Skip to yearly menu bar Skip to main content


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

MON 13 MAY
7 a.m.
Registration Check-in Desk
(ends 5:00 PM)
8:30 a.m.
Break:
Breakfast (for Young Professionals Only)
(ends 9:00 AM)
9 a.m.
Opening Remarks (Young Professional Symposium):
Opening Remarks
(ends 9:05 AM)
9:05 a.m.
Invited Talk:
GenAI Efficiency is About More than Models
Kurt Keutzer
(ends 10:05 AM)
10:10 a.m.
Talk:
Scaling Intelligence
(ends 10:30 AM)
10:30 a.m.
Talk:
Towards Fast and Affordable Serving Systems for Large Language Models
(ends 10:50 AM)
10:50 a.m.
Talk:
Memory-Efficient LLM Training
(ends 11:10 AM)
11:10 a.m.
Break:
Coffee Break
(ends 11:30 AM)
11:30 a.m.
Talk:
Fairness in LLM Serving
(ends 11:50 AM)
11:50 a.m.
Talk:
Hardware-aware algorithms for sequence modeling
(ends 12:10 PM)
12:10 p.m.
Talk:
The Unreasonable Power of Synthetics for Efficient Machine Learning
(ends 12:30 PM)
12:30 p.m.
Break:
Lunch Break
(ends 2:00 PM)
2 p.m.
Lightning Talks:
Sponsor Lightning Talks
(ends 3:00 PM)
3 p.m.
Break:
Coffee Break
(ends 3:30 PM)
3:30 p.m.
Round Table Discussion:
Round Table
(ends 4:15 PM)
4:15 p.m.
Panel:
Discussion Panel
(ends 5:00 PM)
5 p.m.
Student Poster Session [5:00-6:30]
(ends 6:30 PM)

TUE 14 MAY
7 a.m.
Registration Check-in Desk
(ends 5:00 PM)
8:30 a.m.
Break:
Coffee Break
(ends 9:00 AM)
8:45 a.m.
Opening Remarks (Main Conference):
Opening Remarks
(ends 9:00 AM)
9 a.m.
Quantization and Compression 1 [9:00-10:00]
AWQ: Activation-aware Weight Quantization for On-Device LLM Compression and Acceleration
QMoE: Sub-1-Bit Compression of Trillion Parameter Models
Atom: Low-Bit Quantization for Efficient and Accurate LLM Serving
(ends 10:00 AM)
10 a.m.
Break:
Coffee Break
(ends 10:30 AM)
10:30 a.m.
Invited Talk:
Possible Impossibilities and Impossible Possibilities
Yejin Choi
(ends 11:30 AM)
11:30 a.m.
Break:
Lunch Break
(ends 1:00 PM)
1:30 p.m.
Large Language Models 1 [1:30-3:00]
Q-Hitter: A Better Token Oracle for Efficient LLM Inference via Sparse-Quantized KV Cache
Fine-Tuning Language Models Using Formal Methods Feedback: A Use Case in Autonomous Systems
Punica: Multi-Tenant LoRA Serving
SLoRA: Scalable Serving of Thousands of LoRA Adapters
(ends 3:00 PM)
3 p.m.
Break:
Coffee Break
(ends 3:30 PM)
3:30 p.m.
Parallel and Distributed 1 [3:30-4:30]
DiffusionPipe: Training Large Diffusion Models with Efficient Pipelines
Distributed Matrix-Based Sampling for Graph Neural Network Training
L-GreCo: Layerwise-adaptive Gradient Compression For Efficient Data-parallel Deep Learning
(ends 4:30 PM)
4:30 p.m.
Privacy and security [4:30-5:30]
Accelerating ReLU for MPC-Based Private Inference with a Communication-Efficient Sign Estimation
ACCURATE LOW-DEGREE POLYNOMIAL APPROXIMATION OF NON-POLYNOMIAL OPERATORS FOR FAST PRIVATE INFERENCE IN HOMOMORPHIC ENCRYPTION
Proteus: Preserving Model Confidentiality during Graph Optimizations
(ends 5:30 PM)
5:30 p.m.
Reception:
Reception & Poster Session
(ends 8:00 PM)

WED 15 MAY
7 a.m.
Registration Check-in Desk
(ends 5:00 PM)
8:30 a.m.
Break:
Coffee Break
(ends 9:00 AM)
9 a.m.
LLM 2 [9:00-10:00]
FlashDecoding++: Faster Large Language Model Inference with Asynchronization, Flat GEMM Optimization, and Heuristics
Prompt Cache: Modular Attention Reuse for Low-Latency Inference
Keyformer: KV Cache reduction through key tokens selection for Efficient Generative Inference
(ends 10:00 AM)
10 a.m.
Break:
Coffee Break
(ends 10:30 AM)
10:30 a.m.
Invited Talk:
AI Robustness and Security in the Age of LLMs
J. Zico Kolter
(ends 11:30 AM)
11:30 a.m.
Break:
Lunch Break
(ends 1:00 PM)
1:30 p.m.
Quantization and Compression 2 [1:30-3:00]
JIT-Q: Just-in-time Quantization with Processing-In-Memory for Efficient ML Training
Torch2Chip: An End-to-end Customizable Deep Neural Network Compression and Deployment Toolkit for Prototype Hardware Accelerator Design
Schrodinger's FP Training Neural Networks with Dynamic Floating-Point Containers
Efficient Post-training Quantization with FP8 Formats
(ends 3:00 PM)
3 p.m.
Break:
Coffee Break
(ends 3:30 PM)
3:30 p.m.
Federated Learning [3:30-4:30]
FedTrans: Efficient Federated Learning via Multi-Model Transformation
HeteroSwitch: Characterizing and Taming System-Induced Data Heterogeneity in Federated Learning
LIFL: A Lightweight, Event-driven Serverless Platform for Federated Learning
(ends 4:30 PM)
4:30 p.m.
Parallel and Distributed 2 [4:30-5:30]
Lancet: Accelerating Mixture-of-Experts Training by Overlapping Weight Gradient Computation and All-to-All Communication
Disaggregated Multi-Tower: Topology-aware Modeling Technique for Efficient Large Scale Recommendation
HeteGen: Efficient Heterogeneous Parallel Inference for Large Language Models on Resource-Constrained Devices
(ends 5:30 PM)

THU 16 MAY
7 a.m.
Registration Check-in Desk
(ends 5:00 PM)
8:30 a.m.
Break:
Coffee Break
(ends 9:00 AM)
9 a.m.
Performance and Memory [9:00-10:00]
vMCU: Coordinated Memory Management and Kernel Optimization for DNN Inference on MCUs
SiDA: Sparsity-Inspired Data-Aware Serving for Efficient and Scalable Large Mixture-of-Experts Models
ACROBAT: Optimizing Auto-batching of Dynamic Deep Learning at Compile Time
(ends 10:00 AM)
10 a.m.
Break:
Coffee Break
(ends 10:30 AM)
10:30 a.m.
Invited Talk:
Exciting Directions in Systems for Machine Learning
Jeff Dean
(ends 11:30 AM)
11:30 a.m.
Break:
Lunch Break
(ends 1:00 PM)
1:30 p.m.
Measurement and Analysis [1:30-3:00]
CloudEval-YAML: A Practical Benchmark for Cloud Configuration Generation
Does Compressing Activations Help Model Parallel Training?
COMET: Neural Cost Model Explanation Framework
VIDUR: A LARGE-SCALE SIMULATION FRAMEWORK FOR LLM INFERENCE
(ends 3:00 PM)
3 p.m.
Break:
Coffee Break
(ends 3:30 PM)
3:30 p.m.
ML for Systems [3:30-4:50]
On Latency Predictors for Neural Architecture Search
FLASH: Fast Model Adaptation in ML-Centric Cloud Platforms
VQPy: An Object-Oriented Approach to Modern Video Analytics
UniDM: A Unified Framework for Data Manipulation with Large Language Models
(ends 4:50 PM)
5 p.m.
Closing Remarks:
Closing Remarks
(ends 5:15 PM)