<?xml version="1.0" encoding="utf-8"?><feed xmlns="http://www.w3.org/2005/Atom" ><generator uri="https://jekyllrb.com/" version="3.10.0">Jekyll</generator><link href="https://www.deepspeed.ai/feed.xml" rel="self" type="application/atom+xml" /><link href="https://www.deepspeed.ai/" rel="alternate" type="text/html" /><updated>2025-06-02T16:22:59-07:00</updated><id>https://www.deepspeed.ai/feed.xml</id><title type="html">DeepSpeed</title><subtitle>DeepSpeed is a deep learning optimization library that makes distributed training easy, efficient, and effective.</subtitle><entry><title type="html">DeepSpeed-FastGen：通过 MII 和 DeepSpeed-Inference 实现 LLM 高吞吐量文本生成</title><link href="https://www.deepspeed.ai/2023/11/05/deepspeed-fastgen-chinese.html" rel="alternate" type="text/html" title="DeepSpeed-FastGen：通过 MII 和 DeepSpeed-Inference 实现 LLM 高吞吐量文本生成" /><published>2023-11-05T16:00:00-08:00</published><updated>2023-11-05T16:00:00-08:00</updated><id>https://www.deepspeed.ai/2023/11/05/deepspeed-fastgen-chinese</id><content type="html" xml:base="https://www.deepspeed.ai/2023/11/05/deepspeed-fastgen-chinese.html"><![CDATA[]]></content><author><name></name></author><category term="inference" /><category term="Chinese" /></entry><entry><title type="html">DeepSpeed-FastGen: MIIとDeepSpeed-InferenceによるLLMのための高速なテキスト生成</title><link href="https://www.deepspeed.ai/2023/11/05/deepspeed-fastgen-japanese.html" rel="alternate" type="text/html" title="DeepSpeed-FastGen: MIIとDeepSpeed-InferenceによるLLMのための高速なテキスト生成" /><published>2023-11-05T16:00:00-08:00</published><updated>2023-11-05T16:00:00-08:00</updated><id>https://www.deepspeed.ai/2023/11/05/deepspeed-fastgen-japanese</id><content type="html" xml:base="https://www.deepspeed.ai/2023/11/05/deepspeed-fastgen-japanese.html"><![CDATA[]]></content><author><name></name></author><category term="inference" /><category term="Japanese" /></entry><entry><title type="html">DeepSpeed-FastGen: High-throughput Text Generation for LLMs via MII and DeepSpeed-Inference</title><link href="https://www.deepspeed.ai/2023/11/05/deepspeed-fastgen.html" rel="alternate" type="text/html" title="DeepSpeed-FastGen: High-throughput Text Generation for LLMs via MII and DeepSpeed-Inference" /><published>2023-11-05T16:00:00-08:00</published><updated>2023-11-05T16:00:00-08:00</updated><id>https://www.deepspeed.ai/2023/11/05/deepspeed-fastgen</id><content type="html" xml:base="https://www.deepspeed.ai/2023/11/05/deepspeed-fastgen.html"><![CDATA[]]></content><author><name></name></author><category term="inference" /><category term="English" /></entry><entry><title type="html">DeepSpeed-VisualChat：多轮图像+文字，为你展现不一样的AI聊天魅力</title><link href="https://www.deepspeed.ai/2023/10/03/deepspeed-visualchat-chinese.html" rel="alternate" type="text/html" title="DeepSpeed-VisualChat：多轮图像+文字，为你展现不一样的AI聊天魅力" /><published>2023-10-03T17:00:00-07:00</published><updated>2023-10-03T17:00:00-07:00</updated><id>https://www.deepspeed.ai/2023/10/03/deepspeed-visualchat-chinese</id><content type="html" xml:base="https://www.deepspeed.ai/2023/10/03/deepspeed-visualchat-chinese.html"><![CDATA[]]></content><author><name></name></author><category term="training" /><category term="Chinese" /></entry><entry><title type="html">DeepSpeed-VisualChat: 複数ラウンド・複数画像の入力が可能なAIチャット体験を実現</title><link href="https://www.deepspeed.ai/2023/10/03/deepspeed-visualchat-japanese.html" rel="alternate" type="text/html" title="DeepSpeed-VisualChat: 複数ラウンド・複数画像の入力が可能なAIチャット体験を実現" /><published>2023-10-03T17:00:00-07:00</published><updated>2023-10-03T17:00:00-07:00</updated><id>https://www.deepspeed.ai/2023/10/03/deepspeed-visualchat-japanese</id><content type="html" xml:base="https://www.deepspeed.ai/2023/10/03/deepspeed-visualchat-japanese.html"><![CDATA[]]></content><author><name></name></author><category term="training" /><category term="Japanese" /></entry><entry><title type="html">DeepSpeed-VisualChat: Improve Your Chat Experience with Multi-Round Multi-Image Inputs</title><link href="https://www.deepspeed.ai/2023/10/03/deepspeed-visualchat.html" rel="alternate" type="text/html" title="DeepSpeed-VisualChat: Improve Your Chat Experience with Multi-Round Multi-Image Inputs" /><published>2023-10-03T17:00:00-07:00</published><updated>2023-10-03T17:00:00-07:00</updated><id>https://www.deepspeed.ai/2023/10/03/deepspeed-visualchat</id><content type="html" xml:base="https://www.deepspeed.ai/2023/10/03/deepspeed-visualchat.html"><![CDATA[]]></content><author><name></name></author><category term="training" /><category term="English" /></entry><entry><title type="html">DeepSpeed4Science：利用先进的AI系统优化技术实现科学发现</title><link href="https://www.deepspeed.ai/2023/09/18/deepspeed4science-chinese.html" rel="alternate" type="text/html" title="DeepSpeed4Science：利用先进的AI系统优化技术实现科学发现" /><published>2023-09-18T17:00:00-07:00</published><updated>2023-09-18T17:00:00-07:00</updated><id>https://www.deepspeed.ai/2023/09/18/deepspeed4science-chinese</id><content type="html" xml:base="https://www.deepspeed.ai/2023/09/18/deepspeed4science-chinese.html"><![CDATA[]]></content><author><name></name></author><category term="training" /><category term="inference" /><category term="science" /><category term="Chinese" /></entry><entry><title type="html">DeepSpeed4Scienceイニシアティブ: 洗練されたAIシステムのテクノロジーにより大規模な科学的発見を可能に</title><link href="https://www.deepspeed.ai/2023/09/18/deepspeed4science-japanese.html" rel="alternate" type="text/html" title="DeepSpeed4Scienceイニシアティブ: 洗練されたAIシステムのテクノロジーにより大規模な科学的発見を可能に" /><published>2023-09-18T17:00:00-07:00</published><updated>2023-09-18T17:00:00-07:00</updated><id>https://www.deepspeed.ai/2023/09/18/deepspeed4science-japanese</id><content type="html" xml:base="https://www.deepspeed.ai/2023/09/18/deepspeed4science-japanese.html"><![CDATA[]]></content><author><name></name></author><category term="training" /><category term="inference" /><category term="science" /><category term="Japanese" /></entry><entry><title type="html">Announcing the DeepSpeed4Science Initiative: Enabling large-scale scientific discovery through sophisticated AI system technologies</title><link href="https://www.deepspeed.ai/2023/09/18/deepspeed4science.html" rel="alternate" type="text/html" title="Announcing the DeepSpeed4Science Initiative: Enabling large-scale scientific discovery through sophisticated AI system technologies" /><published>2023-09-18T17:00:00-07:00</published><updated>2023-09-18T17:00:00-07:00</updated><id>https://www.deepspeed.ai/2023/09/18/deepspeed4science</id><content type="html" xml:base="https://www.deepspeed.ai/2023/09/18/deepspeed4science.html"><![CDATA[]]></content><author><name></name></author><category term="training" /><category term="inference" /><category term="science" /><category term="English" /></entry><entry><title type="html">Zero Inference</title><link href="https://www.deepspeed.ai/2023/09/12/ZeRO-Inference.html" rel="alternate" type="text/html" title="Zero Inference" /><published>2023-09-12T00:00:00-07:00</published><updated>2023-09-12T00:00:00-07:00</updated><id>https://www.deepspeed.ai/2023/09/12/ZeRO-Inference</id><content type="html" xml:base="https://www.deepspeed.ai/2023/09/12/ZeRO-Inference.html"><![CDATA[<p>title: “ZeRO-Inference: 20X faster inference through weight quantization and KV cache offloading”
excerpt: “”
link: https://github.com/deepspeedai/DeepSpeedExamples/blob/master/inference/huggingface/zero_inference/README.md
date: 2023-09-12 00:09:00
tags: inference ZeRO quantization English
—</p>]]></content><author><name></name></author><summary type="html"><![CDATA[title: “ZeRO-Inference: 20X faster inference through weight quantization and KV cache offloading” excerpt: “” link: https://github.com/deepspeedai/DeepSpeedExamples/blob/master/inference/huggingface/zero_inference/README.md date: 2023-09-12 00:09:00 tags: inference ZeRO quantization English —]]></summary></entry></feed>