vinyesm

arckit

Tools for working with the Abstraction & Reasoning Corpus

ARGA-AAAI23

Abstract Reasoning with Graph Abstractions (ARGA) implementation

cc2dataset

Easily convert common crawl to a dataset of caption and document. Image/text Audio/text Video/text, ...

DiT

Official PyTorch Implementation of "Scalable Diffusion Models with Transformers"

easy

Google Hash Code Paris 2014

fcgan

Fast column generation for atomic norm regularization

learn-lvggm

Gaussian graphical models (GGM) have been widely used in many high-dimensional applications. Problems in which all relevant variables have been identified is extremely rare. The unobserved variables (or latent variables) can induce misleading dependence structure between observed variables. In this paper, we focus on a family of Latent Variable Gaussian graphical models (LVGGM), where the model is conditionally sparse given latent variables, but marginally non-sparse. In order to reveal latent variables and recover the structure of the complete model, we introduce a convex formulation with a new regularization imposing sparsity on latent factors. We propose a tractable convex algorithm and study identifiability conditions. We show promising results on synthetic datasets.

learning-gm

Estimation of Gaussian graphical models

magvit2-pytorch

Implementation of MagViT2 Tokenizer in Pytorch

nanoGPT-for-image

Adaptation of nanoGPT for images

vinyesm/fcgan

Fast column generation for atomic norm regularization

vinyesm/arckit

Tools for working with the Abstraction & Reasoning Corpus

vinyesm/ARGA-AAAI23

Abstract Reasoning with Graph Abstractions (ARGA) implementation

vinyesm/cc2dataset

Easily convert common crawl to a dataset of caption and document. Image/text Audio/text Video/text, ...

vinyesm/DiT

Official PyTorch Implementation of "Scalable Diffusion Models with Transformers"

vinyesm/easy

Google Hash Code Paris 2014

vinyesm/learn-lvggm

Gaussian graphical models (GGM) have been widely used in many high-dimensional applications. Problems in which all relevant variables have been identified is extremely rare. The unobserved variables (or latent variables) can induce misleading dependence structure between observed variables. In this paper, we focus on a family of Latent Variable Gaussian graphical models (LVGGM), where the model is conditionally sparse given latent variables, but marginally non-sparse. In order to reveal latent variables and recover the structure of the complete model, we introduce a convex formulation with a new regularization imposing sparsity on latent factors. We propose a tractable convex algorithm and study identifiability conditions. We show promising results on synthetic datasets.

vinyesm/learning-gm

Estimation of Gaussian graphical models

vinyesm/magvit2-pytorch

Implementation of MagViT2 Tokenizer in Pytorch

vinyesm/nanoGPT-for-image

Adaptation of nanoGPT for images

vinyesm/Open-MAGVIT2

A packaging of Open-MAGVIT2: Democratizing Autoregressive Visual Generation

vinyesm/video2dataset

Easily create large video dataset from video urls

vinyesm/vinyesm.github.io

vinyesm/re-arc

Reverse Engineering the Abstraction and Reasoning Corpus