手握灵珠常奋笔, 心开天籁不吹箫
生成式 AI 对于内容创作者来说， 既是机遇， 也是挑战。 LLM（Large Langage Model) 大语言模型之火愈演愈㤠。 国外的 ChatGPT, 国内的 Kimi 几乎成了与搜索引擎一样必不可少的工具。 AI 不但能理解我们的问题， 理解大篇幅的文本， 还能长篇累牍的生成大块文章， 还能生成图像， 视频， 音频， 这一天来得好快。 现在我写文章， 每当词穷之际， 就会求助于 LLM， 写程序， 忘了函数用法， 也不再查手册， 而去问 LLM, 但是它们是怎么做到的呢。 通过文本生成 token 想想婴儿是怎么学语言的吧， 词汇是基础 , token 就是 LLM 的词汇， BPE（Byte Paire Encoding) 算法广泛用于 LLM ， 从给定的数据集中生成 token 通过 token 进行预测 以这句练打字常用的句子为例 the quick […] →Read more
Test-Driven Development (TDD) is an iterative development cycle that emphasizes writing automated tests before writing the actual feature or function. Put another way, TDD combines building and testing. This process not only helps ensure correctness of the code — but also helps to indirectly evolve the design and architecture of the project at hand. TDD […] →Read more
把大象放进冰箱有三个步骤 打开冰箱门 把大象放进冰箱 关上冰箱门 显然, 这个拆分是不合格的, 第一步和第三步没问题, 第二步太粗. 拆分得尽量小, 顺序安排合理, 做起来就更容易 →Read more
其实也就一个 PDCA 循环 自己为自己设计课程 自己督促自己学习和训练 自己反省自己的问题 自己采取行动改进自己 →Read more
Overview The scheduling mode of a pad defines how data is retrieved from (source) or given to (sink) pads. GStreamer can operate in two scheduling mode, called push- and pull-mode. GStreamer supports elements with pads in any of the scheduling modes where not all pads need to be operating in the same mode. So far, […] →Read more
Elements that synchronize buffers on the pipeline clock will usually measure the current QoS. They will also need to keep some statistics in order to generate the QOS event. 缓冲数据与管道时钟之间的同步通常可以用来衡量 QoS For each buffer that arrives in the sink, the element needs to calculate how late or how early it was. This is called the […] →Read more
TensorRT is a large and flexible project. It can handle a variety of conversion and deployment workflows, and which workflow is best for you will depend on your specific use case and problem setting. TensorRT provides several options for deployment, but all workflows involve the conversion of your model to an optimized representation, which TensorRT […] →Read more
refer to https://github.com/NVIDIA-AI-IOT/deepstream_python_apps →Read more
Goal The previous tutorial showed how to build a pipeline automatically. Now we are going to build a pipeline manually by instantiating each element and linking them all together. In the process, we will learn: What is a GStreamer element and how to create one. 什么是 GStreamer 元件，如何创建它 How to connect elements to each other. […] →Read more
GStreamer 是如此著名的多媒体框架，它的管道和插件模式令人印象深刻，眼界大开，原来管道模式可以玩得这么精妙。 目标 Nothing better to get a first impression about a software library than to print “Hello World” on the screen! 熟悉使用任何一种新的开发软件语言或者软件代码库的方法，最莫过于在屏幕上打印 "Hello world". 而对于多媒体框架来说，播放一段视频比 hello world 更合适于快速上手。 下面代码远多于一个标准的 Hello world, 不过多数是初始化和清理代码，真正起作用的也就四行代码。 废话不多说，还是看代码吧。 程序流程 @startuml start :初始化 gst_init; :构建管道 gst_gst_parse_launch; :开始播放 gst_element_set_state; :等待结束 gst_bus_timed_pop_filtered; :释放资源 gst_object_unref …; stop @enduml 源代码 #include <gst/gst.h> #ifdef __APPLE__ […] →Read more