AI知识库

53AI知识库

学习大模型的前沿技术与行业应用场景


在个人电脑一键运行谷歌最新 Gemma-2-9B 大模型
发布日期:2024-08-24 22:48:43 浏览次数: 1615


谷歌最近发布了9B和27B大小的 Gemma 2模型[1],这是其 Gemma 模型系列的最新型号。根据其技术报告,未来几天将开源一个 Gemma-2-2b 模型。技术报告还显示,Gemma-2-9B模型在多个基准测试中的表现超过了 Mistral-7B、Llama-3-8B和 Gemma 1.5模型。

如果想一键在你的计算机上运行 Gemma-9b-Chat,可以在终端中运行以下命令

bash <(curl -sSfL 'https://raw.githubusercontent.com/LlamaEdge/LlamaEdge/main/run-llm.sh') —model gemma-2-9b-it

本文将以 Gemma-2-9B 为例,手把手教你轻松

  • 在自己的设备上运行 Gemma-2-9B on your own device
  • 为 Gemma-2-9B 创建一个 API server

我们将使用 LlamaEdge[2] (Rust + Wasm技术栈)开发并部署该模型的应用。无需安装复杂的 Python 包或 C++ 工具链!了解我们为什么选择 Rust + Wasm[3]

在你自己的设备上运行 Gemma-2-9B

第一步:用下面的命令行安装 WasmEdge[4].

curl -sSf https://raw.githubusercontent.com/WasmEdge/WasmEdge/master/utils/install_v2.sh | bash -s -- -v 0.13.5

第二步:下载 Gemma-2-9B-it model GGUF 模型[5]文件。模型有6.40 GB,所以下载可能需要一定时间

curl -LO https://huggingface.co/second-state/gemma-2-9b-it-GGUF/resolve/main/gemma-2-9b-it-Q5_K_M.gguf

第三步:下载一个跨平台的可移植 Wasm 文件,是个聊天应用。该应用让你能在命令行中与模型聊天。该应用的 Rust 源代码戳这里[6]

curl -LO https://github.com/LlamaEdge/LlamaEdge/releases/latest/download/llama-chat.wasm

就是这样。可以通过输入以下命令在终端与模型聊天。

wasmedge --dir .:. --nn-preload default:GGML:AUTO:gemma-2-9b-it-Q5_K_M.gguf llama-chat.wasm -p gemma-instruct -c 4096

这个可移植的Wasm应用会自动利用设备上的硬件加速器(例如GPU)。

[You]:
Tell me what is gemma?

[Bot]:
Gemma is an open-weights AI assistant developed by Google DeepMind.

Here are some key things to know about Gemma:

* **Open Weights:** This means the underlying code and model weights are freely accessible to the public. Anyone can inspect, modify, or build upon Gemma.
* **Text-Only:** Gemma is designed primarily for text-based interactions. It can understand and generate human-like text but doesn't have capabilities like image recognition or voice synthesis.
* **Focus on Research & Development:** Gemma is intended as a tool for researchers and developers to explore and experiment with different AI models and techniques.

**Why is Gemma significant?**

* **Transparency and Collaboration:** Open-weights nature promotes transparency in AI development and encourages collaboration among researchers worldwide.
* **Customization:** Users can adapt Gemma to their specific needs by fine-tuning it on their own datasets.
* **Accessibility:**  Makes advanced AI technology accessible to a wider range of individuals and organizations, beyond those with large resources.

**Keep in mind:**

* Gemma is still under development and may have limitations compared to more mature, closed-source models.
* Its capabilities are primarily focused on text processing.

Gemma-2-9b-it[7]创建一个兼容OpenAI的 API server

一个兼容 OpenAI 的API 使得 Llama-3-8B-Chinese 能够与不同的开发框架和工具无缝集成,比如 flows.network[8],  LangChain and LlamaIndex等等,提供更广泛的应用可能。大家也可以参考其代码自己写自己的API服务器或者其它大模型应用。想要启动 API 服务,请按以下步骤操作:下载这个 API 服务器应用。它是一个跨平台的可移植 Wasm 应用,可以在各种 CPU 和 GPU 设备上运行。

curl -LO https://github.com/LlamaEdge/LlamaEdge/releases/latest/download/llama-api-server.wasm

然后,下载聊天机器人 Web UI,从而通过聊天机器人 UI 与模型进行交互。

curl -LO https://github.com/LlamaEdge/chatbot-ui/releases/latest/download/chatbot-ui.tar.gz
tar xzf chatbot-ui.tar.gz
rm chatbot-ui.tar.gz

接下来,使用以下命令行启动模型的 API 服务器。然后,打开浏览器访问 http://localhost:8080[9] 开始聊天!

wasmedge --dir .:. --nn-preload default:GGML:AUTO:gemma-2b-it-Q5_K_M.gguf llama-api-server.wasm -p gemma-instruct -c 4096

另外打开一个终端窗口, 可以使用 curl 与 API 服务器进行交互。

curl -X POST http://localhost:8080/v1/chat/completions \
  -H 'accept:application/json' \
  -H 'Content-Type: application/json' \
  -d '{"messages":[{"role":"system", "content": "You are a sentient, superintelligent artificial general intelligence, here to teach and assist me."}, {"role":"user", "content": "Write a short story about Goku discovering kirby has teamed up with Majin Buu to destroy the world."}], "model":"Gemma-2b-it"}' 

就是这样啦。WasmEdge 是运行 LLM 应用最简单、最快、最安全的方式[10]。快来试试看吧!


53AI,企业落地应用大模型首选服务商

产品:大模型应用平台+智能体定制开发+落地咨询服务

承诺:先做场景POC验证,看到效果再签署服务协议。零风险落地应用大模型,已交付160+中大型企业

联系我们

售前咨询
186 6662 7370
预约演示
185 8882 0121

微信扫码

与创始人交个朋友

回到顶部

 
扫码咨询