Neeraj KumarMemory Service for Agentic ApplicationsLanguage models are stateless: they do not persist information across calls. In contrast, language agents may store and maintain…Oct 24Oct 24
Neeraj KumarLogsumexp trick and Flash attention- Part 2GPU memory hierarchy consists of various types of memory that differ in size and speed, where smaller memory is faster. For instance, the…Sep 5Sep 5
Neeraj KumarGENERATIVE LEARNING TRILEMMAAt present, generative learning frameworks are unable to meet three essential criteria simultaneously, which are often necessary for their…Apr 8, 2023Apr 8, 2023
Neeraj KumarWhy Contrastive loss for unsupervised learning.Unsupervised learning holds out the possibility that we can acquire transferable representations even in the absence of direct human…Sep 5, 2022Sep 5, 2022
Neeraj KumarUnderstanding Variational AutoencoderLet us consider some dataset X = {x (i)} N i=1 consisting of N i.i.d. samples of some continuous or discrete variable x. We assume that the…Aug 10, 2021Aug 10, 2021
Neeraj KumarDocument Ranking using BertDocument ranking is a well-known problem in the NLP domain which has prominent use cases across the various industries in extractive…Aug 7, 20211Aug 7, 20211
Neeraj Kumarknowledge distillation in deep learning — A mathematical perspectiveSome of the problems like object detection, speech recognition are common applications that are used in many industries. Several…Feb 21, 2021Feb 21, 2021