parallelization
Parallelization is a computational technique where tasks are executed simultaneously across multiple processing units. This is vital for scaling up training processes in AI, particularly for large models and datasets.
- A Near-optimal, Scalable and Parallelizable Framework for Stochastic Bandits Robust to Adversarial Corruptions and Beyond
- Efficient Preference-Based Reinforcement Learning: Randomized Exploration meets Experimental Design
- FFN Fusion: Rethinking Sequential Computation in Large Language Models
- Mozart: Modularized and Efficient MoE Training on 3.5D Wafer-Scale Chiplet Architectures
- Practical Kernel Selection for Kernel-based Conditional Independence Test
- SHAP values via sparse Fourier representation
- SPRINT: Enabling Interleaved Planning and Parallelized Execution in Reasoning Models