19 citations · 24 across the 22 of their papers we have counts for
13 papers · 1 filter
When Geometry Aligns: Dihedral Hidden-State Transformations in UNet, ViT, and DiT Architectures
Mojtaba Faramarzi, Alex Lamb, Irina Rish
Diffusion architectures now encompass convolutional UNets as well as transformer-based designs such as Diffusion Transformers (DiTs), inspired by Vision Transformers (ViTs), yet th…
LLM Pretraining Shapes a Generalizable Manifold: Insights into Cross-Modal Transfer to Time Series
Alexis Roger, Prateek Humane, Zhenghan Tai +4
Can language-pretrained transformers become effective time-series forecasters, and why? In this paper, we show that cross-modal transfer arises because language pretraining precond…
Adaptive Batch Sizes Using Non-Euclidean Gradient Noise Scales for Stochastic Sign and Spectral Descent
Hiroki Naganuma, Shagun Gupta, Youssef Briki +4
To maximize hardware utilization, modern machine learning systems typically employ large constant or manually tuned batch size schedules, relying on heuristics that are brittle and…
Warming Up for Zeroth-Order Federated Pre-Training with Low Resource Clients
Gwen Legate, Irina Rish, Eugene Belilovsky
Federated learning enables collaborative model training across numerous edge devices without requiring participants to share data; however, memory and communication constraints on…
A Guide to Robust Generalization: The Impact of Architecture, Pre-training, and Optimization Strategy
Maxime Heuillet, Rishika Bhagwatkar, Jonas Ngnawé +6
Deep learning models operating in the image domain are vulnerable to small input perturbations. For years, robustness to such perturbations was pursued by training models from scra…
Training Dynamics Underlying Language Model Scaling Laws: Loss Deceleration and Zero-Sum Learning
Andrei Mircea, Supriyo Chakraborty, Nima Chitsazan +4
This work aims to understand how scaling improves language models, specifically in terms of training dynamics. We find that language models undergo loss deceleration early in train…
MuLoCo: Muon is a practical inner optimizer for DiLoCo
Benjamin Thérien, Xiaolong Huang, Aaron Defazio +2
DiLoCo is a powerful framework for training large language models (LLMs), enabling larger optimal batch sizes and increased accelerator utilization under networking constraints. Ho…
Continual Pre-training of MoEs: How robust is your router?
Benjamin Thérien, Charles-Étienne Joseph, Zain Sarwar +7
Sparsely-activated Mixture of Experts (MoE) transformers are promising architectures for foundation models. Compared to dense transformers that require the same amount of floating-…
Beyond Cosine Decay: On the effectiveness of Infinite Learning Rate Schedule for Continual Pre-training
Vaibhav Singh, Paul Janson, Paria Mehrbod +4
The ever-growing availability of unlabeled data presents both opportunities and challenges for training artificial intelligence systems. While self-supervised learning (SSL) has em…
Context is Key: A Benchmark for Forecasting with Essential Textual Information
Andrew Robert Williams, Arjun Ashok, Étienne Marcotte +8
Forecasting is a critical task in decision-making across numerous domains. While historical numerical data provide a start, they fail to convey the complete context for reliable an…
Spectra: Surprising Effectiveness of Pretraining Ternary Language Models at Scale
Ayush Kaushal, Tejas Vaidhya, Arnab Kumar Mondal +3
Rapid advancements in GPU computational power has outpaced memory capacity and bandwidth growth, creating bottlenecks in Large Language Model (LLM) inference. Post-training quantiz…
LO: Compute-Efficient Meta-Generalization of Learned Optimizers
Benjamin Thérien, Charles-Étienne Joseph, Boris Knyazev +3
Learned optimizers (LOs) have the potential to significantly reduce the wall-clock training time of neural networks. However, they can struggle to optimize unseen tasks (meta-gener…
Simple and Scalable Strategies to Continually Pre-train Large Language Models
Adam Ibrahim, Benjamin Thérien, Kshitij Gupta +5
Large language models (LLMs) are routinely pre-trained on billions of tokens, only to start the process over again once new data becomes available. A much more efficient solution i…