299 citations · 813 across the 27 of their papers we have counts for
28 papers
SpenseGPT: Practical One-shot Pruning Enabling Sparse and Dense GEMMs for LLM Inference
Jaeseong Lee, Seung-won Hwang, Samyam Rajbhandari
Semi-structured 2:4 sparsity is widely supported by modern accelerators, providing up to a 2x theoretical speedup. However, its strict 50% sparsity constraint often causes non-negl…
FastKernels: Benchmarking GPU Kernel Generation in Production
Gabriele Oliaro, Yichao Fu, Jaeseong Lee +8
LLM-based agents for GPU kernel generation are advancing rapidly, but the benchmarks they optimize against evaluate kernels in isolation, with synthetic inputs and weak baselines,…
MoE-Prefill: Zero Redundancy Overheads in MoE Prefill Serving
Zhaoyuan Su, Olatunji Ruwase, Karthik Ganesan +5
Production LLM workloads increasingly serve discriminative tasks, such as classification, recommendation, and verification, whose answers are read from the logits of a single prefi…
Fast and Accurate Causal Parallel Decoding using Jacobi Forcing
Lanxiang Hu, Siqi Kou, Yichao Fu +5
Multi-token generation has emerged as a promising paradigm for accelerating transformer-based large model inference. Recent efforts primarily explore diffusion Large Language Model…
OWL: Overcoming Window Length-Dependence in Speculative Decoding for Long-Context Inputs
Jaeseong Lee, seung-won hwang, Aurick Qiao +3
Speculative decoding promises faster inference for large language models (LLMs), yet existing methods fail to generalize to real-world settings. Benchmarks typically assume short c…
Shift Parallelism: Low-Latency, High-Throughput LLM Inference for Dynamic Workloads
Mert Hidayetoglu, Aurick Qiao, Michael Wyatt +3
Efficient parallelism is necessary for achieving low-latency, high-throughput inference with large language models (LLMs). Tensor parallelism (TP) is the state-of-the-art method fo…
Arctic Inference with Shift Parallelism: Fast and Efficient Open Source Inference System for Enterprise AI
Samyam Rajbhandari, Mert Hidayetoglu, Aurick Qiao +5
Inference is now the dominant AI workload, yet existing systems force trade-offs between latency, throughput, and cost. Arctic Inference, an open-source vLLM plugin from Snowflake…
Arctic Long Sequence Training: Scalable And Efficient Training For Multi-Million Token Sequences
Stas Bekman, Samyam Rajbhandari, Michael Wyatt +5
Long sequences are critical for applications like RAG, long document summarization, multi-modality, etc., and modern LLMs, like Llama 4 Scout, support max sequence length of up to…
SwiftKV: Fast Prefill-Optimized Inference with Knowledge-Preserving Model Transformation
Aurick Qiao, Zhewei Yao, Samyam Rajbhandari +1
LLM inference for enterprise applications, such as summarization, RAG, and code-generation, typically observe much longer prompt than generations, leading to high prefill cost and…
DeepSpeed-FastGen: High-throughput Text Generation for LLMs via MII and DeepSpeed-Inference
Connor Holmes, Masahiro Tanaka, Michael Wyatt +8
The deployment and scaling of large language models (LLMs) have become critical as they permeate various applications, demanding high-throughput and low-latency serving systems. Ex…
DeepSpeed Ulysses: System Optimizations for Enabling Training of Extreme Long Sequence Transformer Models
Sam Ade Jacobs, Masahiro Tanaka, Chengming Zhang +4
Computation in a typical Transformer-based large language model (LLM) can be characterized by batch size, hidden dimension, number of layers, and sequence length. Until now, system…
DeepSpeed-VisualChat: Multi-Round Multi-Image Interleave Chat via Multi-Modal Causal Attention
Zhewei Yao, Xiaoxia Wu, Conglong Li +6
Most of the existing multi-modal models, hindered by their incapacity to adeptly manage interleaved image-and-text inputs in multi-image, multi-round dialogues, face substantial co…
DeepSpeed-Chat: Easy, Fast and Affordable RLHF Training of ChatGPT-like Models at All Scales
Zhewei Yao, Reza Yazdani Aminabadi, Olatunji Ruwase +16
ChatGPT-like models have revolutionized various applications in artificial intelligence, from summarization and coding to translation, matching or even surpassing human performance…
ZeRO++: Extremely Efficient Collective Communication for Giant Model Training
Guanhua Wang, Heyang Qin, Sam Ade Jacobs +6
Zero Redundancy Optimizer (ZeRO) has been used to train a wide range of large language models on massive GPUs clusters due to its ease of use, efficiency, and good scalability. How…
A Hybrid Tensor-Expert-Data Parallelism Approach to Optimize Mixture-of-Experts Training
Siddharth Singh, Olatunji Ruwase, Ammar Ahmad Awan +3
Mixture-of-Experts (MoE) is a neural network architecture that adds sparsely activated expert blocks to a base model, increasing the number of parameters without impacting computat…
BLOOM: A 176B-Parameter Open-Access Multilingual Language Model
BigScience Workshop, :, Teven Le Scao +391
Large language models (LLMs) have been shown to be able to perform new tasks based on a few demonstrations or natural language instructions. While these capabilities have led to wi…
DeepSpeed Inference: Enabling Efficient Inference of Transformer Models at Unprecedented Scale
Reza Yazdani Aminabadi, Samyam Rajbhandari, Minjia Zhang +8
The past several years have witnessed the success of transformer-based models, and their scale and application scenarios continue to grow aggressively. The current landscape of tra…
Using DeepSpeed and Megatron to Train Megatron-Turing NLG 530B, A Large-Scale Generative Language Model
Shaden Smith, Mostofa Patwary, Brandon Norick +17
Pretrained general-purpose language models can achieve state-of-the-art accuracies in various natural language processing domains by adapting to downstream tasks via zero-shot, few…
DeepSpeed-MoE: Advancing Mixture-of-Experts Inference and Training to Power Next-Generation AI Scale
Samyam Rajbhandari, Conglong Li, Zhewei Yao +5
As the training of giant dense models hits the boundary on the availability and capability of the hardware resources today, Mixture-of-Experts (MoE) models become one of the most p…
Scalable and Efficient MoE Training for Multitask Multilingual Models
Young Jin Kim, Ammar Ahmad Awan, Alexandre Muzio +6
The Mixture of Experts (MoE) models are an emerging class of sparsely activated deep learning models that have sublinear compute costs with respect to their parameters. In contrast…
ZeRO-Infinity: Breaking the GPU Memory Wall for Extreme Scale Deep Learning
Samyam Rajbhandari, Olatunji Ruwase, Jeff Rasley +2
In the last three years, the largest dense deep learning models have grown over 1000x to reach hundreds of billions of parameters, while the GPU memory has only grown by 5x (16 GB…
1-bit LAMB: Communication Efficient Large-Scale Large-Batch Training with LAMB's Convergence Speed
Conglong Li, Ammar Ahmad Awan, Hanlin Tang +2
To train large models (like BERT and GPT-3) on hundreds of GPUs, communication has become a major bottleneck, especially on commodity systems with limited-bandwidth TCP network. On…
1-bit Adam: Communication Efficient Large-Scale Training with Adam's Convergence Speed
Hanlin Tang, Shaoduo Gan, Ammar Ahmad Awan +6
Scalable training of large models (like BERT and GPT-3) requires careful optimization rooted in model design, architecture, and system capabilities. From a system standpoint, commu…
ZeRO-Offload: Democratizing Billion-Scale Model Training
Jie Ren, Samyam Rajbhandari, Reza Yazdani Aminabadi +5
Large-scale model training has been a playing ground for a limited few requiring complex model refactoring and access to prohibitively expensive GPU clusters. ZeRO-Offload changes…