Improving Throughput by Optimising KV Cache Efficiency for Agentic Workloads

Boosting throughput by leveraging characteristics of agentic workloads to increase efficiency of the KV cache.

Fast vLLM: cutting cold starts by 20x

Reducing container start-up time from 117s to 6s, the long way round.

A Deep Dive into Diffusion Models: DDPM

An introduction to diffusion models with a focus on denoising diffusion probabilistic models (DDPM).

Exploring the theory behind Variational Autoencoders (VAEs)

How do VAEs work?