File size: 1,821 Bytes
ab963d5 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 |
---
license: mit
---
# MAGMA -- Multimodal Augmentation of Generative Models through Adapter-based Finetuning
## Paper Authors
Constantin Eichenberg, Sidney Black, Samuel Weinbach, [Aleph Alpha](https://aleph-alpha.com "Independent AI R&D")
Letitia Parcalabescu, Anette Frank, [Heidelberg University](https://www.cl.uni-heidelberg.de "Computational Linguistics at Heidelberg University")
## Abstract
Large-scale pretraining is fast becoming the norm in Vision-Language (VL) modeling. However, prevailing VL approaches are limited by the requirement for labeled data and the use of complex multi-step pretraining objectives. We present MAGMA - a simple method for augmenting generative language models with additional modalities using adapter-based finetuning. Building on Frozen, we train a series of VL models that autoregressively generate text from arbitrary combinations of visual and textual input. The pretraining is entirely end-to-end using a single language modeling objective, simplifying optimization compared to previous approaches. Importantly, the language model weights remain unchanged during training, allowing for transfer of encyclopedic knowledge and in-context learning abilities from language pretraining. MAGMA outperforms Frozen on open-ended generative tasks, achieving state of the art results on the OKVQA benchmark and competitive results on a range of other popular VL benchmarks, while pretraining on 0.2% of the number of samples used to train SimVLM.
Paper on arXiv: https://arxiv.org/abs/2112.05253
## Repository
For the training and inference code, please refer to the [magma repository](https://github.com/Aleph-Alpha/magma) on GitHub.
## Model design
|