<?xml version="1.0" encoding="utf-8" ?><rss version="2.0"><channel><title><![CDATA[chwei20002005的博客]]></title><description><![CDATA[]]></description><link>https://blog.youkuaiyun.com/chwei20002005</link><language>zh-cn</language><generator>https://blog.youkuaiyun.com/</generator><copyright><![CDATA[Copyright &copy; chwei20002005]]></copyright><item><title><![CDATA[世界模型 VS LLM ：概要篇]]></title><link>https://blog.youkuaiyun.com/chwei20002005/article/details/156189958</link><guid>https://blog.youkuaiyun.com/chwei20002005/article/details/156189958</guid><author>chwei20002005</author><pubDate>Tue, 23 Dec 2025 10:58:25 +0800</pubDate><description><![CDATA[AI发展路线之争：世界模型与LLM的核心差异 当前AI领域存在两大发展路线之争：基于统计相关性的LLM（大语言模型）与基于物理因果律的世界模型。LLM擅长文本生成但缺乏物理理解，容易产生&quot;幻觉&quot;；世界模型则能模拟环境变化，具备反事实推理能力，是具身智能和自动驾驶的关键。OpenAI的Sora、Meta的JEPA架构以及自动驾驶系统都展示了世界模型的实际应用。未来趋势可能是两者的融合，构建兼具语言能力和物理推理的超级AI。两种模型底层都依赖transformer架构，展现了AI技术的统一性]]></description><category></category></item><item><title><![CDATA[对比 交叉熵损失（Cross-Entropy Loss）和KL散度（Kullback-Leibler Divergence）]]></title><link>https://blog.youkuaiyun.com/chwei20002005/article/details/149019318</link><guid>https://blog.youkuaiyun.com/chwei20002005/article/details/149019318</guid><author>chwei20002005</author><pubDate>Mon, 30 Jun 2025 11:26:33 +0800</pubDate><description><![CDATA[对比 交叉熵损失（Cross-Entropy Loss）和KL散度（Kullback-Leibler Divergence）]]></description><category></category></item><item><title><![CDATA[Pycharm 配置解释器]]></title><link>https://blog.youkuaiyun.com/chwei20002005/article/details/148428547</link><guid>https://blog.youkuaiyun.com/chwei20002005/article/details/148428547</guid><author>chwei20002005</author><pubDate>Wed, 04 Jun 2025 15:46:52 +0800</pubDate><description><![CDATA[Step 2 这里第二步一定要注意类型要选择python 而不是conda。虽然我的解释器是conda 里面建立的一个环境。今天更新了一版pycharm，因为很久没有配置解释器了，发现一直失败。经过来回试了几次终于成功了，记录一下过程。]]></description><category></category></item><item><title><![CDATA[【DeepSeek-R1】DeepSeek-R1-Distill-Qwen-1.5B 流式与非流式推理调用]]></title><link>https://blog.youkuaiyun.com/chwei20002005/article/details/145434716</link><guid>https://blog.youkuaiyun.com/chwei20002005/article/details/145434716</guid><author>chwei20002005</author><pubDate>Mon, 03 Feb 2025 23:46:57 +0800</pubDate><description><![CDATA[windows 11 下，部署 DeepSeek-R1-Distill-Qwen-1.5B。为了根据用户的需要做了两种形式一种是流式生成一种是非流式。]]></description><category></category></item><item><title><![CDATA[DeepSeek-R1-Distill-Qwen-1.5B 本地部署报错解决]]></title><link>https://blog.youkuaiyun.com/chwei20002005/article/details/145432367</link><guid>https://blog.youkuaiyun.com/chwei20002005/article/details/145432367</guid><author>chwei20002005</author><pubDate>Mon, 03 Feb 2025 19:28:48 +0800</pubDate><description><![CDATA[attn_implementation="eager"  # 禁用可能不稳定的优化。本地部署“DeepSeek-R1-Distill-Qwen-1.5B” 运行时候报错。torch_dtype=torch.bfloat16 即可。里面torch.float16 变为。# 添加稳定性相关参数。]]></description><category></category></item><item><title><![CDATA[langchain 如何使用本地大模型（LLM）]]></title><link>https://blog.youkuaiyun.com/chwei20002005/article/details/138159845</link><guid>https://blog.youkuaiyun.com/chwei20002005/article/details/138159845</guid><author>chwei20002005</author><pubDate>Wed, 24 Apr 2024 15:15:57 +0800</pubDate><description><![CDATA[langchain 如何使用本地的模型及常见的报错解决 ]]></description><category></category></item><item><title><![CDATA[基于Cifar100数据Alexnet改进 （tensorflow版本）]]></title><link>https://blog.youkuaiyun.com/chwei20002005/article/details/129616671</link><guid>https://blog.youkuaiyun.com/chwei20002005/article/details/129616671</guid><author>chwei20002005</author><pubDate>Fri, 17 Mar 2023 11:45:52 +0800</pubDate><description><![CDATA[【代码】基于Cifar100数据Alexnet改进 （tensorflow版本）]]></description><category></category></item><item><title><![CDATA[【Arduino nano 33 ble sense rev2】学习历程1]]></title><link>https://blog.youkuaiyun.com/chwei20002005/article/details/129500109</link><guid>https://blog.youkuaiyun.com/chwei20002005/article/details/129500109</guid><author>chwei20002005</author><pubDate>Mon, 13 Mar 2023 18:08:13 +0800</pubDate><description><![CDATA[比如我的板子是 Arduino nano 33 ble sense rev2，开始我以为选择Arduino Nano，就行了，但是后来编译时候才发现出错，具体情况一会儿再说。在输入需要的板子，比如我的 Arduino nano 33 ble sense rev2 需要的开发板是 Arduino Nano BLE 33。（2）数据线的问题，我开始就是随便找了一根安卓的线，后来发现这个线只是充电线，不是数据线，所以重新上网买了一根数据线，这个买的时候一定问一下。下载成功后，可以选择得出需要的开发板。]]></description><category></category></item><item><title><![CDATA[车载以太网与CAN通信之间的区别]]></title><link>https://blog.youkuaiyun.com/chwei20002005/article/details/129114497</link><guid>https://blog.youkuaiyun.com/chwei20002005/article/details/129114497</guid><author>chwei20002005</author><pubDate>Sun, 19 Feb 2023 20:52:55 +0800</pubDate><description><![CDATA[总的来说，车载以太网在车辆内部的高速数据传输和网络连接方面表现出色，适合于连接车辆内的各种高级电子设备和应用。在车辆内部的高速数据传输和网络连接方面表现出色，适合于连接车辆内的各种高级电子设备和应用。未来，随着汽车电子系统的不断发展，车辆内部的数据交换需求会越来越高，车载以太网的应用会越来越广泛。数据传输质量：车载以太网提供更高的数据传输质量和稳定性，可以防止数据包的丢失和延迟。可以支持更多的功能，例如错误检测和纠正，网络管理，多主机通信等。适合于复杂的控制和通信任务，例如发动机控制和制动系统控制等。]]></description><category></category></item><item><title><![CDATA[[NLP] 对比jieba.cut 与jieba.lcut的区别]]></title><link>https://blog.youkuaiyun.com/chwei20002005/article/details/128218964</link><guid>https://blog.youkuaiyun.com/chwei20002005/article/details/128218964</guid><author>chwei20002005</author><pubDate>Wed, 07 Dec 2022 13:51:57 +0800</pubDate><description><![CDATA[jieba.cut与lcut对比 ]]></description><category></category></item><item><title><![CDATA[Pytorch Bert 中文分类 运行代码时候遇到的问题]]></title><link>https://blog.youkuaiyun.com/chwei20002005/article/details/128172540</link><guid>https://blog.youkuaiyun.com/chwei20002005/article/details/128172540</guid><author>chwei20002005</author><pubDate>Sun, 04 Dec 2022 16:41:06 +0800</pubDate><description><![CDATA[运行bert代码时的报错及解决方案。]]></description><category></category></item><item><title><![CDATA[bert 环境搭建之Pytorch&Transformer 安装]]></title><link>https://blog.youkuaiyun.com/chwei20002005/article/details/128165073</link><guid>https://blog.youkuaiyun.com/chwei20002005/article/details/128165073</guid><author>chwei20002005</author><pubDate>Sat, 03 Dec 2022 19:27:35 +0800</pubDate><description><![CDATA[transformer 安装 & Pytorch 安装]]></description><category></category></item><item><title><![CDATA[Python 图片（.jpg .bmp等）转为指定大小的raw]]></title><link>https://blog.youkuaiyun.com/chwei20002005/article/details/128131694</link><guid>https://blog.youkuaiyun.com/chwei20002005/article/details/128131694</guid><author>chwei20002005</author><pubDate>Thu, 01 Dec 2022 14:39:50 +0800</pubDate><description><![CDATA[把.jpg和.bmp的图片转成指定大小的 raw 格式]]></description><category></category></item><item><title><![CDATA[Python resize 图片并保存为raw 格式]]></title><link>https://blog.youkuaiyun.com/chwei20002005/article/details/128130015</link><guid>https://blog.youkuaiyun.com/chwei20002005/article/details/128130015</guid><author>chwei20002005</author><pubDate>Thu, 01 Dec 2022 12:27:10 +0800</pubDate><description><![CDATA[【代码】Python resize 图片并保存为raw 格式。]]></description><category></category></item><item><title><![CDATA[Tensorflow Transfer Learning (迁移学习) ---云端部署代码方式]]></title><link>https://blog.youkuaiyun.com/chwei20002005/article/details/128127007</link><guid>https://blog.youkuaiyun.com/chwei20002005/article/details/128127007</guid><author>chwei20002005</author><pubDate>Thu, 01 Dec 2022 10:19:47 +0800</pubDate><description><![CDATA[Tensorflow Transfer Learning (迁移学习) ---云端部署代码方式]]></description><category></category></item><item><title><![CDATA[pip git+install 报错解决]]></title><link>https://blog.youkuaiyun.com/chwei20002005/article/details/128126582</link><guid>https://blog.youkuaiyun.com/chwei20002005/article/details/128126582</guid><author>chwei20002005</author><pubDate>Thu, 01 Dec 2022 09:53:21 +0800</pubDate><description><![CDATA[用pip git + install 报错的解决方案。]]></description><category></category></item><item><title><![CDATA[TensorflowLite 模型转为Metadata format -----图片分类模型]]></title><link>https://blog.youkuaiyun.com/chwei20002005/article/details/128119572</link><guid>https://blog.youkuaiyun.com/chwei20002005/article/details/128119572</guid><author>chwei20002005</author><pubDate>Wed, 30 Nov 2022 18:06:53 +0800</pubDate><description><![CDATA[TensorflowLite模型在手机端部署]]></description><category></category></item><item><title><![CDATA[Windows 安装TVM 及各种报错解决！无GPU版本]]></title><link>https://blog.youkuaiyun.com/chwei20002005/article/details/127786861</link><guid>https://blog.youkuaiyun.com/chwei20002005/article/details/127786861</guid><author>chwei20002005</author><pubDate>Thu, 10 Nov 2022 14:56:34 +0800</pubDate><description><![CDATA[TVM安装 模型部署 模型量化]]></description><category></category></item><item><title><![CDATA[tflite 学习——生成.tflite 模型与验证]]></title><link>https://blog.youkuaiyun.com/chwei20002005/article/details/127617772</link><guid>https://blog.youkuaiyun.com/chwei20002005/article/details/127617772</guid><author>chwei20002005</author><pubDate>Tue, 01 Nov 2022 10:26:14 +0800</pubDate><description><![CDATA[用Resnet 模型进行对tensorflow lite验证。]]></description><category></category></item><item><title><![CDATA[Linux(Ubuntu) 安装anaconda环境]]></title><link>https://blog.youkuaiyun.com/chwei20002005/article/details/127532325</link><guid>https://blog.youkuaiyun.com/chwei20002005/article/details/127532325</guid><author>chwei20002005</author><pubDate>Wed, 26 Oct 2022 16:44:05 +0800</pubDate><description><![CDATA[conda create -n Tensorlite python=3.8 (这一块与以前写过Tensorflow 2.* 安装一样)再输入 conda create -n Tensorlite python=3.8 就成功了！这一步选择 增加解释器 选择conda enviroment 下拉菜单就可以发现。cd /pycharm-community-2022.2.3/bin 执行。这一步需要按一堆回车空格，直到出现yes no的选择。./pycharm.sh 就可以了。Step 4 下载pycharm。]]></description><category></category></item></channel></rss>