The model card indicates that this model was trained on a single task - svla_so101_pickplace, but the corresponding paper mentions training on a much larger number of datasets. Has the full model in the paper not been released or is the model card incorrect?
