<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Deep-Learning on Semonan Book</title><link>https://semonan.com/en/tags/deep-learning/</link><description>Recent content in Deep-Learning on Semonan Book</description><generator>Hugo</generator><language>en-US</language><lastBuildDate>Wed, 16 Oct 2024 00:00:00 +0000</lastBuildDate><atom:link href="https://semonan.com/en/tags/deep-learning/rss.xml" rel="self" type="application/rss+xml"/><item><title>Face Analysis based on Deep Learning</title><link>https://semonan.com/en/book/ai/face-processing/face-analysis-deep-learning/</link><pubDate>Wed, 16 Oct 2024 00:00:00 +0000</pubDate><guid>https://semonan.com/en/book/ai/face-processing/face-analysis-deep-learning/</guid><description>&lt;h1 id="face-analysis-based-on-deep-learning"&gt;Face Analysis based on Deep Learning&lt;a class="anchor" href="#face-analysis-based-on-deep-learning"&gt;#&lt;/a&gt;&lt;/h1&gt;
&lt;p&gt;With the advancement of deep learning technology, the performance of face analysis has also improved.&lt;br&gt;
I will introduce a high-performing open-source face analysis library and explain how to use it.&lt;/p&gt;
&lt;h2 id="insightface-2d-and-3d-face-analysis-project"&gt;InsightFace: 2D and 3D Face Analysis Project&lt;a class="anchor" href="#insightface-2d-and-3d-face-analysis-project"&gt;#&lt;/a&gt;&lt;/h2&gt;
&lt;ul&gt;
&lt;li&gt;&lt;a href="https://github.com/deepinsight/insightface" rel="noopener noreferrer" target="_blank"&gt;&lt;code&gt;https://github.com/deepinsight/insightface&lt;/code&gt;&lt;/a&gt;&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;※ License : Please refer to the above site !&lt;/p&gt;
&lt;p&gt;Here, we share various models, and I would like to introduce the &lt;code&gt;buffalo_l&lt;/code&gt; model pack.&lt;br&gt;
&lt;code&gt;buffalo_l&lt;/code&gt; provides blob box, key points, detection score, landmark 2D/3D, gender, age, embedding, and pose information.&lt;br&gt;
 &lt;br&gt;
I will demonstrate the process of face analysis using Python.&lt;/p&gt;</description></item><item><title>LoRA</title><link>https://semonan.com/en/book/ai/fine-tuning/lora/</link><pubDate>Sun, 07 Jul 2024 00:00:00 +0000</pubDate><guid>https://semonan.com/en/book/ai/fine-tuning/lora/</guid><description>&lt;link rel="stylesheet" href="https://semonan.com/katex/katex.min.css" /&gt;&lt;script defer src="https://semonan.com/katex/katex.min.js"&gt;&lt;/script&gt;&lt;script defer src="https://semonan.com/katex/auto-render.min.js" onload="renderMathInElement(document.body, {&amp;#34;delimiters&amp;#34;:[{&amp;#34;left&amp;#34;:&amp;#34;$$&amp;#34;,&amp;#34;right&amp;#34;:&amp;#34;$$&amp;#34;,&amp;#34;display&amp;#34;:true},{&amp;#34;left&amp;#34;:&amp;#34;$&amp;#34;,&amp;#34;right&amp;#34;:&amp;#34;$&amp;#34;,&amp;#34;display&amp;#34;:false},{&amp;#34;left&amp;#34;:&amp;#34;\\[&amp;#34;,&amp;#34;right&amp;#34;:&amp;#34;\\]&amp;#34;,&amp;#34;display&amp;#34;:true},{&amp;#34;left&amp;#34;:&amp;#34;\\(&amp;#34;,&amp;#34;right&amp;#34;:&amp;#34;\\)&amp;#34;,&amp;#34;display&amp;#34;:false}]});"&gt;&lt;/script&gt;
&lt;h1 id="lora-low-rank-adaptation"&gt;LoRA (Low-Rank Adaptation)&lt;a class="anchor" href="#lora-low-rank-adaptation"&gt;#&lt;/a&gt;&lt;/h1&gt;
&lt;p&gt;LoRA is one of the PEFT (Parameter-Efficient Fine-Tuning) techniques.&lt;br&gt;
This technique efficiently fine-tunes large pre-trained models for specific tasks.&lt;/p&gt;
&lt;p&gt;(The following content is referenced from the paper &amp;ldquo;LoRA: Low-Rank Adaptation of Large Language Models.&amp;rdquo;)&lt;/p&gt;
&lt;h2 id="backgroundproblem"&gt;Background/Problem&lt;a class="anchor" href="#backgroundproblem"&gt;#&lt;/a&gt;&lt;/h2&gt;
&lt;p&gt;Models like LLM (Large Language Models) have an extremely large number of parameters.&lt;br&gt;
For example, the llama3 model, released in April 2024, has about 70 billion parameters and a file size of over 40GB, with many models being even larger.&lt;br&gt;
Full fine-tuning of such large models requires high-performance GPUs and considerable training time.&lt;br&gt;
Additionally, fully fine-tuning the base model may potentially degrade the fundamental performance learned during pretraining.&lt;/p&gt;</description></item></channel></rss>