Skip to yearly menu bar Skip to main content


Show Detail
Schedule Sun Tue Wed Thu
Timezone: America/New_York
 
Filter Rooms:  

MON 5 JUN
7:30 a.m.
Registration Desk
(ends 4:00 PM)
8 a.m.
Coffee Break
8:45 a.m.
Opening Remarks:
Opening Remarks
(ends 9:00 AM)
9 a.m.
Parallel and Distributed Systems 1: Parallelism [9:00-10:00]
PipeFisher: Efficient Training of Large Language Models Using Pipelining and Fisher Information Matrices
Tutel: Adaptive Mixture-of-Experts at Scale
Breadth-First Pipeline Parallelism
(ends 10:00 AM)
10 a.m.
Coffee Break
10:30 a.m.
Invited Talk:
Improving the Quality and Factuality of Large Language Model Applications
Matei Zaharia
(ends 11:30 AM)
11:30 a.m.
Lunch Break, on your own
12:45 p.m.
Round Table Discussion:
Round Table Discussion
(ends 1:30 PM)
1:30 p.m.
Memory Optimization [1:30-2:50]
Posters -
Reducing Activation Recomputation in Large Transformer Models
RevBiFPN: The Fully Reversible Bidirectional Feature Pyramid Network
Transcending Runtime-Memory Tradeoffs in Checkpointing by being Fusion Aware
Safe Optimized Static Memory Allocation for Parallel Deep Learning
(ends 2:50 PM)
2:50 p.m.
Coffee Break
3:20 p.m.
Correctness and Security [3:20-4:40]
Posters -
Validating Large Language Models with ReLM
SysNoise: Exploring and Benchmarking Training-Deployment System Inconsistency
Be Careful with PyPI Packages: You May Unconsciously Spread Backdoor Model Weights
Building Verified Neural Networks for Computer Systems with Ouroboros
(ends 4:40 PM)
4:40 p.m.
Sparsity 1: Models and Algorithms [4:40-5:40]
MegaBlocks: Efficient Sparse Training with Mixture-of-Experts
Uniform Sparsity in Deep Neural Networks
Cuttlefish: Low-Rank Model Training without All the Tuning
(ends 5:40 PM)
5:40 p.m.
Poster Session/Reception [5:40-8:00]
(ends 8:00 PM)

TUE 6 JUN
7:30 a.m.
Registration Desk
(ends 4:00 PM)
8:30 a.m.
Coffee Break
9 a.m.
Measurement and Analysis [9:00-10:00]
Efficiently Scaling Transformer Inference
Hotline Profiler: Automatic Annotation and A Multi-Scale Timeline for Visualizing Time-Use in DNN Training
ApproxCaliper: A Programmable Framework for Application-aware Neural Network Optimization
(ends 10:00 AM)
10 a.m.
Coffee Break
10:30 a.m.
Invited Talk:
Do we need Attention?
Alexander Rush
(ends 11:30 AM)
11:30 a.m.
Lunch Break, on your own
1:30 p.m.
Parallel and Distributed Systems 2: Communication [1:30-2:50]
Posters -
Cupcake: A Compression Scheduler for Scalable Communication-Efficient Distributed Training
Communication-Efficient Graph Neural Networks with Probabilistic Neighborhood Expansion Analysis and Caching
Adaptive Message Quantization and Parallelization for Distributed Full-graph GNN Training
On Optimizing the Communication of Model Parallelism
(ends 2:50 PM)
2:50 p.m.
Coffee Break
3:20 p.m.
Federated Learning [3:20-4:40]
Posters -
FedTree: A Federated Learning System For Trees
FLINT: A Platform for Federated Learning Integration
On Noisy Evaluation in Federated Hyperparameter Tuning
GlueFL: Reconciling Client Sampling and Model Masking for Bandwidth Efficient Federated Learning
(ends 4:40 PM)
4:40 p.m.
ML for Systems [4:40-6:00]
Posters -
AutoScratch: ML-Optimized Cache Management for Inference-Oriented GPUs
GiPH: Generalizable Placement Learning for Adaptive Heterogeneous Computing
Learning to Parallelize with OpenMP by Augmented Heterogeneous AST Representation
Virtual Machine Allocation with Lifetime Predictions
(ends 6:00 PM)

WED 7 JUN
7:30 a.m.
Registration Desk
(ends 4:00 PM)
8:30 a.m.
Coffee Break
9 a.m.
Compilers [9:00-10:00]
ALCOP: Automatic Load-Compute Pipelining in Deep Learning Compiler for AI-GPUs
SIRIUS: Harvesting Whole-Program Optimization Opportunities for DNNs
X-RLFLOW: GRAPH REINFORCEMENT LEARNING FOR NEURAL NETWORK SUBGRAPHS TRANSFORMATION
(ends 10:00 AM)
10 a.m.
Coffee Break
10:30 a.m.
Emerging Models and Domains [10:30-11:30]
XRBench: An Extended Reality (XR) Machine Learning Benchmark Suite for the Metaverse
Renee: END-TO-END TRAINING OF EXTREME CLASSIFICATION MODELS
HyperGef: A Framework Enabling Efficient Fusion for Hypergraph Neural Network on GPUs
(ends 11:30 AM)
11:30 a.m.
Lunch Break, on your own
1:30 p.m.
Sparsity 2: Systems [1:30-3:20]
Posters -
Efficient GPU Kernels for N:M-Sparse Weights in Deep Learning
Unified Convolution Framework: A compiler-based approach to support sparse convolutions
Exploiting Hardware Utilization and Adaptive Dataflow for Efficient Sparse Convolution in 3D Point Clouds
Sparsity-Aware Memory Interface Architecture using Stacked XORNet Compression for Accelerating Pruned-DNN Models
(ends 3:20 PM)
2:50 p.m.
Coffee Break
3:20 p.m.
Storage, Scheduling, and Networking [3:20-4:40]
Posters -
Pre-train and Search: Efficient Embedding Table Sharding with Pre-trained Neural Cost Models
μ-TWO: 3× Faster Multi-Model Training with Orchestration and Memory Optimization
PyTorch RPC: Distributed Deep Learning Built on Tensor-Optimized Remote Procedure Calls
RecD: Deduplication for End-to-End Deep Learning Recommendation Model Training Infrastructure
(ends 4:40 PM)
4:40 p.m.
Edge [4:40-5:40]
Practical Edge Kernels for Integer-Only Vision Transformers Under Post-training Quantization
Edge Impulse: An MLOps Platform for Tiny Machine Learning
SUBGRAPH STATIONARY HARDWARE-SOFTWARE INFERENCE CO-DESIGN
(ends 5:40 PM)

THU 8 JUN
7 a.m.
Registration Desk
(ends 12:00 PM)
8 a.m.
Workshop:
2nd Workshop on Practical Adoption Challenges of ML for Systems in Industry
(ends 5:00 PM)
Workshop:
Benchmarking Machine Learning Workloads on Emerging Hardware
(ends 5:00 PM)
8:30 a.m.
Workshop:
Research On Algorithms & Data Structures (ROADS) to Mega-AI Models
(ends 5:00 PM)
8:45 a.m.
Workshop:
Workshop on Decentralized and Collaborative Learning
(ends 6:20 PM)
8:50 a.m.
Workshop:
Workshop on Federated Learning Systems
(ends 5:05 PM)
9 a.m.
Workshop:
Resource-Constrained Learning in Wireless Networks
(ends 3:00 PM)
Workshop:
The 3rd On-Device Intelligence Workshop
(ends 5:15 PM)
9:50 a.m.
Workshop:
Workshop on Systems for Next-Gen AI Paradigms
(ends 5:00 PM)
10 a.m.
Coffee Break
3 p.m.
Coffee Break