andersonbcdefg

dpo-lora

direct preference optimization with only 1 model copy :)

aiq

ai for jq

cramBERT

Training BERT with relatively few compute resources.

rewardmodeling

Train reward models for reinforcement learning from human feedback (RLHF).

gptemail

mlx-soft-moe

implementation of soft mixture-of-experts in mlx

promptbreeder-plus

andersonbcdefg

Config files for my GitHub profile.

andersonbcdefg.github.io

My webpages

bpt

bufferpiece

SentencePiece tokenizer that operates on utf-8 bytes.

vision-models

My implementation of recent cutting-edge computer vision models.

brrrotary-embedding

PyTorch rotary position embedding (RoPE) that goes brrrr.

contrastive_losses

functional implementations of contrastive losses

cs229-final-project

cs448b-final-project

CTranslate2

Fast inference engine for Transformer models

embedding-laser

compressing embedding models

ezblock

simple website blocking CLI tool leveraging etc/hosts

instruct-pythia-ptuning

Using p-tuning (a form of parameter-efficient fine-tuning) to tune Pythia models on natural language instructions.

llm_social_engineering

mlx-data

Efficient framework-agnostic data loading

mlx-examples

Examples in the MLX framework

podsearch

podsearch-client

shoggoth-chat

Chat with different alien entities in the latent space of gpt-3.5-turbo-0301

signals-ai

AI Journal

simlm

Large-scale Self-supervised Pre-training Across Tasks, Languages, and Modalities

stats205-final-project

t5