Download & Install ollama 一個可以運行大模型的工具
Download web address: https://ollama.com/
Install it step by step after download success.
Download a AI model (name: llama3) & run it in command window.
招待指定大模型的格式: ollama run 大模型名稱
e.g. ollama run llama3
其他可執行的大模型參考: https://ollama.com/library
這行命令既是下載, 也是運行(如果下載好了)
退出: /bye
ollama后臺執行
ollama server
啟動一個WebUI 操作界面
docker run -d -p 3001:8088/tcp --add-host=host.docker.internal:host-gateway -v open-webui:/app/backend/data --name open-webui --restart always ghcr.io/open-webui/open-webui:main
訪問WebUI操作界面 ???
小技巧:
- 如果要讀取網站鏈接, 要在鏈接前加一個‘#’.
- 在 "設置" > "通用" > "系統提示詞" 里面輸入 "不能說英文, 必須用中文回復用戶", 這樣兒以后回答的內容都會以中文回答.
安裝本地知識庫 --- Anything LLM
下載 & 安裝
下載地址: https://useanything.com/download
選擇(本地-ollama)模型 和 向量數據庫
只要ollama服務在后臺啟用著, AnythingLLM就會識別.
配置&指定本地模型
Ollama模型: llama3
- Embedding Providers 嵌入提供器(會把上傳的文件轉為低維向量數據) 我們使用默認的
AnythingLLM Embedder -
Vector Database 連接向量數據庫
-
前置設置: 輸入 workspace 名字, 用于指定知識庫服務于那個workspace
上傳知識庫
注意: 上傳的文件名必須是英文, 中文文件名上傳會一直讀取的狀態.
問: Mackbook Pro多少錢?
How to create a standard SD order in SAP GUI?
不能說英文, 必須用中文回復用戶
最好是上傳文本性文檔, 以便知識庫的質量高一些.
深度調整
ollama - Models
Here are the AI models categorized and described in Chinese using Markdown format:
代碼生成模型 (Code Generation Models)
Code 34B: 17.8K Pulls 16 Tags Updated 8 months ago
Codegeex4: A versatile model for AI software development scenarios, including code completion.
- Magicoder ??: A family of 7B parameter models trained on 75K synthetic instruction data using OSS-Instruct, a novel approach to enlightening LLMs with open-source code snippets.
Chat模型 (Chat Models)
Alfred: A robust conversational model designed to be used for both chat and instruct use cases.
Everythinglm: Uncensored Llama2 based model with support for a 16K context window.
Internlm2: A 7B parameter model tailored for practical scenarios with outstanding reasoning capability.
Megadolphin: MegaDolphin-2.2-120b is a transformation of Dolphin-2.2-70b created by interleaving the model with itself.
Mistrallite: MistralLite is a fine-tuned model based on Mistral with enhanced capabilities of processing long contexts.
Nexus Raven: A 13B instruction tuned model for function calling tasks.
Notus: A 7B chat model fine-tuned with high-quality data and based on Zephyr.
Open-orca-platypus2: Merge of the Open Orca OpenChat model and the Garage-bAInd Platypus 2 model. Designed for chat and code generation.
工具模型 (Tool Models)
DBRX: An open, general-purpose LLM created by Databricks.
Firefunction-v2: An open weights function calling model based on Llama 3, competitive with GPT-4o function calling capabilities.
Llama3-groq-tool-use: A series of models from Groq that represent a significant advancement in open-source AI capabilities for tool use/function calling.
MathΣtral: A 7B model designedx for math reasoning and scientific discovery by Mistral AI.
Nuextract: A 3.8B model fine-tuned on a private high-quality synthetic dataset for information extraction, based on Phi-3.
語言模型 (Language Models)
Goliath: A language model created by combining two fine-tuned Llama 2 70B models into one.
Notux: A top-performing mixture of experts model, fine-tuned with high-quality data.
Stablelm-zephyr: A lightweight chat model allowing accurate, and responsive output without requiring high-end hardware.
其他模型 (Other Models)
Falcon2: An 11B parameters causal decoder-only model built by TII and trained over 5T tokens.
Llama3.1: A new state-of-the-art model.
Wizard Vicuna: A 13B parameter model based on Llama 2 trained by MelodysDreamj.
