diff --git a/README.md b/README.md index b25a917d7..6f6044b30 100644 --- a/README.md +++ b/README.md @@ -76,31 +76,31 @@ Jan is an open-source ChatGPT alternative that runs 100% offline on your compute Experimental (Nightly Build) - + jan.exe - + Intel - + M1/M2 - + jan.deb - + jan.AppImage diff --git a/extensions/monitoring-extension/src/node/index.ts b/extensions/monitoring-extension/src/node/index.ts index 1d65704de..25f151112 100644 --- a/extensions/monitoring-extension/src/node/index.ts +++ b/extensions/monitoring-extension/src/node/index.ts @@ -2,17 +2,17 @@ import { GpuSetting, GpuSettingInfo, ResourceInfo } from '@janhq/core' import { getJanDataFolderPath, log } from '@janhq/core/node' import { mem, cpu } from 'node-os-utils' import { exec } from 'child_process' -import { writeFileSync, existsSync, readFileSync } from 'fs' +import { writeFileSync, existsSync, readFileSync, mkdirSync } from 'fs' import path from 'path' +/** + * Path to the settings directory + **/ +export const SETTINGS_DIR = path.join(getJanDataFolderPath(), 'settings') /** * Path to the settings file **/ -export const GPU_INFO_FILE = path.join( - getJanDataFolderPath(), - 'settings', - 'settings.json' -) +export const GPU_INFO_FILE = path.join(SETTINGS_DIR, 'settings.json') /** * Default GPU settings @@ -136,6 +136,11 @@ export const updateNvidiaInfo = async () => { try { JSON.parse(readFileSync(GPU_INFO_FILE, 'utf-8')) } catch (error) { + if (!existsSync(SETTINGS_DIR)) { + mkdirSync(SETTINGS_DIR, { + recursive: true, + }) + } writeFileSync(GPU_INFO_FILE, JSON.stringify(DEFAULT_SETTINGS, null, 2)) } diff --git a/extensions/tensorrt-llm-extension/models.json b/extensions/tensorrt-llm-extension/models.json index bc6a78256..31bb11a9e 100644 --- a/extensions/tensorrt-llm-extension/models.json +++ b/extensions/tensorrt-llm-extension/models.json @@ -33,10 +33,10 @@ "description": "LlamaCorn is a refined version of TinyLlama-1.1B, optimized for conversational quality, running on consumer devices through TensorRT-LLM", "format": "TensorRT-LLM", "settings": { - "ctx_len": 2048 + "ctx_len": 2048, + "text_model": false }, "parameters": { - "stream": true, "max_tokens": 4096 }, "metadata": { diff --git a/extensions/tensorrt-llm-extension/src/index.ts b/extensions/tensorrt-llm-extension/src/index.ts index 076951c3f..e3014b447 100644 --- a/extensions/tensorrt-llm-extension/src/index.ts +++ b/extensions/tensorrt-llm-extension/src/index.ts @@ -19,6 +19,7 @@ import { systemInformations, LocalOAIEngine, fs, + MessageRequest, } from '@janhq/core' import models from '../models.json' @@ -144,4 +145,10 @@ export default class TensorRTLLMExtension extends LocalOAIEngine { ) return Promise.resolve() } + + inference(data: MessageRequest): void { + // TensorRT LLM Extension supports streaming only + if (data.model) data.model.parameters.stream = true + super.inference(data) + } } diff --git a/web/screens/Chat/ChatInput/index.tsx b/web/screens/Chat/ChatInput/index.tsx index c90a12cd2..8707e8bcd 100644 --- a/web/screens/Chat/ChatInput/index.tsx +++ b/web/screens/Chat/ChatInput/index.tsx @@ -244,16 +244,13 @@ const ChatInput: React.FC = () => {
  • { if ( - !activeThread?.assistants[0].model.settings - .vision_model || activeThread?.assistants[0].model.settings .text_model !== false ) { diff --git a/web/screens/Settings/CoreExtensions/TensorRtExtensionItem.tsx b/web/screens/Settings/CoreExtensions/TensorRtExtensionItem.tsx index e4c8ae04f..b363f79be 100644 --- a/web/screens/Settings/CoreExtensions/TensorRtExtensionItem.tsx +++ b/web/screens/Settings/CoreExtensions/TensorRtExtensionItem.tsx @@ -123,6 +123,7 @@ const TensorRtExtensionItem: React.FC = ({ item }) => { {item.description}

    + {(!compatibility || compatibility['platform']?.includes(PLATFORM)) && isGpuSupported ? (
    @@ -185,15 +186,14 @@ const InstallStateIndicator: React.FC = ({ onInstallClick, onCancelClick, }) => { - // TODO: NamH support dark mode for this if (installProgress !== -1) { const progress = installProgress * 100 return ( -
    +
    -
    +
    {progress.toFixed(0)}%