← Search

Karthick Gopalswamy

2 accepted papers

2024

Collage: Light-Weight Low-Precision Strategy for LLM Training

ICML 2024poster

Large models training is plagued by the intense compute cost and limited hardware memory. A practical solution is low-precision representation but is troubled by loss in numerical accuracy and unstable training rendering the model less useful. We argue that low-precision floating points can perform…

2024

Pessimistic Off-Policy Multi-Objective Optimization

AISTATS 2024poster

Multi-objective optimization is a class of optimization problems with multiple conflicting objectives. We study offline optimization of multi-objective policies from data collected by a previously deployed policy. We propose a pessimistic estimator for policy values that can be easily plugged into e…

Cited by 0SourcePDFScholar