+ const g = de.generationStats;
+ const decode = { tokens: 3060, tokPerS: 58.25, draftN: 5030, draftAccepted: 1791 };
+ assert.equal(
+ formatLiveDecode(decode, tDe, "de-DE"),
+ [
+ g.tokens.replace("{count}", "3.060"),
+ "58,3 tok/s",
+ g.draftAccepted.replace("{accepted}", "1.791").replace("{drafted}", "5.030").replace("{percent}", "36"),
+ ].join(" · "),
+ );
+ assert.equal(formatLiveDecode({ ...decode, tokens: 1, tokPerS: null, draftN: null }, tDe, "de-DE"), g.tokensOne.replace("{count}", "1"));
+ const prompt = prefill({ processed: 4118, total: 18880, cached: 0, tokPerS: 1401.2 });
+ assert.equal(formatPrefillTimeLeft(prompt, tDe), g.timeLeft.replace("{eta}", "11s"));
+ assert.equal(
+ formatPrefillDetails(prompt, tDe, "de-DE"),
+ `${g.tokensOf.replace("{done}", "4.118").replace("{total}", "18.880")} · 1.401 tok/s`,
+ );
+ const metal = { offloaded: 33, total: 33, overflowLayers: null, gpuWeightsMib: 2603.5, hostWeightsMib: 497.31, expertCacheMib: null };
+ assert.deepEqual(describePlacement(metal, tDe, "unified"), {
+ layers: g.placement.allOnGpu.replace("{total}", "33"),
+ memory: g.placement.inUnifiedMemory.replace("{amount}", "3.0 GiB"),
+ });
+ assert.notEqual(g.placement.allOnGpu, en.generationStats.placement.allOnGpu);
+});
+
+test("languages whose noun form depends on the number put counts after a label", async () => {
+ // A one/other pair cannot say "2 токена" and "5 токенов" in Russian, so these read "токены: 21" instead.
+ for (const file of ["ru", "uk", "pl", "cs", "ar"]) {
+ const catalog = (await import(`../src/i18n/locales/${file}.ts`)) as Record<string, { generationStats: typeof en.generationStats }>;
+ const g = Object.values(catalog)[0].generationStats;
+ for (const template of [g.tokens, g.tokensOf, g.timing.steps, g.timing.blocks, g.placement.expertSpill]) {
+ assert.doesNotMatch(template, /\{(count|total)\} \p{L}/u, `${file}: ${template}`);
+ }
+ }
+});
+
+test("the prompt line, live line, timing tooltip and placement row hold no English of their own", () => {
+ const thread = readSrc("components/assistant-ui/thread.tsx");
+ const status = thread.slice(thread.indexOf("const GeneratingStatus"), thread.indexOf("// Placeholder when stop fires"));
+ assert.doesNotMatch(status, />\s*(Working\.\.\.|Processing prompt)/);
+ assert.match(status, /t\("generationStats\.processingPrompt", \{ percent \}\)/);
+ const timing = readSrc("components/assistant-ui/message-timing.tsx");
+ assert.doesNotMatch(timing, /text-muted-foreground">[A-Za-z]/, "a tooltip row label left as a literal");
+ assert.doesNotMatch(timing, /"(End of turn|Length limit|Tool call|Content filter|Stopped by you|Message timing)"/);
+ const row = readSrc("features/model-picker/components/loaded-placement-row.tsx");
+ assert.doesNotMatch(row, /Weight placement/);
+ const progress = readSrc("features/chat/lib/generation-progress.ts");
+ // The sentences main built around localized numbers: "} tokens`", "} left`", "draft ${…} accepted (".
+ assert.doesNotMatch(progress, /\} tokens`|\} left`|`draft \$\{|"token" : "tokens"/);
+});
+
test("both live lines wrap instead of clipping on a narrow chat column", () => {
// .generating-glint is nowrap and overflow: hidden; at 390 px the prefill details and the decode line ran past the edge.
const thread = readSrc("components/assistant-ui/thread.tsx");
@@ -294,20 +371,20 @@ test("live progress is per thread, transient, and re-keyed with the thread's oth
test("the placement readout names expert weights and gives both memory figures", () => {
// gemma-4-26B-A4B Q8_0 on a 3090: --fit's verdict "31 layers ( 8 overflowing)", the first of them GATE-only.
assert.deepEqual(
- describePlacement({ offloaded: 31, total: 31, overflowLayers: 8, gpuWeightsMib: 20709.15, hostWeightsMib: 5639.4, expertCacheMib: null }),
+ describePlacement({ offloaded: 31, total: 31, overflowLayers: 8, gpuWeightsMib: 20709.15, hostWeightsMib: 5639.4, expertCacheMib: null }, t),
{ layers: "All 31 layers on the GPU, expert weights of 8 layers in RAM", memory: "20 GiB in VRAM · 5.5 GiB in RAM" },
);
assert.equal(
- describePlacement({ offloaded: 31, total: 31, overflowLayers: 1, gpuWeightsMib: 20000, hostWeightsMib: 900, expertCacheMib: null })
+ describePlacement({ offloaded: 31, total: 31, overflowLayers: 1, gpuWeightsMib: 20000, hostWeightsMib: 900, expertCacheMib: null }, t)
.layers,
"All 31 layers on the GPU, expert weights of 1 layer in RAM",
);
assert.deepEqual(
- describePlacement({ offloaded: 41, total: 41, overflowLayers: 0, gpuWeightsMib: 8161.75, hostWeightsMib: 417.3, expertCacheMib: null }),
+ describePlacement({ offloaded: 41, total: 41, overflowLayers: 0, gpuWeightsMib: 8161.75, hostWeightsMib: 417.3, expertCacheMib: null }, t),
{ layers: "All 41 layers on the GPU", memory: "8.0 GiB in VRAM · 0.4 GiB in RAM" },
);
assert.deepEqual(
- describePlacement({ offloaded: 20, total: 31, overflowLayers: null, gpuWeightsMib: 12288, hostWeightsMib: 5120, expertCacheMib: null }),
+ describePlacement({ offloaded: 20, total: 31, overflowLayers: null, gpuWeightsMib: 12288, hostWeightsMib: 5120, expertCacheMib: null }, t),
{ layers: "20 of 31 layers on the GPU, 11 on the CPU", memory: "12 GiB in VRAM · 5.0 GiB in RAM" },
);
// A Manual load reports its expert offload as n_cpu_moe, not through --fit.
@@ -328,7 +405,7 @@ test("the placement readout names expert weights and gives both memory figures",
test("a GPU build that fell back to the CPU is not reported as GPU layers", () => {
// CUDA build, no usable device: llama.cpp still logs "offloaded 29/29 layers to GPU", buffers AMX + CPU_Mapped.
assert.deepEqual(
- describePlacement({ offloaded: 29, total: 29, overflowLayers: null, gpuWeightsMib: 0, hostWeightsMib: 499.8, expertCacheMib: null }),
+ describePlacement({ offloaded: 29, total: 29, overflowLayers: null, gpuWeightsMib: 0, hostWeightsMib: 499.8, expertCacheMib: null }, t),
{ layers: "All 29 layers on the CPU", memory: "0.5 GiB in RAM" },
);
});
@@ -345,7 +422,7 @@ test("--moe-cache-mib auto reads as every expert in RAM with a GPU cache", () =>
expert_cache_mib: 6144,
});
assert.ok(placement);
- assert.deepEqual(describePlacement(placement), {
+ assert.deepEqual(describePlacement(placement, t), {
layers: "All 31 layers on the GPU, all experts in RAM (6.0 GiB expert cache on the GPU)",
memory: "3.8 GiB in VRAM · 22 GiB in RAM",
});
@@ -354,14 +431,14 @@ test("--moe-cache-mib auto reads as every expert in RAM with a GPU cache", () =>
test("one memory pool gets one figure, and amounts under 0.1 GiB read in MiB", () => {
// Qwen3.5-4B Q4_K_M under Metal on an M1 Mac mini: MTL0 2,603.5 MiB and CPU 497.3 MiB, one pool.
const metal = { offloaded: 33, total: 33, overflowLayers: null, gpuWeightsMib: 2603.5, hostWeightsMib: 497.31, expertCacheMib: null };
- assert.deepEqual(describePlacement(metal, "unified memory"), {
+ assert.deepEqual(describePlacement(metal, t, "unified"), {
layers: "All 33 layers on the GPU",
memory: "3.0 GiB in unified memory",
});
- assert.equal(describePlacement(metal).memory, "2.5 GiB in VRAM · 0.5 GiB in RAM");
+ assert.equal(describePlacement(metal, t).memory, "2.5 GiB in VRAM · 0.5 GiB in RAM");
// A small model's token embeddings would otherwise print as "0.0 GiB in RAM".
assert.equal(
- describePlacement({ offloaded: 31, total: 31, overflowLayers: 0, gpuWeightsMib: 112.6, hostWeightsMib: 28.7, expertCacheMib: null }).memory,
+ describePlacement({ offloaded: 31, total: 31, overflowLayers: 0, gpuWeightsMib: 112.6, hostWeightsMib: 28.7, expertCacheMib: null }, t).memory,
"0.1 GiB in VRAM · 29 MiB in RAM",
);
});