<?xml version="1.0" encoding="utf-8" ?><rss version="2.0"><channel><title><![CDATA[越努力，越幸运]]></title><description><![CDATA[认真的人 自带光芒]]></description><link>https://blog.csdn.net/TYUT_xiaoming</link><language>zh-cn</language><generator>https://blog.csdn.net/</generator><copyright><![CDATA[Copyright &copy; TYUT_xiaoming]]></copyright><item><title><![CDATA[Docker 运行cuda11.8 + TensorRT8.5.3]]></title><link>https://blog.csdn.net/TYUT_xiaoming/article/details/163369958</link><guid>https://blog.csdn.net/TYUT_xiaoming/article/details/163369958</guid><author>TYUT_xiaoming</author><pubDate>Fri, 31 Jul 2026 16:04:51 +0800</pubDate><description><![CDATA[宿主机系统 CUDA Toolkit=12.8；NVIDIA 驱动是向下兼容的（高驱动可以跑容器内 CUDA11.8 程序）。]]></description><category></category></item><item><title><![CDATA[yolo模型参数调优思路]]></title><link>https://blog.csdn.net/TYUT_xiaoming/article/details/162755178</link><guid>https://blog.csdn.net/TYUT_xiaoming/article/details/162755178</guid><author>TYUT_xiaoming</author><pubDate>Fri, 10 Jul 2026 10:56:32 +0800</pubDate><description><![CDATA[]]></description><category></category></item><item><title><![CDATA[yolo 模型训练及数据集迭代管理方案]]></title><link>https://blog.csdn.net/TYUT_xiaoming/article/details/162671028</link><guid>https://blog.csdn.net/TYUT_xiaoming/article/details/162671028</guid><author>TYUT_xiaoming</author><pubDate>Tue, 07 Jul 2026 17:56:01 +0800</pubDate><description><![CDATA[【代码】yolo 模型训练及数据集迭代管理方案。]]></description><category></category></item><item><title><![CDATA[CUDA Execution Provider]]></title><link>https://blog.csdn.net/TYUT_xiaoming/article/details/162628572</link><guid>https://blog.csdn.net/TYUT_xiaoming/article/details/162628572</guid><author>TYUT_xiaoming</author><pubDate>Mon, 06 Jul 2026 11:42:48 +0800</pubDate><description><![CDATA[包的设计理念是能够与 PyTorch 无缝配合使用，前提是两者都是基于相同版本的 CUDA 和 cuDNN 构建的。在安装支持 CUDA 功能的 PyTorch 时，系统中会自动包含所需的 CUDA 和 cuDNN DLL 文件，因此无需单独安装 CUDA 工具包或 cuDNN。为确保 ONNX Runtime 能够使用 PyTorch 所安装的 DLL 文件，你可以在创建推理会话之前预先加载这些库。Configuration Options   配置选项。]]></description><category></category></item><item><title><![CDATA[TensorRT Execution Provider]]></title><link>https://blog.csdn.net/TYUT_xiaoming/article/details/162627212</link><guid>https://blog.csdn.net/TYUT_xiaoming/article/details/162627212</guid><author>TYUT_xiaoming</author><pubDate>Mon, 06 Jul 2026 11:10:26 +0800</pubDate><description><![CDATA[在 ONNX Runtime 中，TensorRT 执行引擎利用了 NVIDIA 的 TensorRT 深度学习推理引擎，从而提升了基于该引擎的 ONNX 模型在 NVIDIA GPU 上的运行效率。微软与 NVIDIA 密切合作，将 TensorRT 执行引擎成功集成到了 ONNX Runtime 中。，这样 Onnx Runtime 就能将那些 TensorRT 不支持的节点分配给 CUDA 执行提供程序来处理。要使用 TensorRT 执行提供程序，必须在实例化。]]></description><category></category></item><item><title><![CDATA[conda 环境安装onnx-runtime]]></title><link>https://blog.csdn.net/TYUT_xiaoming/article/details/162509695</link><guid>https://blog.csdn.net/TYUT_xiaoming/article/details/162509695</guid><author>TYUT_xiaoming</author><pubDate>Thu, 02 Jul 2026 10:27:07 +0800</pubDate><description><![CDATA[直接安装支持 CUDA 12 + cuDNN 8 的专用版本：安装完成后执行import onnxruntime,会有如下报错信息这个报错是因为你的环境中安装了（具体为 2.2.6），而当前安装的或其他依赖模块是基于编译的，两者不兼容。解决这个问题的最简单有效的方法是将 NumPy 降级到 1.x 版本。]]></description><category></category></item><item><title><![CDATA[NVIDIA Jetson Orin]]></title><link>https://blog.csdn.net/TYUT_xiaoming/article/details/162278283</link><guid>https://blog.csdn.net/TYUT_xiaoming/article/details/162278283</guid><author>TYUT_xiaoming</author><pubDate>Wed, 24 Jun 2026 18:07:26 +0800</pubDate><description><![CDATA[启动后，默认进入综合视图。]]></description><category></category></item><item><title><![CDATA[SwanLabAI训练分析平台]]></title><link>https://blog.csdn.net/TYUT_xiaoming/article/details/162246034</link><guid>https://blog.csdn.net/TYUT_xiaoming/article/details/162246034</guid><author>TYUT_xiaoming</author><pubDate>Tue, 23 Jun 2026 22:02:17 +0800</pubDate><description><![CDATA[等功能，帮助团队快速发现训练问题，加速模型迭代。SwanLab 是一款。，面向模型训练团队，提供。]]></description><category></category></item><item><title><![CDATA[沉浸式翻译插件]]></title><link>https://blog.csdn.net/TYUT_xiaoming/article/details/162151045</link><guid>https://blog.csdn.net/TYUT_xiaoming/article/details/162151045</guid><author>TYUT_xiaoming</author><pubDate>Sat, 20 Jun 2026 19:13:58 +0800</pubDate><description><![CDATA[沉浸式翻译 - AI 双语网页翻译 | PDF翻译 | 视频翻译 | 漫画&图片翻译]]></description><category></category></item><item><title><![CDATA[tmux命令]]></title><link>https://blog.csdn.net/TYUT_xiaoming/article/details/162047031</link><guid>https://blog.csdn.net/TYUT_xiaoming/article/details/162047031</guid><author>TYUT_xiaoming</author><pubDate>Tue, 16 Jun 2026 22:56:24 +0800</pubDate><description><![CDATA[]]></description><category></category></item><item><title><![CDATA[文章转载：视觉Transformer发展简史：从ViT到DINOv2的技术演进]]></title><link>https://blog.csdn.net/TYUT_xiaoming/article/details/162009570</link><guid>https://blog.csdn.net/TYUT_xiaoming/article/details/162009570</guid><author>TYUT_xiaoming</author><pubDate>Mon, 15 Jun 2026 16:23:07 +0800</pubDate><description><![CDATA[https://zhuanlan.zhihu.com/p/1932783167590340035]]></description><category></category></item><item><title><![CDATA[yolo模型训练]]></title><link>https://blog.csdn.net/TYUT_xiaoming/article/details/161997807</link><guid>https://blog.csdn.net/TYUT_xiaoming/article/details/161997807</guid><author>TYUT_xiaoming</author><pubDate>Mon, 15 Jun 2026 11:51:29 +0800</pubDate><description><![CDATA[空闲 GPU 训练能够自动选择多 GPU 系统中利用率最低的 GPU，从而在无需手动选择的情况下优化资源使用。此功能根据利用率指标和 VRAM 可用性来识别可用 GPU。]]></description><category></category></item><item><title><![CDATA[pip install -U 解释]]></title><link>https://blog.csdn.net/TYUT_xiaoming/article/details/161996592</link><guid>https://blog.csdn.net/TYUT_xiaoming/article/details/161996592</guid><author>TYUT_xiaoming</author><pubDate>Mon, 15 Jun 2026 10:18:13 +0800</pubDate><description><![CDATA[【代码】pip install -U 解释。]]></description><category></category></item><item><title><![CDATA[Copilot 场景 → 模式选择 → 提示词模板]]></title><link>https://blog.csdn.net/TYUT_xiaoming/article/details/156711392</link><guid>https://blog.csdn.net/TYUT_xiaoming/article/details/156711392</guid><author>TYUT_xiaoming</author><pubDate>Thu, 08 Jan 2026 09:29:04 +0800</pubDate><description><![CDATA[请帮我初始化一个【项目类型，如 React/Vue/Express】项目，并配置好基本结构和依赖。请帮我实现一个【功能描述】，包括必要的文件创建、代码编写和依赖安装。请帮我将【旧模块】迁移到【新模块】，并更新所有相关文件的引用。请帮我设计【功能】所需的数据库表结构，并说明字段含义和关系。请解释这段代码的功能、输入输出、逻辑流程和可能的边界情况。请帮我重构这段代码，使其更清晰、可维护，并保持功能不变。适合：改代码、加功能、重构、补测试，所有修改需你确认。请对比【方案A】和【方案B】的优缺点，并给出推荐方案。]]></description><category></category></item><item><title><![CDATA[Copilot 的agent、ask、edit、plan模式有什么区别]]></title><link>https://blog.csdn.net/TYUT_xiaoming/article/details/156693728</link><guid>https://blog.csdn.net/TYUT_xiaoming/article/details/156693728</guid><author>TYUT_xiaoming</author><pubDate>Wed, 07 Jan 2026 18:12:10 +0800</pubDate><description><![CDATA[Copilot 的 ask、edit、agent、plan 四种模式，核心区别在于。，以下从定义、工作机制、核心特点、典型场景与操作流程展开，帮你快速区分并选对模式。]]></description><category></category></item><item><title><![CDATA[零样本目标检测及分割模型]]></title><link>https://blog.csdn.net/TYUT_xiaoming/article/details/156676790</link><guid>https://blog.csdn.net/TYUT_xiaoming/article/details/156676790</guid><author>TYUT_xiaoming</author><pubDate>Wed, 07 Jan 2026 11:06:30 +0800</pubDate><description><![CDATA[（Self-Supervised Vision Foundation Model），是 DINO 系列（DINO → DINOv2 → DINOv3）的集大成者。它无需任何人工标注，仅通过海量无标签图像学习通用、高分辨率、语义丰富的视觉特征，在。（frozen backbone）的情况下，即可在多种下游任务（如分类、分割、深度估计、医学影像等）中达到 SOTA 或接近 SOTA 的性能。DINOv3 是 Meta AI 在。]]></description><category></category></item><item><title><![CDATA[MediaMTX]]></title><link>https://blog.csdn.net/TYUT_xiaoming/article/details/156648329</link><guid>https://blog.csdn.net/TYUT_xiaoming/article/details/156648329</guid><author>TYUT_xiaoming</author><pubDate>Tue, 06 Jan 2026 14:48:32 +0800</pubDate><description><![CDATA[MediaMTX（原名）是一款轻量级、高性能、零依赖的开源实时媒体服务器和媒体代理，专为现代流媒体场景设计。它支持多种主流流媒体协议，能够实现视频/音频流的发布、读取、代理、录制、回放以及跨协议转换，适用于直播、视频监控、远程医疗、工业视觉、新闻回传等多种应用场景。]]></description><category></category></item><item><title><![CDATA[CANN 8.0.0 调研记录]]></title><link>https://blog.csdn.net/TYUT_xiaoming/article/details/156448894</link><guid>https://blog.csdn.net/TYUT_xiaoming/article/details/156448894</guid><author>TYUT_xiaoming</author><pubDate>Wed, 31 Dec 2025 15:58:51 +0800</pubDate><description><![CDATA[【代码】CANN 8.0.0 调研记录。]]></description><category></category></item><item><title><![CDATA[NVIDIA DGX Spark]]></title><link>https://blog.csdn.net/TYUT_xiaoming/article/details/155570587</link><guid>https://blog.csdn.net/TYUT_xiaoming/article/details/155570587</guid><author>TYUT_xiaoming</author><pubDate>Thu, 04 Dec 2025 14:32:03 +0800</pubDate><description><![CDATA[支持高达 200B 参数级别的AI模型（双 DGX Spark 互联可支持高达 405B 参数的模型）先进的连接功能，包括支持高性能 NVIDIA Connect-X™ 网络技术、 WiFi 7 等。基于 NVIDIA GB10 Grace Blackwell 超级芯片构建。NVIDIA Grace CPU 采用⾼性能 20-core Arm 架构。128 GB 统一寻址系统内存。]]></description><category></category></item><item><title><![CDATA[cuda和cudnn下载地址]]></title><link>https://blog.csdn.net/TYUT_xiaoming/article/details/154383804</link><guid>https://blog.csdn.net/TYUT_xiaoming/article/details/154383804</guid><author>TYUT_xiaoming</author><pubDate>Tue, 04 Nov 2025 09:22:43 +0800</pubDate><description><![CDATA[使用以下命令安装CUDA 11.6（推荐）]]></description><category></category></item></channel></rss>