eliahuhorwitz/ProbeX

This project helps machine learning researchers and MLOps engineers understand what an existing, undocumented model does simply by looking at its internal structure. You input the weights of various models (like Vision Transformers or Stable Diffusion LoRAs), and it outputs predictions about the categories of data the models were trained on or enables searching models by text descriptions. It's designed for those who need to quickly categorize or find suitable pre-trained models without extensive testing or documentation.