We walk through the architecture, dataflow, and some design decisions of the vLLM library.

1

We investigate the internals of ollama, a widely-adored local LLM inference management platform built atop ggml.

2

On-device, low-latency language model inference have become increasingly important in designing production systems. Here, we dive deep into one leading framework for performant inference.

3

Undirected graphical models formed a large part of the initial push for machine intelligence, and remain relevant today. Here, I motivate and derive Monte Carlo-based learning algorithms for such models.

4

We motivate and derive the generalized backpropagation algorithm for arbitrarily structured networks.

5

We discuss the fundamental ideas behind fully convolutional networks, including the transformation of fully connected layers to convolutional layers and upsampling via transposed convolutions ("deconvolutions").

6

We motivate and derive the backpropagation learning algorithm for convolutional networks.

7

We motivate and derive the backpropagation learning algorithm for feedforward networks.

8

A detailed look at one of my favorite software security papers, and its implications on bootstrapping trust.

9

Directed latent variable models provide a powerful way to represent complex distributions by combining simple ones. However, they often have intractable log-likelihoods, yielding complicated learning algorithms. In this post, we build intuition for these concepts.

10