aiXpert-优快云博客

原创 OO面向对象再认识

Objects in programming languages often embody both and .PropertiesFieldsInstance variablesProceduresOperationsActionsMember functions了解一个东西的属性和行为，就表示对该东西有了认识和概念（）。尽管有些东西并不存在，但只要对其有概念，就是对象，如古代神话中的龙、凤凰、月中白兔、嫦娥等都是我们熟悉的对象。但对于没有听过嫦娥奔月故事的外国人来说，嫦娥并不是对象。

2024-04-26 12:49:29 347

原创大厂100 NLP interview questions外企

CLASSIC NLPTF-IDF & ML (8)1. Write TF-IDF from scratch.2. What is normalization in TF-IDF ?3. Why do you need to know about TF-IDF in our time, and how can you use it in complex models?4. Explain how Naive Bayes works. What can you use it for?5.

2024-04-12 20:11:16 740

原创前端的演化Fundamentals of Web apps

React是由Facebook开发并且开源的UI库，换言之，

2024-04-08 07:33:13 1087

原创 Gemma尝鲜：本地笔记本流畅运行Gemma

各种办法尝试个遍，这样弄保证你能开始玩儿。

2024-02-28 10:54:40 1137 2

原创 Windows11(非WSL)安装Installing llama-cpp-python with GPU Support

直接安装，只支持CPU。想支持GPU，麻烦一些。

2024-02-18 13:43:18 3039 3

原创 WSL从C盘挪到其他盘

suwhoami。

2024-02-17 23:18:31 898

原创 OpenAI ChatGPT-4开发笔记2024-08：windows本地环境下载Llama 2

Meta提供的下载时一个.sh文件，Windows无法直接使用。，在git bash里面可以跑.sh。下载后放入path当中。

2024-02-16 08:08:03 1714

原创 OpenAI ChatGPT-4开发笔记2024-07：Embedding之Text Similarity文本相似度

语义相似度语义相似度，如同字面意思一样，就是形容两句话的语义是否相似，是不是表达着同样的意思。在上面所介绍的两类分类问题中，都需要用到语义相似度的计算。第一类需要计算问题与n个候选答案之间的语义相似度，第二类需要计算问题与n个候选问题之间的语义相似度。目前有许多方法可以用来计算语义相似度，例如余弦相似度（Cosine Similarity）、欧几里得距离（Euclidean Distance）、指数（exponential）、曼哈顿距离（Manhatta

2024-01-25 12:22:24 960

原创真假AI技术

AI, ML, DL, automation and robotics are transforming our world. Understanding the nature of AI/ML/DL is the critical step in building real/genuine/true AI systems.The domineering assumption of AI as emulating, mimicking, simulating, or replicating the huma

2024-01-25 04:30:23 444

原创 CSS明显比XPATH更性感！CSS再学一点儿

【代码】CSS明显比XPATH更性感！CSS再学一点儿。

2024-01-20 21:57:03 410

原创 WebDriverWait太强大

一会儿是拒绝，一会儿是验证码，一会儿是验证图形，一会儿又是直接进public profile。login validation逻辑不太好写。有了implicitly之后，基本上不再关注网速之类的影响。请WebDriverWait出场。

2024-01-20 21:39:04 582

原创为了这口醋，包的这饺子。为了Selenium，学有限的CSS，逐步替换XPATH

locatorChropathXPath简单，但不可靠。任何软件升级都有可能造成Path的变化。更不用说反扒技术的存在了。ID最好，但不常有。颠来倒去，还是CSS靠谱些。但学起来费劲。不过，为了Selenium的目的，CSS还是有限费劲，可整。

2024-01-14 11:58:44 1054

原创 XPath Locators Cheat Sheet

如果有id的话，谁还用XPath呢？

2024-01-13 23:43:26 1120

原创 Jupyter Markdown格式

Jupyter Notebook（此前被称为 IPython notebook）是一个交互式笔记本，支持运行 40 多种编程语言。Jupyter Notebook 的本质是一个 Web 应用程序，便于创建和共享程序文档，支持实时代码，数学方程，可视化和 markdown。用途包括：数据清理和转换，数值模拟

2024-01-13 05:50:23 713 1

原创 selenium不自动关闭chrome，selenium hello world

分析原因：代码2运行完之所以会关闭chrome浏览器，是因为对应chrome浏览器厂家提供的浏览器源生驱动文件（chromedriver.exe）自身逻辑设置引起的，方法运行完会自动关闭回收方法中定义的局部变量dr。selenium版本太高了，默认安装的是最新版本，将版本降低后，就不会自动关闭浏览器了（我原本安装的是4.8.2，卸载装了4.0.0）以上三种代码统一换成Firefox浏览器的话，均不会自动关闭浏览器。代码1：chrome浏览器不会自动关闭。代码3：chrome浏览器不会自动关闭。

2024-01-12 04:55:23 1054 1

原创 OpenAI ChatGPT-4开发笔记2024-06：最简Embedding

在文本摘要任务中，将输入文本嵌入为向量有助于模型理解文本的重要信息，从而更好地生成摘要。

2024-01-12 01:48:12 670

原创 OpenAI ChatGPT-4开发笔记2024-05：windows下anaconda中设置visual studio code workspace

选择./anaconda/envs/ai312/python/bin。打开ai312.code-workspace。进入workspace。关闭workspace，关闭VSCode。重新打开VSCode。安装extension, Material Icon Theme。

2024-01-05 12:10:45 668 1

原创 OpenAI ChatGPT-4开发笔记2024-04：Chat之Tool之2：multiple functions

1.设定目标#1.定义parameters for text completion"content": f"汇总3个function的aiXpert的结果"#2.定义self-function,3个return f"#1.设定目标 import json import openai #1.定义parameters for text completion ai_prompt = [ {

2024-01-04 12:18:36 1141 1

原创 OpenAI ChatGPT-4开发笔记2024-03：Chat之Tool和Tool_Call（含前function call）

In another world,原来的function call都不再正常工作了，必须全部重写。function和function call全部由tool和tool_choice取代。2023年11月之前关于function call的代码都准备翘翘。干嘛要整个tool出来取代function呢？原因有很多，不再赘述。作为程序员，我们真正关心的是：怎么改？简单来说，就是整合chatgpt的能力和你个人的能力通过这个tools。怎么做呢？第二步，调用chatgpt模型让chatgpt干活儿。问问chat

2024-01-03 08:44:22 4855 3

原创 OpenAI ChatGPT-4开发笔记2024-02：Chat之text generation之completions

大模型封装在库里，库放在服务器上，服务器放在微软的云上。我们能做的，仅仅是通过API这个小小的缝隙，窥探ai的奥妙。从程序员的角度而言，水平的高低，就体现在对openai的这几个api的理解程度上。

2024-01-03 00:02:02 924 1

原创 OpenAI ChatGPT-4开发笔记2024-01：开发环境

工欲善其事，必先利其器.其他目的各种平台罗卜青菜，学习肯定要用JupyterLab，没有之一。

2024-01-02 23:27:19 753 1

原创高中物理：正弦波sin纯音puretone原理

review一个Android语音代码，对纯音正弦波的算法产生了兴趣。翻开高中物理，内牛满面。目标生成一个纯音正弦波。函数：f(x) = sin(x)找到每个x对应的f(x)。#方法数字时代，抛弃演绎。用最笨的办法：取点，测值。所谓的取点，就是采样。采样越多，测出来的值就越多，和原来的sin就越匹配。正弦波必备三样：频率frequency采样率sample rate振幅level我们的例子采用：100Hz44.1kHz sampling rate16 bit正弦波一个

2021-02-07 10:24:56 1317

原创纯音语音资料免费下载

纯音语音资料免费下载下载地址文件名下载地址链接: 时长3s 10s.链接: 时长5s.内容规格费用免费时长3s, 5s, 10s波形正弦频率125Hz - 8kHz音量0dbfs格式wav, mp3采样率44100文件名正弦波125hz-8000hz, 时长：3s和10s，0dbfs。seeingvoice.com_sin_125Hz_0dBFS_3s.wav音频播放器seeingvoice.com_sin_125H

2021-02-06 02:28:28 1642

原创华为云DevCloud管理App开发之二：Git Flow设计

华为云DevCloud管理App开发之三：Git Flow、Xcode、华为代码托管CodeHub（接上回）1 建立Git Flow（接上回）Xcode和华为云DevCloud连上后，开始着手引进现有iOS项目。1 建立Git Flow通过giOS，团队熟悉新的Git流程，iOS App开发纳入正轨。法律上讲求事实正义和程序正义。开发App也是一样。代码质量固然重要，过程的科学、可控也同样不可忽视。从风控角度来看，甚至更重要一些。小团队，iOS和Android两个版本。从老板的角度，希望能够实现

2020-05-19 07:16:56 352

原创华为云DevCloud管理App开发之一：Git连接华为代码托管CodeHub，iOS和Android本地和云端开发环境搭建

直播：iOS、Android、Git、DevCloud一锅炖，不知名App开发项目管理进行时，XCode11, Android Studio 3.6.3, 华为云, 代码托管CodeHub,Git背景第一步建立项目第二步建仓库3.1 必须注册Apple Developer Program3.2 在App Store Connect 登记一个App3. 用Xcode11上传App到App Sto...

2020-05-12 21:55:55 1722

原创非对称加密和SSH、HTTPS的道理和逻辑

通俗讲解对称加密、非对称加密及其应用SSH、RSA、SSL、HTTPS背景1 对称密码2 非对称密码第三步本地Xcode通过Git连接华为云LiOS仓库第四步本地Android Studio通过Git连接华为云LAndroid仓库背景编程用到保密通信，绕不开SSH、HTTPS。深入了解背后的道理，才能在工作中得心应手。不涉及具体加密算法，只是想讲明白这是些什么东东。最古老的信息传递，没有密码，比如烽火。后来需要保密了，就有了密码系统。最容易想到的加密办法，就是我用一把钥匙锁上门，然后把钥匙给你，

2020-05-11 21:49:25 455

转载 2019云栖大会精品资料下载

https://developer.aliyun.com/article/719452?spm=a313p.198.plec.1067371257172&short_name=I3.B6QL&app=firefox公开课小能手 2019-09-26 31545浏览量简介： 9月25日-27日，2019杭州云栖大会圆满成功。大会包含130+场峰会和专场、上千位重量级分享嘉宾以及硬...

2019-10-09 19:57:19 822

原创 Kaldi单步完美运行AIShell v1 S5之五：DNN (chain)

@[TOC](Kaldi单步完美运行AIShell v1 S5之五：DNN (chain))致谢感谢AIShell在商业化道路上的探索。期待着v3的到来。Kaldi单步完美运行AIShell v1 S5之五：DNN (chain)终篇。Chain Model的结果可以线上、实时，才有独立的商用价值。第14部分：DNN Chain Model先看结果。错词率达到36.69%。monoph...

2019-01-15 10:53:38 4077 4

原创 Kaldi单步完美运行AIShell v1 S5之四：DNN (nnet3、xent、MPE)

@[TOC](Kaldi单步完美运行AIShell v1 S5之四：DNN (nnet3、xent、MPE))致谢感谢AIShell在商业化道路上的探索。期待着v3的到来。Kaldi下AIShell v1详细输出之四：DNN (nnet3、xent、MPE)一网打尽。第11部分：nnet3 DNN先看结果。错词率达到36.69%。monophone效果一般般。sv@HP:~/lkal...

2019-01-14 14:31:23 3838 2

原创 Kaldi单步完美运行AIShell v1 S5之三：三音tri1,tri2,tri3,tri4,tri5

Kaldi单步完美运行AIShell v1 S5之三：三音tri1 2 3 4致谢Kaldi下AIShell v1详细输出之三：三音triphone第六部分：结果更新第七部分：三音Tri1训练、解码、校准第八部分：三音素tri2（delta+delta-deltas）第九部分：三音素tri3a[LDA+MLLT]第十部分：三音素tri4第11部分：三音素tri5（SAT）第12部分：nnet3致...

2019-01-13 16:48:36 3237 3

原创 Kaldi单步完美运行AIShell v1 S5之二：单音素MonoPhone

Kaldi单步完美运行AIShell v1 S5之二：MONO致谢Kaldi下AIShell v1详细输出之二：monophone第三部分：单音素结果第四部分：Monophone训练、解码、校准第四部分：三音素致谢感谢AIShell在商业化道路上的探索。期待着v3的到来。Kaldi下AIShell v1详细输出之二：monophone一网打尽。第三部分：单音素结果先看结果。错词率达到3...

2019-01-13 11:28:38 1341 4

原创 Kaldi单步完美运行AIShell v1 S5之一：MONO前

Kaldi单步完美运行AIShell v1 S5之一：MONO前机器配置Kaldi下TIMIT详细输出第一部分：数据准备第二部分：MFCC &amp;amp;amp; CMVN第三部分：单音素机器配置sv@HP:~$ cat /proc/cpuinfo | grep model\ namemodel name : Intel(R) Core(TM) i7-8700 CPU @ 3.20GHzmodel n...

2019-01-13 09:34:48 2702

原创 Kaldi完美运行TIMIT完整结果（含DNN）

Kaldi完美运行TIMIT完整结果（含DNN）完全完整含DNN的TIMIT结果RESULTS机器配置Kaldi下TIMIT详细输出第一部分：数据准备第二部分：MFCC &amp;amp; CMVN第三部分：单音素第四部分：tri1: Deltas第五部分：LDA + MLLT第六部分：LDA +MLLT + SAT第七部分：SGMM2第八部分：MＭＩ + SGMM2第九部分：DNN第十部分：ＤＮＮ+SG...

2019-01-04 16:31:15 4203 7

firebase3.6.1

firebase最新版。下载后，plugins，右键，选择install from local，点这个zip就行了。持续更新，请关注！

2020-04-29

嵌入推理在ARM架构的实现案例分析Enabling Embedded Inference Engine with the ARM

嵌入推理在ARM架构的实现案例Enabling Embedded Inference Engine with the ARM Compute Library A Case Study.pdf

2019-08-26

雾计算在电信语音的应用.pdf

FIT: A Fog Computing Device for Speech TeleTreatments*

2019-08-26

5G边缘计算应用、技术调查.pdf

A Survey of Multi-Access Edge Computing in 5G and Beyond Fundamentals, Technology Integration, and State-of-the-Art

2019-08-26

Building Speech Recognition Systems with the Kaldi Toolkit

Sanjeev Khudanpur, Dan Povey and Jan Trmal Johns Hopkins University Center for Language and Speech Processing

2019-03-17

The Application of Hidden Markov Models in Speech Recognition

Hidden Markov Models (HMMs) provide a simple and effective framework for modelling time-varying spectral vector sequences. As a consequence, almost all present day large vocabulary continuous speech recognition (LVCSR) systems are based on HMMs. Whereas the basic principles underlying HMM-based LVCSR are rather straightforward, the approximations and simplifying assumptions involved in a direct implementation of these principles would result in a system which has poor accuracy and unacceptable sensitivity to changes in operating environment. Thus, the practical application of HMMs in modern systems involves considerable sophistication. The aim of this review is first to present the core architecture of a HMM-based LVCSR system and then describe the various refinements which are needed to achieve state-of-the-art performance. These refinements include feature projection, improved covariance modelling, discriminative parameter estimation, adaptation and normalisation, noise compensation and multi-pass system combination. The review concludes with a case study of LVCSR for Broadcast News and Conversation transcription in order to illustrate the techniques described.

2019-03-17

DNN-HMM Based Multilingual Recognizer of Telephone Speech

This thesis deals with the multilingual acoustic modeling problem based on the shared global phones inventory for five East Eurpoean languages: Czech, Russian, Hungarian, Slovak and Polish which are available within SpeechDat-E, i.e. the set of telephone speech databases. Because the SAMPA with unnormalized convention is used to represent the phonetic content of the particular languages and different symbols are in several cases representing the same phone, the mapping to the general X-SAMPA phonetic alphabet was proposed in the first step. The impact of a multilingual acoustic modeling was analyzed on the basis of a continuous speech recognition. The analysis of the acoustic modeling in the LVCSR task was performed for the GMM-HMM system and for the DNN-GMM approach. The experiments were performed for the LVCSR with the language specific acoustic model same as for the multilingual system. The particular recognizers were implemented via the Kaldi toolkit. One of this thesis goals is to provide a tutorial-style description of the Kaldi usage and create the recipe for the SpeechDat databases. Depending on the language, the best obtained accuracy of HMM recognizers was 18%-28%WER. DNN-HMM improved the results about 4%WER on average. The results for the multilingual HMM system reached the values from 25%-37%WER.The DNN approached had significant impact on the speech recognition accuracy for the multilingual system as well and it reduced theWER about 9% on average.

2019-01-22

The HTK Book 3.5

HTK is a toolkit for building Hidden Markov Models (HMMs). HMMs can be used to model any time series and the core of HTK is similarly general-purpose. However, HTK is primarily designed for building HMM-based speech processing tools, in particular recognisers. Thus, much of the infrastructure support in HTK is dedicated to this task. As shown in the picture above, there are two major processing stages involved. Firstly, the HTK training tools are used to estimate the parameters of a set of HMMs using training utterances and their associated transcriptions.

2019-01-17

A KALDI-DNN-based ASR system for Italian

The KALDI ASR engine adapted to Italian is described and the results obtained so far on some children speech ASR experiments are reported. We give a brief overview of KALDI, we describe in detail its DNN implementation, we introduce the acoustic model (AM) training procedure and we end describing some experiments on Italian children speech together with the final test procedures.

2019-01-14

自动语音识别ASR讲义（397页，全）Steve Renals & Hiroshi Shimodaira

Automatic Speech Recognition| ASR Lecture. About 18 lectures, plus a couple of extra lectures on basic introduction to neural networks. Lecturers: Steve Renals and Hiroshi Shimodaira.

2019-01-14

Kaldi语音识别实验讲义（全）University of Edinburgh

The main goal of this lab is to get acquainted with Kaldi. We will begin by creating and exploring a data directory for the TIMIT dataset. Then we will extract features for TIMIT upon which we can train a complete speech recognition system in the coming labs. An underlying goal of this lab is to get you acquainted with Kaldi. Notes on UNIX commands are included in boxes; feel free to skip them if you're already familiar. Most importantly, don't be afraid to ask questions when you get stuck.

2019-01-14

空空如也

TA创建的收藏夹 TA关注的收藏夹

TA关注的人