Blog

VQGAN+CLIP: Hallucinating Images from Text

pytorch
vqgan
clip
gan

CLIP can score how well an image matches a text prompt, but on its own it has no way to produce one. VQGAN+CLIP closes that gap by optimizing a latent vector fed through a pretrained VQGAN decoder, nudging the decoded image’s CLIP embedding toward a text prompt with gradient ascent. The decoder supplies a learned prior over realistic image structure: the optimizer never touches a pixel directly, only the latent that generates it.

Jul 11, 2026
21 min

Diffusion Models

pytorch
diffusion

A walk through the NVIDIA DLI workshop on generative AI with diffusion models, covering U-Net architecture, DDPM, iterative optimizations, and classifier-free guidance: with a closer look at the math and design choices the course glosses over.

Jul 3, 2026
15 min

Down the Rabbit Hole: Setting Up a Local LLM with llama.cpp

tech
LLM
hardware
My local LLM journey began with a graphics card, specifically an NVIDIA GeForce RTX 3090 with 24 GB of VRAM. I did not buy it new or unbox it from a retail package. Instead…
Jun 7, 2026
17 min

Reinforcement Learning: Tabular Q-Learning

code
python
mdp
reinforcement learning

I’ve decided to start learning reinforcement learning with Frozenlake game and Q-Learning.

Dec 17, 2025
4 min

Bernoulli Multi-Armed Bandit Problem

python
code
MAB
reinforcement learning

The multi-armed bandit (MAB) problem is a classic reinforcement learning problem that illustrates the exploitation-exploration dilemma.

Dec 16, 2025
18 min

Building a No-Fluff Report Template in LaTeX

LaTeX

On a few occasions, I had to write a LaTeX/PDF report or documentation on what I was doing. I often got annoyed with TeX templates, which had lots of wasted whitespace around the title and edges of the document. So, I created a template that improves density, flexibility, and clarity.

May 7, 2025
3 min

Visualizing Feature Maps from VGG11 and ResNet50 in PyTorch

pytorch
nn

In convolutional neural networks (CNNs), intermediate layers capture increasingly abstract representations of the input image. In this post, we visualize feature maps from VGG and ResNet using forward hooks in PyTorch.

May 6, 2025
3 min

Research Compute Infrastructure

compute
infrastructure

Our research lab needed a more scalable approach to sharing computing resources among researchers. To address this, we built a scalable computing infrastructure on Kubernetes. Researchers can request resources from the pool and use a familiar workflow with Jupyter Notebooks. Resources are returned to the pool once a task is complete or idle for too long.

Nov 20, 2023
26 min

Generative Adversarial Networks

pytorch
GAN

Generative Adversarial Networks (GANs) represent an innovative class of unsupervised neural networks that have revolutionized the field of artificial intelligence. Eager to learn how they work, I’ve implemented foundational “vanilla” GAN and its more complex counterpart, the Deep Convolutional GAN (DCGAN), from scratch. I’ve put them on a test run on MNIST Digits and Fashion toy datasets.

Oct 10, 2023
5 min

Neural Style Transfer

pytorch
NST

Neural Style Transfer (NST) was a breakthrough deep learning approach that can transfer artistic style from one image to another. Eager to learn how it works, I’ve implemented the original approach from scratch.

Sep 15, 2023
8 min
No matching items