Picture for Yuxiong He

Yuxiong He

FP6-LLM: Efficiently Serving Large Language Models Through FP6-Centric Algorithm-System Co-Design

Add code
Jan 25, 2024
Viaarxiv icon

DeepSpeed-FastGen: High-throughput Text Generation for LLMs via MII and DeepSpeed-Inference

Add code
Jan 09, 2024
Viaarxiv icon

ZeroQuant(4+2): Redefining LLMs Quantization with a New FP6-Centric Strategy for Diverse Generative Tasks

Add code
Dec 18, 2023
Figure 1 for ZeroQuant(4+2): Redefining LLMs Quantization with a New FP6-Centric Strategy for Diverse Generative Tasks
Figure 2 for ZeroQuant(4+2): Redefining LLMs Quantization with a New FP6-Centric Strategy for Diverse Generative Tasks
Figure 3 for ZeroQuant(4+2): Redefining LLMs Quantization with a New FP6-Centric Strategy for Diverse Generative Tasks
Figure 4 for ZeroQuant(4+2): Redefining LLMs Quantization with a New FP6-Centric Strategy for Diverse Generative Tasks
Viaarxiv icon

ZeroQuant-HERO: Hardware-Enhanced Robust Optimized Post-Training Quantization Framework for W8A8 Transformers

Add code
Oct 26, 2023
Viaarxiv icon

DeepSpeed4Science Initiative: Enabling Large-Scale Scientific Discovery through Sophisticated AI System Technologies

Oct 11, 2023
Figure 1 for DeepSpeed4Science Initiative: Enabling Large-Scale Scientific Discovery through Sophisticated AI System Technologies
Figure 2 for DeepSpeed4Science Initiative: Enabling Large-Scale Scientific Discovery through Sophisticated AI System Technologies
Figure 3 for DeepSpeed4Science Initiative: Enabling Large-Scale Scientific Discovery through Sophisticated AI System Technologies
Figure 4 for DeepSpeed4Science Initiative: Enabling Large-Scale Scientific Discovery through Sophisticated AI System Technologies
Viaarxiv icon

DeepSpeed-VisualChat: Multi-Round Multi-Image Interleave Chat via Multi-Modal Causal Attention

Add code
Sep 29, 2023
Viaarxiv icon

DeepSpeed Ulysses: System Optimizations for Enabling Training of Extreme Long Sequence Transformer Models

Sep 25, 2023
Viaarxiv icon

RenAIssance: A Survey into AI Text-to-Image Generation in the Era of Large Model

Sep 02, 2023
Figure 1 for RenAIssance: A Survey into AI Text-to-Image Generation in the Era of Large Model
Figure 2 for RenAIssance: A Survey into AI Text-to-Image Generation in the Era of Large Model
Figure 3 for RenAIssance: A Survey into AI Text-to-Image Generation in the Era of Large Model
Figure 4 for RenAIssance: A Survey into AI Text-to-Image Generation in the Era of Large Model
Viaarxiv icon

DeepSpeed-Chat: Easy, Fast and Affordable RLHF Training of ChatGPT-like Models at All Scales

Add code
Aug 02, 2023
Figure 1 for DeepSpeed-Chat: Easy, Fast and Affordable RLHF Training of ChatGPT-like Models at All Scales
Figure 2 for DeepSpeed-Chat: Easy, Fast and Affordable RLHF Training of ChatGPT-like Models at All Scales
Figure 3 for DeepSpeed-Chat: Easy, Fast and Affordable RLHF Training of ChatGPT-like Models at All Scales
Figure 4 for DeepSpeed-Chat: Easy, Fast and Affordable RLHF Training of ChatGPT-like Models at All Scales
Viaarxiv icon

ZeroQuant-FP: A Leap Forward in LLMs Post-Training W4A8 Quantization Using Floating-Point Formats

Add code
Jul 20, 2023
Figure 1 for ZeroQuant-FP: A Leap Forward in LLMs Post-Training W4A8 Quantization Using Floating-Point Formats
Figure 2 for ZeroQuant-FP: A Leap Forward in LLMs Post-Training W4A8 Quantization Using Floating-Point Formats
Figure 3 for ZeroQuant-FP: A Leap Forward in LLMs Post-Training W4A8 Quantization Using Floating-Point Formats
Figure 4 for ZeroQuant-FP: A Leap Forward in LLMs Post-Training W4A8 Quantization Using Floating-Point Formats
Viaarxiv icon