The code used to train and run inference with the ColVision models, e.g. ColPali, ColQwen2, and ColSmol.
Terminal score: 0–1000 raw, weighted across 4 dimensions. Public score: 0–10 normalized (shown in the 30-day stars chart above).
