Fine-tune any Hugging Face LLM or VLM on day-0 using PyTorch-native features for GPU-accelerated distributed training with superior performance and memory efficiency.

VILA is a family of state-of-the-art vision language models (VLMs) for diverse multimodal AI tasks across the edge, data center, and cloud.

Easily turn large sets of image urls to an image dataset. Can download, resize and package 100M urls in 20h on one machine.

An open source implementation of CLIP.

Call all LLM APIs using the OpenAI format. Use Bedrock, Azure, OpenAI, Cohere, Anthropic, Ollama, Sagemaker, HuggingFace, Replicate (100+ LLMs)

Easily compute clip embeddings and build a clip retrieval system with them

Machine Learning Engineering Guides and Tools

PyTorch code and models for the DINOv2 self-supervised learning method.

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

Tackling the Generative Learning Trilemma with Denoising Diffusion GANs https://arxiv.org/abs/2112.07804

Turn any collection of files into a dataset

PyTorch implementation of MAE https//arxiv.org/abs/2111.06377

Reproducible scaling laws for contrastive language-image learning (https://arxiv.org/abs/2212.07143)

Official DeiT repository