RT @Luke_Metz: We have a new paper on learned optimizers! We used thousands of tasks (and a lot of compute 😬) to train general purpose lear…