happyz
/
llama.cpp
mirror of https://github.com/ggerganov/llama.cpp.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

webui: fix prompt progress ETA calculation (#18468) 

* webui: fix prompt progress ETA calculation

* handle case done === 0
This commit is contained in:
Xuan-Son Nguyen 2025-12-29 21:42:11 +01:00 committed by GitHub
parent c9a3b40d65
commit 51a48720b8
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
3 changed files with 23 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

BIN
tools/server/public/index.html.gz
View File

Binary file not shown.

16
tools/server/webui/src/lib/hooks/use-processing-state.svelte.ts
View File

@ -6,6 +6,7 @@ export interface LiveProcessingStats {
totalTokens: number;
timeMs: number;
tokensPerSecond: number;
etaSecs?: number;
}
export interface LiveGenerationStats {
@ -82,6 +83,15 @@ export function useProcessingState(): UseProcessingStateReturn {
}
});
function getETASecs(done: number, total: number, elapsedMs: number): number | undefined {
const elapsedSecs = elapsedMs / 1000;
const progressETASecs =
done === 0 || elapsedSecs < 0.5
? undefined // can be the case for the 0% progress report
: elapsedSecs * (total / done - 1);
return progressETASecs;
}
function startMonitoring(): void {
if (isMonitoring) return;
isMonitoring = true;
@ -178,6 +188,12 @@ export function useProcessingState(): UseProcessingStateReturn {
const actualProcessed = processed - cache;
const actualTotal = total - cache;
const percent = Math.round((actualProcessed / actualTotal) * 100);
const eta = getETASecs(actualProcessed, actualTotal, processingState.promptProgress.time_ms);
if (eta !== undefined) {
const etaSecs = Math.ceil(eta);
return `Processing ${percent}% (ETA: ${etaSecs}s)`;
}
return `Processing ${percent}%`;
}

8
tools/server/webui/src/lib/stores/chat.svelte.ts
View File

@ -303,11 +303,17 @@ class ChatStore {
const currentConfig = config();
const outputTokensMax = currentConfig.max_tokens || -1;
// Note: for timings data, the n_prompt does NOT include cache tokens
const contextUsed = promptTokens + cacheTokens + predictedTokens;
const outputTokensUsed = predictedTokens;
// Note: for prompt progress, the "processed" DOES include cache tokens
// we need to exclude them to get the real prompt tokens processed count
const progressCache = promptProgress?.cache || 0;
const progressActualDone = (promptProgress?.processed ?? 0) - progressCache;
const progressActualTotal = (promptProgress?.total ?? 0) - progressCache;
const progressPercent = promptProgress
? Math.round((promptProgress.processed / promptProgress.total) * 100)
? Math.round((progressActualDone / progressActualTotal) * 100)
: undefined;
return {