Andy Lo's repositories
cuda-mnist
Training MLP on MNIST in 2 seconds with pure CUDA
simpleGEMM
The simplest but fast implementation of matrix multiplication in CUDA.
jax-diffusion
Implementation of Denoising Diffusion Probabilistic Models (DDPM) in JAX and Flax.
Text2Emoji
Text to emoji translation via DeepDreaming
jax-vqvae-gpt
Implementation of VQ-VAE with a GPT-style sampler in the JAX and Haiku ecosystem.
ml-weight-decay
Code for exploring effects of weight decay on neural network generalization
YourMumBot
YourMumBot is a discord bot that makes your mum jokes based on content sent by users.
score-sde-inverse
Solving all kinds of (linear) inverse problems with diffusion models.
jax-logging
Imperative-like logging for JAX
Learn-CUDA
Learning how to write CUDA
speedtoxify
Fast :speak_no_evil: Detoxify inference with ONNX runtime
adaptive-generation
Adaptive-compute generative modeling. Combining PonderNet and VAE/Diffusion models for better computation efficiency. Undergraduate dissertation @Cambridge.
datasets
TFDS is a collection of datasets ready to use with TensorFlow, Jax, ...
sparse-dropout
SparseDrop is a simple, structured, and hardware-friendly variant of dropout that can benefit from sparsity on GPUs.
cutlass
CUDA Templates for Linear Algebra Subroutines
gpt-neox
An implementation of model parallel autoregressive transformers on GPUs, based on the DeepSpeed library.
ivy
The Unified Machine Learning Framework
jax-monads
Playing around with monads in JAX
llm-ol
Building a multilingual machine ontologist with language models
nanoGPT
The simplest, fastest repository for training/finetuning medium-sized GPTs.
nanotron
Minimalistic large language model 3D-parallelism training
prolog-tic-tac-toe
Optimal Tic-Tac-Toe player implemented in Prolog
tic-tac-gpt
Interpreting sequence models of Tic Tac Toe games.
transformer-lens-docker
Base docker image for Mech Intern research