Lucas van Walstijn - Another Machine Learning Blog

Categories

All (20)

Adam (1)

PyTorch (1)

RMSProp (1)

Resnet (1)

activations (1)

attention (1)

axolotl (1)

batch normalization (1)

bptt (1)

collation (1)

competition (1)

computer vision (5)

convolutions (1)

cross entropy (1)

data augmentation (1)

dataloading (1)

ddpm (1)

diffusers (3)

embeddings (2)

fine-tuning (1)

foundations (2)

generative (7)

initialization (1)

jarvis (1)

langchain (1)

langsmith (1)

learning rate finder (1)

learning rate scheduler (1)

linear algebra (1)

llama3 (1)

llm (2)

loss functions (1)

lstm (1)

math (1)

mlops (1)

mlp (1)

momentum (2)

n-gram (1)

nll (1)

nlp (5)

nn.Module (1)

ollama (1)

rnn (2)

sampler (1)

setup (2)

softmax (1)

stable diffusion (3)

subscribers (1)

training (2)

transformers (1)

vectorization (1)

weight decay (1)

Having fun with llama3:8b

llm

ollama

llama3

langchain

langsmith

In this blog post I want to play around with Meta’s Llama-3:8b (instruct/chat) model, not so much as a way to rigorously benchmark it, but more as a way to get familiar with…

Lucas van Walstijn

LLM fine-tuning 101

llm

fine-tuning

axolotl

jarvis

As part of the conference Mastering LLMs: A Conference For Developers & Data Scientists, I wanted to document the process of fine-tuning my very first model.

Lucas van Walstijn

Transformers

nlp

transformers

attention

Transformers were introduced in the paper Attention is all you need in 2017 by Vaswani et al, and as the title suggests focusses heavily on a something called attention.…

Lucas van Walstijn

Denoising Diffusion Probabilistic Models

ddpm

generative

stable diffusion

diffusers

In this post, we will return to the topic of stable diffusion and discuss the Denoising Diffusion Probabilitic Models (DPPM) paper by Ho et al.

Lucas van Walstijn

Long short-term memory

nlp

rnn

lstm

generative

In the previous post we went in detail through Recurrent Neural Networks (RNN’s). We looked at how they are similar to MLPs and how they allow for a large improvement in…

Lucas van Walstijn

Recurrent Neural Networks

nlp

rnn

bptt

generative

In the previous post we went through a simple MLP language model. We looked in detail at embeddings, and how they can be used to overcome the curse of dimensionality encounte…

Lucas van Walstijn

Multilayer Perceptron language model

embeddings

mlp

nlp

generative

In the previous post we had a first look at language models and discussed the n-gram model. Specifically, we:

Lucas van Walstijn

N-gram language models

n-gram

nlp

embeddings

generative

In this post, I’ll discuss a very simple language model: n-grams! To keep things simple we will be concerned with single words, and not for example complete sentences. The…

Lucas van Walstijn

`nntrain` (4/n): Accelerated optimization

weight decay

momentum

RMSProp

Adam

Resnet

learning rate scheduler

data augmentation

computer vision

In this series, I want to discuss the creation of a small PyTorch based library for training neural networks: nntrain. It’s based off the excellent part 2 of Practical Deep…

Lucas van Walstijn

`nntrain` (3/n): Activations, Initialization and Normalization

convolutions

activations

initialization

batch normalization

computer vision

In this series, I want to discuss the creation of a small PyTorch based library for training neural networks: nntrain. It’s based off the excellent part 2 of Practical Deep…

Lucas van Walstijn

`nntrain` (2/n): Learner

training

momentum

subscribers

learning rate finder

computer vision

In this series, I want to discuss the creation of a small PyTorch based library for training neural networks: nntrain. It’s based off the excellent part 2 of Practical Deep…

Lucas van Walstijn

`nntrain` (1/n): Datasets and Dataloaders

dataloading

training

collation

sampler

In this series, I want to discuss the creation of a small PyTorch based library for training neural networks: nntrain. It’s based off the excellent part 2 of Practical Deep…

Lucas van Walstijn

`nntrain` (0/n): Preliminaries

foundations

PyTorch

nn.Module

In this series, I want to discuss the creation of a small PyTorch based library for training neural networks: nntrain. It’s based off the excellent part 2 of Practical Deep…

Lucas van Walstijn

Introduction to Stable Diffusion - Code

generative

stable diffusion

diffusers

computer vision

In the previous blog post, the main components and some intuition behind Stable Diffusion were introduced. Now, let’s see how we can use the HuggingFace diffusers library to…

Lucas van Walstijn

Introduction to Stable Diffusion - Concepts

generative

stable diffusion

diffusers

computer vision

Stable Diffusion, a generative deep learning algorithm developed in 2022, is capable of creating images from prompts. For example, when presented the prompt: A group of…

Lucas van Walstijn

Cross entropy any which way

loss functions

softmax

nll

cross entropy

Cross entropy is one of the most commonly used loss functions. In this post, we will have a look at how it works, and compute it in a couple of different ways.

Lucas van Walstijn

Fast matrix multiplications

foundations

math

vectorization

linear algebra

Matrix multiplications are kind of boring, so why write a blog post about them? Well, matrix multiplications are the most basic computation that is being performed by neural…

Lucas van Walstijn

First competition🏅

competition

In the past couple of weeks I have participated in the first ever Hugging Face competition: aiornot. And as a matter of fact, it was also my first competition to participate…

Lucas van Walstijn

Paperspace setup

setup

mlops

Most people don’t have a GPU installed in their working machine that is suited for Deep Learning, and in fact you don’t need to. It’s quite easy to setup a remote GPU server…

Lucas van Walstijn

Blog setup

setup

In this blog post I’ll explain how I created this blog, using Quarto and GitHub.

Lucas van Walstijn