rlx-distributed v0.2.19 — RLX multi-node pipeline-parallel execution: graph partition into weight-sharded stages + tensor transport, to pool RAM across machines for models larger than any single node.

Sign in to join the discussion — your thebay.events account works here.