Code Repository for the ICML 2024 paper: "Towards Scalable and Versatile Weight Space Learning".

Regime-adaptive speculative weight prediction for accelerating neural network training

A collection of weight space learning including papers, codes, and datasets.

dLLM: Simple Diffusion Language Modeling

A Large-Scale, High-Quality Math Dataset for Reinforcement Learning in Language Models

Visualize neural network with or without weights

incenτivised inτerneτ-wide τraining

Code for the paper Don't Pay Attention

[ICML 2025] Retraining-Free Merging of Sparse MoE via Hierarchical Clustering

Oscillatory State-Space Models

Code for Celo: Training Versatile Learned Optimizers on a Compute Diet

An efficient implementation of learned optimizers in PyTorch

[KDD'2024] "LLM4Graph: A Survey of Large Language Models for Graphs"

[ICML 2024] LLM and Simulation as Bilevel Optimizers: A New Paradigm to Advance Physical Scientific Discovery

GraphLLM: Boosting Graph Reasoning Ability of Large Language Model (IEEE Transactions on Big Data)

[ICLR 2023] "Learning to Grow Pretrained Models for Efficient Transformer Training" by Peihao Wang, Rameswar Panda, Lucas Torroba Hennigen, Philip Greengard, Leonid Karlinsky, Rogerio Feris, David …

Official source code for "Graph Neural Networks for Learning Equivariant Representations of Neural Networks". In ICLR 2024 (oral).

[TMLR 2025] Meta-learning Optimizers for Communication-Efficient Learning

Automatic gradient descent

Model Zoos published at the NeurIPS 2022 Dataset & Benchmark track: "Model Zoos: A Dataset of Diverse Populations of Neural Network Models"

Custom distributed implementation of our proposed DTP algorithm parallelizing feedback weight training across GPUs (ICML 2022)

Code Repository for the NeurIPS 2021 paper: "Self-Supervised Representation Learning on Neural Network Weights for Model Characteristic Prediction".

Framework for defining machine learning models, including feature generation and transformations, as directed acyclic graphs (DAGs).

Tutorial on amortized optimization for learning to optimize over continuous domains

Code for Parameter Prediction for Unseen Deep Architectures (NeurIPS 2021)

Official repository for the paper "On Evaluation Metrics for Graph Generative Models"