【Hacker News搬运】Llama 3.2：通过开放、可定制的模型彻底改变边缘人工智能和视觉
-
Title: Llama 3.2: Revolutionizing edge AI and vision with open, customizable models
Llama 3.2：通过开放、可定制的模型彻底改变边缘人工智能和视觉
Text:
Url: https://ai.meta.com/blog/llama-3-2-connect-2024-vision-edge-mobile-devices/?_fb_noscript=1
很抱歉，我无法直接访问外部链接来获取网页内容。不过，我可以告诉你如何使用 JinaReader 这样的工具来抓取和分析网页内容，以及如何处理非中文内容。 以下是一个使用 Python 和 JinaReader（假设这是一个可以用于内容抓取和分析的库）的基本步骤： 1. **安装 JinaReader**（假设你可以安装库）: ```bash pip install jinareader-
使用 JinaReader 抓取内容:
from jinareader import JinaReader reader = JinaReader() url = "https://ai.meta.com/blog/llama-3-2-connect-2024-vision-edge-mobile-devices/?_fb_noscript=1" content = reader.get_content(url) print(content) -
分析内容（假设你可以分析内容）:
# 假设有一些分析函数 summary = analyze_content(content) print(summary) -
处理非中文内容:
如果抓取到的内容不是中文，你可以使用翻译API来将其翻译成中文。以下是一个使用 Google Translate API 的例子（请注意，你需要有有效的API密钥）:from googletrans import Translator translator = Translator() translated_content = translator.translate(content, dest='zh-cn').text print(translated_content) -
总结内容:
在得到翻译后的中文内容后，你可以使用自然语言处理工具来生成总结。
请注意，以上代码仅为示例，实际使用时你需要根据实际情况调整库的调用方式和参数。
如果你需要我帮助你分析一个网页的具体内容，你可以将网页的HTML内容复制粘贴到代码中，我可以帮你分析并总结。
## Post by: nmwnmw ### Comments: **simonw**: I'm absolutely amazed at how capable the new 1B model is, considering it's just a 1.3GB download (for the Ollama GGUF version).<p>I tried running a full codebase through it (since it can handle 128,000 tokens) and asking it to summarize the code - it did a surprisingly decent job, incomplete but still unbelievable for a model that tiny: <a href="https://gist.github.com/simonw/64c5f5b111fe473999144932bef4218b" rel="nofollow">https://gist.github.com/simonw/64c5f5b111fe473999144932bef42...</a><p>More of my notes here: <a href="https://simonwillison.net/2024/Sep/25/llama-32/" rel="nofollow">https://simonwillison.net/2024/Sep/25/llama-32/</a><p>I've been trying out the larger image models to using the versions hosted on <a href="https://lmarena.ai/" rel="nofollow">https://lmarena.ai/</a> - navigate to "Direct Chat" and you can select them from the dropdown and upload images to run prompts. > **simonw**: 我；考虑到新的1B型号的性能，我感到非常惊讶；这只是1.3GB的下载（适用于Ollama GGUF版本）<p> 我尝试通过它运行一个完整的代码库（因为它可以处理128000个令牌），并要求它总结代码——它做得非常好，虽然不完整，但对于一个如此小的模型来说仍然令人难以置信：<a href=“https:#x2F；gist.github.com#x2F ; simonw#x2F 64c5f5b111fe47399144932bef4218b”rel=“nofollow”>https:/；github.com；simonw；64c5f5b111fe473999144932bef42…</a><p>我在这里的更多笔记：<a href=“https:”simonwillison.net“2024年9月25日”llama-32“rel=”nofollow“>https:”/；simonwillison.net；2024年；9月；25°F；llama-32</a> <p>我；我一直在尝试使用更大的图像模型，以使用<a href=“https:”lmarena.ai“rel=”nofollow“>https:”/；lmarena.ai</a> -导航到"；直接聊天"；您可以从下拉列表中选择它们并上传图像以运行提示。 **opdahl**: I'm blown away with just how open the Llama team at Meta is. It is nice to see that they are not only giving access to the models, but they at the same time are open about how they built them. I don't know how the future is going to go in the terms of models, but I sure am grateful that Meta has taken this position, and are pushing more openness. > **opdahl**: 我；我对Meta的Llama团队的开放程度感到震惊。很高兴看到他们不仅提供了对模型的访问权限，同时也对如何构建模型持开放态度。我不知道；我不知道未来在模型方面会如何发展，但我很感激Meta采取了这一立场，并正在推动更多的开放。 **a_wild_dandan**: "The Llama jumped over the ______!" (Fence? River? Wall? Synagogue?)<p>With 1-hot encoding, the answer is "wall", with 100% probability. Oh, you gave plausibility to "fence" too? WRONG! ENJOY MORE PENALTY, SCRUB!<p>I believe this unforgiving dynamic is why model distillation works well. The original teacher model had to learn via the "hot or cold" game on <i>text</i> answers. But when the child instead imitates the teacher's predictions, it learns <i>semantically rich</i> answers. That strikes me as vastly more compute-efficient. So to me, it makes sense why these Llama 3.2 edge models punch so far above their weight(s). But it still blows my mind thinking how far models have advanced from a year or two ago. Kudos to Meta for these releases. > **a_wild_dandan**: &“；骆驼跳过了______&“；（栅栏？河？墙？犹太教堂？）<p>使用1点编码，答案是"；“墙”；，概率为100%。哦，你为"；围栏"；也？错了！享受更多的惩罚，擦洗<p> 我相信这种无情的动态是模型蒸馏效果良好的原因。最初的教师模式必须通过"；热或冷”；游戏对<i>文本</i>的回答。但当孩子转而模仿老师时；在进行预测时，它会学习语义丰富的<i>答案。这让我觉得计算效率要高得多。所以对我来说，这些Llama 3.2边缘模型的冲击力远远超过它们的重量是有道理的。但想到模型比一两年前进步了多远，我仍然很震惊。Meta的这些发布值得称赞。 **freedomben**: If anyone else is looking for the bigger models on ollama and wondering where they are, the Ollama blog post answered that for me. The are "coming soon" so they just aren't ready quite yet[1]. I was a little worried when I couldn't find them but sounds like we just need to be patient.<p>[1]: <a href="https://ollama.com/blog/llama3.2">https://ollama.com/blog/llama3.2</a> > **freedomben**: 如果有人正在寻找OLAMA上更大的模型，并想知道它们在哪里，ollama博客文章为我回答了这个问题；即将到来"；所以它们只是不是；我还没准备好[1]。当我无法做到时，我有点担心；我找不到他们，但听起来我们只需要有耐心<p> [1]：<a href=“https:/；ollama.comG；blogM；llama3.2”>https:"/；ollama.com；博客/；骆驼3.2</a> **dhbradshaw**: Tried out 3B on ollama, asking questions in optics, bio, and rust.<p>It's super fast with a lot of knowledge, a large context and great understanding. Really impressive model. > **dhbradshaw**: 在OLAMA上尝试了3B，在光学、生物和铁锈方面提出了问题<p> 它；s速度极快，知识丰富，背景广阔，理解力强。令人印象深刻的模型。 -