David Heineman

🌾 Universal, customizable and deployable fine-grained evaluation for text generation.

Interactive rubiks cube solver with LLMs (AGI House Open Source Hackathon 2nd Place)

Success and Failure Linguistic Simplification Annotation 💃

solver for berghain challenge

Beam search & sampling visualization plugin for HuggingFace.

The innards of my website.

the Beaker tooling you always wanted

search the anthology with ColBERT

Code and data for "Improving Minimum Bayes Risk Decoding with Multi-Prompt" at EMNLP 2024.

👨‍💻 Powerful text simplification ranking, rating and span fixing interface

LoRA adapter that encodes my email

daily research from the arXiv

No description.

No description.

Deviously useful utilities

my macos setup

converts BPE models to byte-level models for a correct BPB measure

GRPO in \le 500 lines of Python

Tools for finding interesting papers in large conference proceedings (COLM 25, NeurIPS 25, ...)

No description.

(100% generated by cursor) alpha zero for chess

6.5x speedup to MolmoBot inference code (graph capture, torch compile, flash attention, kv caching)

data mixing for RL

migrate olmo eval tasks with claude code

Bipelines: pipelines for beaker

the ai2 meat markov chain

Python SDK for trmnl.com

A benchmark for LLMs on complicated tasks in the terminal

No description.

No description.

run ShareGPT throughput benchmark on Hazy's megakernels

Build beaker://davidh/davidh-interactive

Tinker RL trainer for papergym

eval MoE experts with Shapley values

A high-throughput and memory-efficient inference and serving engine for LLMs

Mini version of https://github.com/test-time-training/discover

envs for paper replication

scripts for olmo baseline exps

a simple eval library

Open-source framework for the research and development of foundation models.

Legible, Scalable, Reproducible Foundation Models with Named Tensors and Jax

Named Tensors for Legible Deep Learning in JAX

Docker service for LLM environments

No description.

Replication of an ML model in concrete mixing prediction

Compute MMI between text sets with zstd compression

Annotate traces from thinking LLMs

OAI's Preparedness Team's Eval Harness

+- 2% on Minerva 500 by changing vLLM version

Compute branching factor using vLLM

Nano vLLM

Training tokenizers

Dataset of NSF awards 1960-present

HF Dataset for Jane Street Puzzles

better pre-training evaluation

Pull BlueSky data with atproto

D3.js coffee map w/ Google Maps API

CUDA/C++ implementation of Fast Feed-Forward NNs

Viewer for WIMBD (What's in My Big Data)

Code for "Towards a path dependent account of category fluency" at CogSci 2024

Estimate the quality of instruction fine-tuning data

No description.

Real-time video stereoscopy implemented in CUDA

My Solutions for GT's GreyHat CTF 2022

No description.

Complaint Explorer for Environmental Protection Division of Georgia (EPD)

Minimal PyTorch re-implementation

E First Technical

Create a photo mosaic

Host a LLaMA model with an API endpoint using Flask

📱 Web demo and API of a hashtag segmentation model, HashtagMaster

No description.

📝 Implementation of Essay Scoring Models

No description.

🎶 NFC-based Raspberry Pi Spotify Player

🚩 Phrase-level error annotation interface for text simplification

No description.

🔎 Logic parsing using 🌲 dependency trees

📝 Notes for all my classes at Tech

🎄 Advent of Code '21

Data Generation & Annotation for Lexical Simplification Pipeline Testing

🎥 Face alignment tool for generating timelapses

💬 English version of lexical simplification model capable of phrase substitutions

Finds the avg GPA and difficulty of any Georgia Tech class schedule

Chrome extension to apply Plaza.one to chrome newtab. Includes various goodies & additions.