Modal (@modal)

Training large models with RL typically runs on a fixed homogeneous cluster of the latest accelerators for numerical stability and faster throughput. This throws away inference optimization work done for other vendors and generations, limits the capacity you can utilize, and is unnecessarily expensi...

<i>This post was written by Colin Weld and Connor Adams. Read it here or </i><a target="_blank" href="https://modal.com/blog/scaling-to-1-million-concurrent-sandboxes-in-seconds" color="blue">on our blog</a><i>.</i> At Modal, we build sandboxes, among other things. Agents run in sandboxes, and agen...