Reproduction and critical analysis of “Generalization in Diffusion Models Arises from Geometry-Adaptive Harmonic Representations” (Kadkhodaie et al., 2024). Includes reproducible pipeline, datasets…

This is an ongoing project where I'm building a small LLaMA-like LLM from scratch.

Our implementation of SpodNet: the first neural network layer to guarantee jointly sparse and SPD output matrices.

Self-contained implementation of a Byte-Pair Encoding (BPE) tokenizer covering the full pipeline from training to encoding / decoding.

A summary of related works about flow matching, stochastic interpolants

Code for the paper https://arxiv.org/abs/2402.04997

Official pytorch repository for "Diffusion Posterior Sampling for General Noisy Inverse Problems"

Student version of Assignment 1 for Stanford CS336 - Language Modeling From Scratch

Diffusion distribution matching for inverse problems

Optimal Transport for Machine Learners

Interactive visualizations of the geometric intuition behind diffusion models.

PyTorch code for FLD (Feature Likelihood Divergence), FID, KID, Precision, Recall, etc. using DINOv2, InceptionV3, CLIP, etc.

Educational implementation of the Discrete Flow Matching paper

A PyTorch library for implementing flow matching algorithms, featuring continuous and discrete flow matching implementations. It includes practical examples for both text and image modalities.

Code for MGDM algorithm, ICML 2025, https://arxiv.org/abs/2502.03332

Repository for CARTE: Context-Aware Representation of Table Entries

Code for PnP-Flow: Plug-and-Play image restoration with Flow Matching (ICLR 2025)

Python implementation of soft-DTW.

Domain Adaptation benchmark

⚡ TabPFN: Foundation Model for Tabular Data ⚡

El0ps: An Exact L0-Problem Solver

Make your tweets more punchy with Mixtral and GPT