From ee2ccc3e9b037e65fe14d1d6d156785e72c76bf9 Mon Sep 17 00:00:00 2001
From: linonetwo <linonetwo012@gmail.com>
Date: Wed, 10 Apr 2024 14:14:23 +0800
Subject: [PATCH] feat: decode token

---
 localization/locales/zh_CN/translation.json      |  3 ++-
 src/services/languageModel/index.ts              |  2 +-
 src/services/languageModel/llmWorker/llamaCpp.ts | 11 +++++++++--
 3 files changed, 12 insertions(+), 4 deletions(-)

diff --git a/localization/locales/zh_CN/translation.json b/localization/locales/zh_CN/translation.json
index 6000d059..e1a6d2ec 100644
--- a/localization/locales/zh_CN/translation.json
+++ b/localization/locales/zh_CN/translation.json
@@ -479,7 +479,8 @@
     "ModelNotExist": "找不到模型",
     "ModelNotExistDescription": "尝试使用你给的这个路径加载模型，但在这个位置其实没有所需要的模型",
     "OpenThisPath": "打开该位置",
-    "GenerationTimeout": "模型生成超时，已中止。"
+    "GenerationTimeout": "模型生成超时，已中止。",
+    "ModalDisposed": "模型已卸载，生成中止，需要重新加载模型。"
   },
   "Help": {
     "Alternatives": "其它源",
diff --git a/src/services/languageModel/index.ts b/src/services/languageModel/index.ts
index 9e88f5c9..274b94b7 100644
--- a/src/services/languageModel/index.ts
+++ b/src/services/languageModel/index.ts
@@ -120,7 +120,7 @@ export class LanguageModel implements ILanguageModelService {
           return;
         }
         let observable;
-        const texts = { timeout: i18n.t('LanguageModel.GenerationTimeout') };
+        const texts = { timeout: i18n.t('LanguageModel.GenerationTimeout'), disposed: i18n.t('LanguageModel.ModalDisposed') };
         switch (runner) {
           case LanguageModelRunner.llamaCpp: {
             observable = worker.runLLama({ completionOptions, loadConfig: { ...config, modelPath }, conversationID }, texts);
diff --git a/src/services/languageModel/llmWorker/llamaCpp.ts b/src/services/languageModel/llmWorker/llamaCpp.ts
index b4fd2568..776de685 100644
--- a/src/services/languageModel/llmWorker/llamaCpp.ts
+++ b/src/services/languageModel/llmWorker/llamaCpp.ts
@@ -59,7 +59,7 @@ export function runLLama(
     conversationID: IRunLLAmaOptions['id'];
     loadConfig: IRunLLAmaOptions['loadConfig'];
   },
-  texts: { timeout: string },
+  texts: { disposed: string; timeout: string },
 ): Observable<ILanguageModelWorkerResponse> {
   const { conversationID, completionOptions, loadConfig } = options;
 
@@ -105,8 +105,15 @@ export function runLLama(
           ...completionOptions,
           signal: abortController.signal,
           onToken: (tokens) => {
+            if (modalInstance === undefined) {
+              abortController.abort();
+              runnerAbortControllers.delete(conversationID);
+              subscriber.next({ type: 'result', token: texts.timeout, id: conversationID });
+              subscriber.complete();
+              return;
+            }
             updateTimeout();
-            subscriber.next({ type: 'result', token, id: conversationID });
+            subscriber.next({ type: 'result', token: modalInstance.detokenize(tokens), id: conversationID });
           },
         });
         // completed