LLM training systems

Differentially Private Training in the LLM Era

Di Wang, Assistant Professor, Computer Science

Sep 28, 12:00 - 13:00

B9 R2325

privacy-preserving AI LLM LLM training systems stochastic gradient descent optimization

This seminar explores how to make differentially private training practical and scalable for large language models and new methods that enable stronger privacy guarantees without sacrificing performance.

Algorithmically Faithful and System-Efficient Optimization for Large Language Models

Liangyu Wang, Ph.D. Student, Computer Science

Sep 8, 13:30 - 16:30

B4 R5209; Zoom Meeting 99767080624

LLM Neural Network training algorithms Optimization for Machine Learning LLM training systems

The thesis develops an optimizer-aware systems view in which these methods are treated as structured computations with algorithmic invariants, memory behavior, communication objects, and scheduling constraints.