MAGIC
CapDec
Stars - the number of stars that a project has on GitHub. Growth - month over month growth in stars.
Activity is a relative number indicating how actively a project is being developed. Recent commits have higher weight than older ones.
For example, an activity of 9.0 indicates that a project is amongst the top 10% of the most actively developed projects that we are tracking.
MAGIC
-
What if TTI was ran backwards? Like showing an image and asking what prompt it would need and conditions (temperature, top_k, etc) to generate that image. This might give us a better glimpse at how it wants to receive prompts.
yeah you could force a model to try to fill in a provided prompt template like that. Check this out: https://github.com/yxuansu/MAGIC
-
Cambridge AI Researchers Propose ‘MAGIC’: A Training-Free Framework That Plugs Visual Controls Into The Generation Of A Language Model
Github: https://github.com/yxuansu/magic
CapDec
- Open source – Unsupervised captioning getting closer to supervised captioning
-
Reverse engineer Stable Diffusion images
Cool! I also how a project that does image captioning: https://github.com/DavidHuji/CapDec
- CapDec: SOTA Zero Shot Image Captioning Using Clip and GPT2
What are some alternatives?
Auto-GPT - Auto-GPT + CLIP vision for stable v0.3.1
mmf - A modular framework for vision & language multimodal research from Facebook AI Research (FAIR)
OFA - Official repository of OFA (ICML 2022). Paper: OFA: Unifying Architectures, Tasks, and Modalities Through a Simple Sequence-to-Sequence Learning Framework
pytorch-widedeep - A flexible package for multimodal-deep-learning to combine tabular data with text and images using Wide and Deep models in Pytorch
CLIP-Caption-Reward - PyTorch code for "Fine-grained Image Captioning with CLIP Reward" (Findings of NAACL 2022)
3DCoMPaT-v2 - 3DCoMPaT++: An improved large-scale 3D vision dataset for compositional recognition
GPT2-Chinese - Chinese version of GPT2 training code, using BERT tokenizer.
DeepViewAgg - [CVPR'22 Best Paper Finalist] Official PyTorch implementation of the method presented in "Learning Multi-View Aggregation In the Wild for Large-Scale 3D Semantic Segmentation"
cappr - Completion After Prompt Probability. Make your LLM make a choice
x-clip - A concise but complete implementation of CLIP with various experimental improvements from recent papers