A rescaling-invariant Lipschitz bound based on path-metrics for modern ReLU network parameterizationsby Antoine Gonon, Nicolas…
PV-Tuning: Beyond Straight-Through Estimation for Extreme LLM Compressionby Vladimir Malinovskii, Denis Mazur, Ivan Ilin, Denis…
TerDiT: Ternary Diffusion Models with Transformersby Xudong Lu, Aojun Zhou, Ziyi Lin, Qi Liu, Yuhui…
MultiCast: Zero-Shot Multivariate Time Series Forecasting Using LLMsby Georgios Chatzigeorgakidis, Konstantinos Lentzos, Dimitrios SkoutasFirst submitted…
Integer Scale: A Free Lunch for Faster Fine-grained Quantization of LLMsby Qingyuan Li, Ran Meng,…
Mitigating Quantization Errors Due to Activation Spikes in GLU-Based LLMsby Jaewoo Yang, Hayun Kim, Younghoon…
ZipCache: Accurate and Efficient KV Cache Quantization with Salient Token Identificationby Yefei He, Luoming Zhang,…
Rate-Adaptive Quantization: A Multi-Rate Codebook Adaptation for Vector Quantization-based Generative Modelsby Jiwan Seo, Joonhyuk KangFirst…
eXmY: A Data Type and Technique for Arbitrary Bit Precision Quantizationby Aditya Agrawal, Matthew Hedlund,…
Communication-Efficient Federated Learning via Clipped Uniform Quantizationby Zavareh Bozorgasl, Hao ChenFirst submitted to arxiv on:…