Sayeh Sharify

Cnvlutin2: Ineffectual-Activation-and-Weight-Free Deep Neural Network Computing

2017 · arXiv (Cornell University)

We discuss several modifications and extensions over the previous proposed Cnvlutin (CNV) accelerator for convolutional and fully-connected layers of Deep Learning Network. We first describe different encodings of the activations that are deemed ineffectual. The …

Scaling Laws for Post Training Quantized Large Language Models

2024 · arXiv (Cornell University)

Generalization abilities of well-trained large language models (LLMs) are known to scale predictably as a function of model size. In contrast to the existence of practical scaling laws governing pre-training, the quality of LLMs after …