8 papers
Optimistic Dual Averaging Unifies Modern Optimizers
Thomas Pethick, Wanyun Xie, Roman Machacek +1
We introduce SODA, a generalization of Optimistic Dual Averaging, which provides a common perspective on state-of-the-art optimizers like Muon, Lion, AdEMAMix and NAdam, showing th…
MaD-Mix: Multi-Modal Data Mixtures via Latent Space Coupling for Vision-Language Model Training
Wanyun Xie, Francesco Tonin, Volkan Cevher
Vision-Language Models (VLMs) are typically trained on a diverse set of multi-modal domains, yet current practices rely on costly manual tuning. We propose MaD-Mix, a principled an…
Generalized Gradient Norm Clipping & Non-Euclidean -Smoothness
Thomas Pethick, Wanyun Xie, Mete Erdogan +3
This work introduces a hybrid non-Euclidean optimization method which generalizes gradient norm clipping by combining steepest descent and conditional gradient approaches. The meth…
Chameleon: A Flexible Data-mixing Framework for Language Model Pretraining and Finetuning
Wanyun Xie, Francesco Tonin, Volkan Cevher
Training data mixtures greatly impact the generalization performance of large language models. Existing domain reweighting methods often rely on costly weight computations and requ…
Training Deep Learning Models with Norm-Constrained LMOs
Thomas Pethick, Wanyun Xie, Kimon Antonakopoulos +3
In this work, we study optimization methods that leverage the linear minimization oracle (LMO) over a norm-ball. We propose a new stochastic family of algorithms that uses the LMO…
SAMPa: Sharpness-aware Minimization Parallelized
Wanyun Xie, Thomas Pethick, Volkan Cevher
Sharpness-aware minimization (SAM) has been shown to improve the generalization of neural networks. However, each SAM update requires \emph{sequentially} computing two gradients, e…
Improving SAM Requires Rethinking its Optimization Formulation
Wanyun Xie, Fabian Latorre, Kimon Antonakopoulos +2
This paper rethinks Sharpness-Aware Minimization (SAM), which is originally formulated as a zero-sum game where the weights of a network and a bounded perturbation try to minimize/…
Stable Nonconvex-Nonconcave Training via Linear Interpolation
Thomas Pethick, Wanyun Xie, Volkan Cevher
This paper presents a theoretical analysis of linear interpolation as a principled method for stabilizing (large-scale) neural network training. We argue that instabilities in the…