<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Performance on Jaehun's Blog</title><link>https://jaehun.me/tags/performance/</link><description>Recent content in Performance on Jaehun's Blog</description><generator>Hugo</generator><language>ko-kr</language><lastBuildDate>Sat, 19 Sep 2026 01:44:51 +0900</lastBuildDate><atom:link href="https://jaehun.me/tags/performance/index.xml" rel="self" type="application/rss+xml"/><item><title>FLEX ATTENTION: A PROGRAMMING MODEL FOR GENERATING OPTIMIZED ATTENTION KERNELS</title><link>https://jaehun.me/posts/flex-attention-a-programming-model-for-generating-optimized-attention-kernels/</link><pubDate>Mon, 07 Apr 2025 00:00:00 +0900</pubDate><guid>https://jaehun.me/posts/flex-attention-a-programming-model-for-generating-optimized-attention-kernels/</guid><description>&lt;p&gt;&lt;a&#10; href="https://arxiv.org/abs/2412.05496v1"target="_blank"&#10; class="inline-flex items-center gap-1"&#10; &gt;논문 링크&lt;svg class="h-3 w-3 flex-shrink-0" id="external-link" xmlns="http://www.w3.org/2000/svg" viewBox="0 0 24 24"&gt;&lt;path fill="none" stroke="currentColor" stroke-linecap="round" stroke-linejoin="round" stroke-width="2" d="M15 3h6v6m-11 5L21 3m-3 10v6a2 2 0 0 1-2 2H5a2 2 0 0 1-2-2V8a2 2 0 0 1 2-2h6"/&gt;&lt;/svg&gt;&#10; &lt;/a&gt;&lt;/p&gt;&#10;&lt;h2 id="강점과-독창성-핵심-알고리즘-그리고-한계"&gt;강점과 독창성, 핵심 알고리즘, 그리고 한계&lt;a href="#%ea%b0%95%ec%a0%90%ea%b3%bc-%eb%8f%85%ec%b0%bd%ec%84%b1-%ed%95%b5%ec%8b%ac-%ec%95%8c%ea%b3%a0%eb%a6%ac%ec%a6%98-%ea%b7%b8%eb%a6%ac%ea%b3%a0-%ed%95%9c%ea%b3%84" class="heading-anchor" aria-label="이 섹션에 대한 링크"&gt;&lt;/a&gt;&lt;/h2&gt;&#10; &lt;blockquote&#10; class="border-primary bg-muted/30 hover:bg-muted/40 my-6 rounded-r-lg border-l-4 py-4 pr-4 pl-6 italic transition-all duration-200 ease-out hover:translate-x-1"&#10; &gt;&#10; &lt;div class="prose prose-sm max-w-none"&gt;&#10; &lt;p&gt;물어본 것: 이 논문을 아주 자세하게 읽고 논문의 강점과 독창적인 지점을 설명해주고 핵심 알고리즘을 예시 입력을 들어서 전체적인 과정을 설명해줘 추가적으로 논문의 한계점에 대해서도 알려줘&lt;/p&gt;</description></item><item><title>QServe: W4A8KV4 Quantization and System Co-design for Efficient LLM Serving</title><link>https://jaehun.me/posts/qserve-w4a8kv4-quantization-and-system-co-design-for-efficient-llm-serving/</link><pubDate>Tue, 18 Mar 2025 00:00:00 +0900</pubDate><guid>https://jaehun.me/posts/qserve-w4a8kv4-quantization-and-system-co-design-for-efficient-llm-serving/</guid><description>&lt;p&gt;&lt;a&#10; href="https://arxiv.org/abs/2405.04532"target="_blank"&#10; class="inline-flex items-center gap-1"&#10; &gt;논문 링크&lt;svg class="h-3 w-3 flex-shrink-0" id="external-link" xmlns="http://www.w3.org/2000/svg" viewBox="0 0 24 24"&gt;&lt;path fill="none" stroke="currentColor" stroke-linecap="round" stroke-linejoin="round" stroke-width="2" d="M15 3h6v6m-11 5L21 3m-3 10v6a2 2 0 0 1-2 2H5a2 2 0 0 1-2-2V8a2 2 0 0 1 2-2h6"/&gt;&lt;/svg&gt;&#10; &lt;/a&gt;&lt;/p&gt;&#10;&lt;h2 id="강점과-독창성-핵심-알고리즘-그리고-한계"&gt;강점과 독창성, 핵심 알고리즘, 그리고 한계&lt;a href="#%ea%b0%95%ec%a0%90%ea%b3%bc-%eb%8f%85%ec%b0%bd%ec%84%b1-%ed%95%b5%ec%8b%ac-%ec%95%8c%ea%b3%a0%eb%a6%ac%ec%a6%98-%ea%b7%b8%eb%a6%ac%ea%b3%a0-%ed%95%9c%ea%b3%84" class="heading-anchor" aria-label="이 섹션에 대한 링크"&gt;&lt;/a&gt;&lt;/h2&gt;&#10; &lt;blockquote&#10; class="border-primary bg-muted/30 hover:bg-muted/40 my-6 rounded-r-lg border-l-4 py-4 pr-4 pl-6 italic transition-all duration-200 ease-out hover:translate-x-1"&#10; &gt;&#10; &lt;div class="prose prose-sm max-w-none"&gt;&#10; &lt;p&gt;물어본 것: 이 논문을 아주 자세하게 읽고 논문의 강점과 독창적인 지점을 설명해주고 핵심 알고리즘을 예시 입력을 들어서 전체적인 과정을 설명해줘 추가적으로 논문의 한계점에 대해서도 알려줘&lt;/p&gt;</description></item><item><title>ClusterKV: Manipulating LLM KV Cache in Semantic Space for Recallable Compression</title><link>https://jaehun.me/posts/clusterkv-manipulating-llm-kv-cache-in-semantic-space-for-recallable-compression/</link><pubDate>Fri, 20 Dec 2024 00:00:00 +0900</pubDate><guid>https://jaehun.me/posts/clusterkv-manipulating-llm-kv-cache-in-semantic-space-for-recallable-compression/</guid><description>&lt;p&gt;&lt;a&#10; href="https://arxiv.org/abs/2412.03213v1"target="_blank"&#10; class="inline-flex items-center gap-1"&#10; &gt;논문 링크&lt;svg class="h-3 w-3 flex-shrink-0" id="external-link" xmlns="http://www.w3.org/2000/svg" viewBox="0 0 24 24"&gt;&lt;path fill="none" stroke="currentColor" stroke-linecap="round" stroke-linejoin="round" stroke-width="2" d="M15 3h6v6m-11 5L21 3m-3 10v6a2 2 0 0 1-2 2H5a2 2 0 0 1-2-2V8a2 2 0 0 1 2-2h6"/&gt;&lt;/svg&gt;&#10; &lt;/a&gt;&lt;/p&gt;&#10;&lt;h2 id="강점과-독창성-핵심-알고리즘-그리고-한계"&gt;강점과 독창성, 핵심 알고리즘, 그리고 한계&lt;a href="#%ea%b0%95%ec%a0%90%ea%b3%bc-%eb%8f%85%ec%b0%bd%ec%84%b1-%ed%95%b5%ec%8b%ac-%ec%95%8c%ea%b3%a0%eb%a6%ac%ec%a6%98-%ea%b7%b8%eb%a6%ac%ea%b3%a0-%ed%95%9c%ea%b3%84" class="heading-anchor" aria-label="이 섹션에 대한 링크"&gt;&lt;/a&gt;&lt;/h2&gt;&#10; &lt;blockquote&#10; class="border-primary bg-muted/30 hover:bg-muted/40 my-6 rounded-r-lg border-l-4 py-4 pr-4 pl-6 italic transition-all duration-200 ease-out hover:translate-x-1"&#10; &gt;&#10; &lt;div class="prose prose-sm max-w-none"&gt;&#10; &lt;p&gt;물어본 것: 이 논문을 아주 자세하게 읽고 논문의 강점과 독창적인 지점을 설명해주고 핵심 알고리즘을 예시 입력을 들어서 전체적인 과정을 설명해줘 추가적으로 논문의 한계점에 대해서도 알려줘&lt;/p&gt;</description></item><item><title>Efficient LLM Inference with I/O-Aware Partial KV Cache Recomputation</title><link>https://jaehun.me/posts/efficient-llm-inference-with-i/o-aware-partial-kv-cache-recomputation/</link><pubDate>Fri, 20 Dec 2024 00:00:00 +0900</pubDate><guid>https://jaehun.me/posts/efficient-llm-inference-with-i/o-aware-partial-kv-cache-recomputation/</guid><description>&lt;p&gt;&lt;a&#10; href="https://arxiv.org/abs/2411.17089v1"target="_blank"&#10; class="inline-flex items-center gap-1"&#10; &gt;논문 링크&lt;svg class="h-3 w-3 flex-shrink-0" id="external-link" xmlns="http://www.w3.org/2000/svg" viewBox="0 0 24 24"&gt;&lt;path fill="none" stroke="currentColor" stroke-linecap="round" stroke-linejoin="round" stroke-width="2" d="M15 3h6v6m-11 5L21 3m-3 10v6a2 2 0 0 1-2 2H5a2 2 0 0 1-2-2V8a2 2 0 0 1 2-2h6"/&gt;&lt;/svg&gt;&#10; &lt;/a&gt;&lt;/p&gt;&#10;&lt;h2 id="강점과-독창성-핵심-알고리즘-그리고-한계"&gt;강점과 독창성, 핵심 알고리즘, 그리고 한계&lt;a href="#%ea%b0%95%ec%a0%90%ea%b3%bc-%eb%8f%85%ec%b0%bd%ec%84%b1-%ed%95%b5%ec%8b%ac-%ec%95%8c%ea%b3%a0%eb%a6%ac%ec%a6%98-%ea%b7%b8%eb%a6%ac%ea%b3%a0-%ed%95%9c%ea%b3%84" class="heading-anchor" aria-label="이 섹션에 대한 링크"&gt;&lt;/a&gt;&lt;/h2&gt;&#10; &lt;blockquote&#10; class="border-primary bg-muted/30 hover:bg-muted/40 my-6 rounded-r-lg border-l-4 py-4 pr-4 pl-6 italic transition-all duration-200 ease-out hover:translate-x-1"&#10; &gt;&#10; &lt;div class="prose prose-sm max-w-none"&gt;&#10; &lt;p&gt;물어본 것: 이 논문을 아주 자세하게 읽고 논문의 강점과 독창적인 지점을 설명해주고 핵심 알고리즘을 예시 입력을 들어서 전체적인 과정을 설명해줘 추가적으로 논문의 한계점에 대해서도 알려줘&lt;/p&gt;</description></item><item><title>SparseInfer: Training-free Prediction of Activation Sparsity for Fast LLM Inference</title><link>https://jaehun.me/posts/sparseinfer-training-free-prediction-of-activation-sparsity-for-fast-llm-inference/</link><pubDate>Fri, 20 Dec 2024 00:00:00 +0900</pubDate><guid>https://jaehun.me/posts/sparseinfer-training-free-prediction-of-activation-sparsity-for-fast-llm-inference/</guid><description>&lt;p&gt;&lt;a&#10; href="https://arxiv.org/abs/2411.12692v1"target="_blank"&#10; class="inline-flex items-center gap-1"&#10; &gt;논문 링크&lt;svg class="h-3 w-3 flex-shrink-0" id="external-link" xmlns="http://www.w3.org/2000/svg" viewBox="0 0 24 24"&gt;&lt;path fill="none" stroke="currentColor" stroke-linecap="round" stroke-linejoin="round" stroke-width="2" d="M15 3h6v6m-11 5L21 3m-3 10v6a2 2 0 0 1-2 2H5a2 2 0 0 1-2-2V8a2 2 0 0 1 2-2h6"/&gt;&lt;/svg&gt;&#10; &lt;/a&gt;&lt;/p&gt;&#10;&lt;h2 id="강점과-독창성-핵심-알고리즘-그리고-한계"&gt;강점과 독창성, 핵심 알고리즘, 그리고 한계&lt;a href="#%ea%b0%95%ec%a0%90%ea%b3%bc-%eb%8f%85%ec%b0%bd%ec%84%b1-%ed%95%b5%ec%8b%ac-%ec%95%8c%ea%b3%a0%eb%a6%ac%ec%a6%98-%ea%b7%b8%eb%a6%ac%ea%b3%a0-%ed%95%9c%ea%b3%84" class="heading-anchor" aria-label="이 섹션에 대한 링크"&gt;&lt;/a&gt;&lt;/h2&gt;&#10; &lt;blockquote&#10; class="border-primary bg-muted/30 hover:bg-muted/40 my-6 rounded-r-lg border-l-4 py-4 pr-4 pl-6 italic transition-all duration-200 ease-out hover:translate-x-1"&#10; &gt;&#10; &lt;div class="prose prose-sm max-w-none"&gt;&#10; &lt;p&gt;물어본 것: 이 논문을 아주 자세하게 읽고 논문의 강점과 독창적인 지점을 설명해주고 핵심 알고리즘을 예시 입력을 들어서 전체적인 과정을 설명해줘 추가적으로 논문의 한계점에 대해서도 알려줘&lt;/p&gt;</description></item><item><title>DeepSpeed-FastGen: High-throughput Text Generation for LLMs via MII and DeepSpeed-Inference</title><link>https://jaehun.me/posts/deepspeed-fastgen-high-throughput-text-generation-for-llms-via-mii-and-deepspeed-inference/</link><pubDate>Mon, 16 Dec 2024 00:00:00 +0900</pubDate><guid>https://jaehun.me/posts/deepspeed-fastgen-high-throughput-text-generation-for-llms-via-mii-and-deepspeed-inference/</guid><description>&lt;p&gt;&lt;a&#10; href="https://arxiv.org/abs/2401.08671"target="_blank"&#10; class="inline-flex items-center gap-1"&#10; &gt;논문 링크&lt;svg class="h-3 w-3 flex-shrink-0" id="external-link" xmlns="http://www.w3.org/2000/svg" viewBox="0 0 24 24"&gt;&lt;path fill="none" stroke="currentColor" stroke-linecap="round" stroke-linejoin="round" stroke-width="2" d="M15 3h6v6m-11 5L21 3m-3 10v6a2 2 0 0 1-2 2H5a2 2 0 0 1-2-2V8a2 2 0 0 1 2-2h6"/&gt;&lt;/svg&gt;&#10; &lt;/a&gt;&lt;/p&gt;&#10;&lt;h2 id="강점과-독창성-핵심-알고리즘-그리고-한계"&gt;강점과 독창성, 핵심 알고리즘, 그리고 한계&lt;a href="#%ea%b0%95%ec%a0%90%ea%b3%bc-%eb%8f%85%ec%b0%bd%ec%84%b1-%ed%95%b5%ec%8b%ac-%ec%95%8c%ea%b3%a0%eb%a6%ac%ec%a6%98-%ea%b7%b8%eb%a6%ac%ea%b3%a0-%ed%95%9c%ea%b3%84" class="heading-anchor" aria-label="이 섹션에 대한 링크"&gt;&lt;/a&gt;&lt;/h2&gt;&#10; &lt;blockquote&#10; class="border-primary bg-muted/30 hover:bg-muted/40 my-6 rounded-r-lg border-l-4 py-4 pr-4 pl-6 italic transition-all duration-200 ease-out hover:translate-x-1"&#10; &gt;&#10; &lt;div class="prose prose-sm max-w-none"&gt;&#10; &lt;p&gt;물어본 것: 이 논문을 아주 자세하게 읽고 논문의 강점과 독창적인 지점을 설명해주고 핵심 알고리즘을 예시 입력을 들어서 전체적인 과정을 설명해줘 추가적으로 논문의 한계점에 대해서도 알려줘&lt;/p&gt;</description></item><item><title>KIVI: A Tuning-Free Asymmetric 2bit Quantization for KV Cache</title><link>https://jaehun.me/posts/kivi-a-tuning-free-asymmetric-2bit-quantization-for-kv-cache/</link><pubDate>Mon, 09 Dec 2024 00:00:00 +0900</pubDate><guid>https://jaehun.me/posts/kivi-a-tuning-free-asymmetric-2bit-quantization-for-kv-cache/</guid><description>&lt;p&gt;&lt;a&#10; href="https://arxiv.org/abs/2402.02750"target="_blank"&#10; class="inline-flex items-center gap-1"&#10; &gt;논문 링크&lt;svg class="h-3 w-3 flex-shrink-0" id="external-link" xmlns="http://www.w3.org/2000/svg" viewBox="0 0 24 24"&gt;&lt;path fill="none" stroke="currentColor" stroke-linecap="round" stroke-linejoin="round" stroke-width="2" d="M15 3h6v6m-11 5L21 3m-3 10v6a2 2 0 0 1-2 2H5a2 2 0 0 1-2-2V8a2 2 0 0 1 2-2h6"/&gt;&lt;/svg&gt;&#10; &lt;/a&gt;&lt;/p&gt;&#10;&lt;h2 id="강점과-독창성-핵심-알고리즘-그리고-한계"&gt;강점과 독창성, 핵심 알고리즘, 그리고 한계&lt;a href="#%ea%b0%95%ec%a0%90%ea%b3%bc-%eb%8f%85%ec%b0%bd%ec%84%b1-%ed%95%b5%ec%8b%ac-%ec%95%8c%ea%b3%a0%eb%a6%ac%ec%a6%98-%ea%b7%b8%eb%a6%ac%ea%b3%a0-%ed%95%9c%ea%b3%84" class="heading-anchor" aria-label="이 섹션에 대한 링크"&gt;&lt;/a&gt;&lt;/h2&gt;&#10; &lt;blockquote&#10; class="border-primary bg-muted/30 hover:bg-muted/40 my-6 rounded-r-lg border-l-4 py-4 pr-4 pl-6 italic transition-all duration-200 ease-out hover:translate-x-1"&#10; &gt;&#10; &lt;div class="prose prose-sm max-w-none"&gt;&#10; &lt;p&gt;물어본 것: 이 논문을 아주 자세하게 읽고 논문의 강점과 독창적인 지점을 설명해주고 핵심 알고리즘을 예시 입력을 들어서 전체적인 과정을 설명해줘 추가적으로 논문의 한계점에 대해서도 알려줘&lt;/p&gt;</description></item><item><title>ALISA: Accelerating Large Language Model Inference via Sparsity-Aware KV Caching</title><link>https://jaehun.me/posts/alisa-accelerating-large-language-model-inference-via-sparsity-aware-kv-caching/</link><pubDate>Fri, 06 Dec 2024 00:00:00 +0900</pubDate><guid>https://jaehun.me/posts/alisa-accelerating-large-language-model-inference-via-sparsity-aware-kv-caching/</guid><description>&lt;p&gt;&lt;a&#10; href="https://arxiv.org/abs/2403.17312"target="_blank"&#10; class="inline-flex items-center gap-1"&#10; &gt;논문 링크&lt;svg class="h-3 w-3 flex-shrink-0" id="external-link" xmlns="http://www.w3.org/2000/svg" viewBox="0 0 24 24"&gt;&lt;path fill="none" stroke="currentColor" stroke-linecap="round" stroke-linejoin="round" stroke-width="2" d="M15 3h6v6m-11 5L21 3m-3 10v6a2 2 0 0 1-2 2H5a2 2 0 0 1-2-2V8a2 2 0 0 1 2-2h6"/&gt;&lt;/svg&gt;&#10; &lt;/a&gt;&lt;/p&gt;&#10;&lt;h2 id="강점과-독창성-핵심-알고리즘-그리고-한계"&gt;강점과 독창성, 핵심 알고리즘, 그리고 한계&lt;a href="#%ea%b0%95%ec%a0%90%ea%b3%bc-%eb%8f%85%ec%b0%bd%ec%84%b1-%ed%95%b5%ec%8b%ac-%ec%95%8c%ea%b3%a0%eb%a6%ac%ec%a6%98-%ea%b7%b8%eb%a6%ac%ea%b3%a0-%ed%95%9c%ea%b3%84" class="heading-anchor" aria-label="이 섹션에 대한 링크"&gt;&lt;/a&gt;&lt;/h2&gt;&#10; &lt;blockquote&#10; class="border-primary bg-muted/30 hover:bg-muted/40 my-6 rounded-r-lg border-l-4 py-4 pr-4 pl-6 italic transition-all duration-200 ease-out hover:translate-x-1"&#10; &gt;&#10; &lt;div class="prose prose-sm max-w-none"&gt;&#10; &lt;p&gt;물어본 것: 이 논문을 아주 자세하게 읽고 논문의 강점과 독창적인 지점을 설명해주고 핵심 알고리즘을 예시 입력을 들어서 전체적인 과정을 설명해줘 추가적으로 논문의 한계점에 대해서도 알려줘&lt;/p&gt;</description></item><item><title>FASTDECODE: High-Throughput GPU-Efficient LLM Serving using Heterogeneous Pipelines</title><link>https://jaehun.me/posts/fastdecode-high-throughput-gpu-efficient-llm-serving-using-heterogeneous-pipelines/</link><pubDate>Fri, 06 Dec 2024 00:00:00 +0900</pubDate><guid>https://jaehun.me/posts/fastdecode-high-throughput-gpu-efficient-llm-serving-using-heterogeneous-pipelines/</guid><description>&lt;p&gt;&lt;a&#10; href="https://arxiv.org/abs/2403.11421"target="_blank"&#10; class="inline-flex items-center gap-1"&#10; &gt;논문 링크&lt;svg class="h-3 w-3 flex-shrink-0" id="external-link" xmlns="http://www.w3.org/2000/svg" viewBox="0 0 24 24"&gt;&lt;path fill="none" stroke="currentColor" stroke-linecap="round" stroke-linejoin="round" stroke-width="2" d="M15 3h6v6m-11 5L21 3m-3 10v6a2 2 0 0 1-2 2H5a2 2 0 0 1-2-2V8a2 2 0 0 1 2-2h6"/&gt;&lt;/svg&gt;&#10; &lt;/a&gt;&lt;/p&gt;&#10;&lt;h2 id="강점과-독창성-핵심-알고리즘-그리고-한계"&gt;강점과 독창성, 핵심 알고리즘, 그리고 한계&lt;a href="#%ea%b0%95%ec%a0%90%ea%b3%bc-%eb%8f%85%ec%b0%bd%ec%84%b1-%ed%95%b5%ec%8b%ac-%ec%95%8c%ea%b3%a0%eb%a6%ac%ec%a6%98-%ea%b7%b8%eb%a6%ac%ea%b3%a0-%ed%95%9c%ea%b3%84" class="heading-anchor" aria-label="이 섹션에 대한 링크"&gt;&lt;/a&gt;&lt;/h2&gt;&#10; &lt;blockquote&#10; class="border-primary bg-muted/30 hover:bg-muted/40 my-6 rounded-r-lg border-l-4 py-4 pr-4 pl-6 italic transition-all duration-200 ease-out hover:translate-x-1"&#10; &gt;&#10; &lt;div class="prose prose-sm max-w-none"&gt;&#10; &lt;p&gt;물어본 것: 이 논문을 아주 자세하게 읽고 논문의 강점과 독창적인 지점을 설명해주고 핵심 알고리즘을 예시 입력을 들어서 전체적인 과정을 설명해줘 추가적으로 논문의 한계점에 대해서도 알려줘&lt;/p&gt;</description></item><item><title>FlexGen High-Throughput Generative Inference of Large Language Models with a Single GPU</title><link>https://jaehun.me/posts/flexgen-high-throughput-generative-inference-of-large-language-models-with-a-single-gpu/</link><pubDate>Fri, 01 Nov 2024 00:00:00 +0900</pubDate><guid>https://jaehun.me/posts/flexgen-high-throughput-generative-inference-of-large-language-models-with-a-single-gpu/</guid><description>&lt;p&gt;논문 : &lt;a&#10; href="https://arxiv.org/abs/2303.06865"target="_blank"&#10; class="inline-flex items-center gap-1"&#10; &gt;https://arxiv.org/abs/2303.06865&lt;svg class="h-3 w-3 flex-shrink-0" id="external-link" xmlns="http://www.w3.org/2000/svg" viewBox="0 0 24 24"&gt;&lt;path fill="none" stroke="currentColor" stroke-linecap="round" stroke-linejoin="round" stroke-width="2" d="M15 3h6v6m-11 5L21 3m-3 10v6a2 2 0 0 1-2 2H5a2 2 0 0 1-2-2V8a2 2 0 0 1 2-2h6"/&gt;&lt;/svg&gt;&#10; &lt;/a&gt;&lt;/p&gt;&#10;&lt;p&gt;개인의견 : 타겟은 다르지만 deepspeed(zero) 오프로딩을 한다는 측면에서는 비슷한 느낌을 받았다 그리고 정말 높은 throughtput을 위해서 quantization과 선형계획법으로 효율적으로 디바이스 사용을 가능하게 하는 방법으로 보인다. 다만 한계점으로 말하듯 TTFT(Time to First Token)가 증가하거나 quantization으로 인한 에러가 발생할 수 있는 구조로 생각이 된다.&lt;/p&gt;</description></item></channel></rss>