🚀 We are thrilled to introduce TextImage Data Augmentation, developed in collaboration with Albumentations AI! ✨ This multimodal technique modifies document images and text simultaneously, enhancing Vision Language Models (VLMs) for high-text datasets.
The Document AI team (@Molbap, @rwightman, @danaaubakirova) at Hugging Face is developing a new multimodal data augmentation pipeline utilising both visual and textual aspects of document images.
We release Idefics2-8B, a foundation vision language model with SOTA results for its size on many benchmarks.
For Idefics2, we adopted a simple architecture: -Images are fed to a vision encoder, then to a modality projection to match the input dimension of the LLM, and finally to a perceiver resampler for efficient pooling. -Interleaved image-text data are then passed to the LLM.
During the pre-training: -The modality projection and perceiver resampler weights are newly initialized. -We start with pre-trained models for the vision encoder and the LLM, and continue the training with LoRA. -In total, we see 1.5T images!
We pre-train on 3 types of data, all publicly available: -Interleaved image-text documents: our dataset OBELICS HuggingFaceM4/OBELICS -Image caption pairs: only synthetic captions! -PDF documents: IDL and PDFA
We kept the aspect ratio of the images with the Patch n' Pack strategy, with a resolution of up to 980x980. At inference, it's also more efficient for lower-resolution images.
For the SFT, we build The Cauldron, a collection of 50 high-quality datasets in the user/assistant format. It is a ready-to-use dataset for the fine-tuning of any VLM. HuggingFaceM4/the_cauldron
Most current models, like LLaVA-NeXT, encode images with an excessive number of tokens, like 2880. Instead, we put a focus on being efficient at inference by training on a mix of images encoded with 64 tokens, and 320 tokens. The result is that we perform favorably compared to the best models in our size class, while being efficient at inference.
