5 accepted papers
Large language models (LLMs) excel at capturing global token dependencies via self-attention but face prohibitive compute and memory costs on lengthy inputs. While sub-quadratic methods (e.g., linear attention) can reduce these costs, they often degrade accuracy due to overemphasizing recent tokens.…
The proliferation of large language models (LLMs) has led to the adoption of Mixture-of-Experts (MoE) architectures that dynamically leverage specialized subnetworks for improved efficiency and performance. Despite their benefits, MoE models face significant challenges during inference, including in…
Autoregressive Large Language Models (e.g., LLaMa, GPTs) are omnipresent achieving remarkable success in language understanding and generation. However, such impressive capability typically comes with a substantial model size, which presents significant challenges for autoregressive token-by-token g…
Training data and model sizes are increasing exponentially. One way to reduce training time and resources is to train with a carefully selected subset of the full dataset. Prior work uses the gradient signals obtained during a warm-up or ``pre-training" phase over the full dataset, for determining t…
Post-training quantization compresses a neural network within few hours with only a small unlabeled calibration set. However, so far it has been only discussed and empirically demonstrated in the context of uniform quantization on convolutional neural networks. We thus propose a new post-training no…