Replace vision memorization set with no-text-id geometric image cards

Remove probe ids from vision memorization prompts

Upload dataset