julian-q

kgene

Knowledge graph extraction from webcrawled plaintext.

facecam

auto-compile

BIGCHAIR

BIGCHAIR: BImodal Graph-CHAracter learnIng for Retrieval. Stanford CS 224N final project by Pino Cholsaipant, Sidney Hough, + Julian Quevedo.

dotfiles

hgcn

Increased-Efficiency Hyperbolic Graph Convolutional Networks with PyTorch Geometric.

laser-control

Automatic laser power stabilization using a motorized waveplate mount.

fmri-analysis

Graph Attention Transformer Encoder: QKV attention mechanism for GNNs.

GradCache

Run Effective Large Batch Contrastive Learning Beyond GPU/TPU Memory Constraint

hugo-PaperMod

A fast, clean, responsive Hugo theme.

policy-gradient

Reinforcement Learning: A NumPy implementation of two simple policy gradient algorithms.

ArCHer

Research Code for "ArCHer: Training Language Model Agents via Hierarchical Multi-Turn RL"

attention_experiments

composer

Train neural networks up to 7x faster

cs140e-24win

all class materials for 140e

cs149gpt

FasterTransformer

Transformer related optimization, including BERT, GPT

FlexGen

Running large language models on a single GPU for throughput-oriented scenarios.

gemm

Improving GEMM performance with language models

graph-meta-rl-for-amod

Official implementation of "Graph Meta-Reinforcement Learning for TransferableAutonomous Mobility-on-Demand"

GraphGPT

Extrapolating knowledge graphs from unstructured text using GPT-3 🕵️‍♂️

kliu.io

A blog.

llm-awq

AWQ: Activation-aware Weight Quantization for LLM Compression and Acceleration

object-oriented-NN

Object-oriented neural network built from scratch with NumPy.

plants

Exploring Lindenmayer systems. BIO 20Q project by Pino Cholsaipant + Julian Quevedo.

pytorch_geometric

Graph Neural Network Library for PyTorch

sketches

programmatic animations with processing

smoothquant

SmoothQuant: Accurate and Efficient Post-Training Quantization for Large Language Models

vector-fields

A simple vector field / differential equation plotter with Euler's method solution trajectories.

vllm

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