mlfoundations

An open source implementation of CLIP.

An open-source framework for training large multimodal models.

DataComp for Language Models

🍃 MINT-1T: A one trillion token multimodal interleaved dataset.

DataComp: In search of the next generation of multimodal datasets

Robust fine-tuning of zero-shot models

Automatic evals for LLMs

Editing Models with Task Arithmetic

A repository for research on medium sized language models.

Model soups: averaging weights of multiple fine-tuned models improves accuracy without increasing inference time

Simple large-scale training of stable diffusion with multi-node support.

Language models scale reliably with over-training and on downstream tasks

Patching open-vocabulary models by interpolating weights

A benchmark for distribution shift in tabular data

Research on Tabular Foundation Models

Release of ImageNet-Captions