Rate this Page
★ ★ ★ ★ ★

Examples#

  • ping_pong.py: Demonstrates the basics of Monarch’s Actor/endpoint API with a ping-pong communication example

  • crawler.py: Demonstrates Monarch’s actor API and many-to-one communications with a web crawler example

  • spmd_ddp.py: Shows how to run PyTorch’s Distributed Data Parallel (DDP) using SPMDActor

  • Interactive SPMD Job: Shows how to use serve() and run_spmd() for interactive SPMD training with job caching and debugging

  • kubernetes_ddp.py: Extends the DDP example to run on Kubernetes using MonarchMesh CRD and operator

  • grpo_actor.py: Implements a distributed PPO-like reinforcement learning algorithm using the Monarch actor framework

  • distributed_tensors.py: Shows how to dispatch tensors and tensor level operations to a distributed mesh of workers and GPUs

  • debugging.py: Shows how to use the Monarch debugger to debug a distributed program

  • Multinode Slurm Tutorial: Multinode distributed training tutorial using Monarch and Slurm to run an SPMD training job.

  • Running on Kubernetes using Skypilot: Run Monarch on Kubernetes and cloud VMs via SkyPilot.