Skip to main content
A few are purely demonstrative, but most are verifiable against a concrete performance score.

​
Recipes

End-to-end training workflows — the place to start.
  • geo3k_vlm: Training VLMs with FSDP using GRPO on the GEO3K dataset.
    • multi_turn: The same dataset over multiple turns, with the model cropping images through an interactive environment.
  • lora: LoRA fine-tuning with the Megatron backend.
  • multi_lora: Fully-async multi-adapter LoRA training with a slot-keyed adapter page table.
  • on_policy_distillation: Teacher–student distillation on the student’s own rollouts, run inside the on-policy training loop.
    • qwen3_5_35b_selfdistill: Two-phase self-distillation of Qwen3.5-35B-A3B on one 8xH200 node, with an in-process Megatron teacher.
  • ppo: Actor-critic PPO with GAE advantages, where the critic shares the actor’s train GPUs.
  • retool_v2: Tool-enabled language model generation with sandboxed Python code execution interleaved with thinking.
  • swe-agent-harbor-docker: Trains coding and terminal agents with Harbor-managed local Docker sandboxes and verifier rewards.

​
Infra Features

Runtime and infrastructure plumbing rather than training recipes — how miles moves data and weights around.
  • fully_async: Demonstrates fully asynchronous rollout generation for higher efficiency.
  • low_precision: Examples of FP8 training and inference, plus INT4 QAT, for improved throughput and stability.
  • p2p_weight_transfer: Point-to-point weight transfer between training and rollout engines.
  • random_async: Dataset-free stress test of the async rollout ↔ trainer loop.
  • train_infer_mismatch_helper: Algorithmic methods for rollout correction (e.g., TIS, MIS).
  • true_on_policy: Ensures strictly equal log probabilities between inference (SGLang) and training engines.

​
Experimental

Not fully verified — for experimental and development use.
  • agentenv: Rollouts against AgentENV, a self-hosted platform running agent sandboxes on Firecracker microVMs.
  • DrGRPO: Custom reducer for Dr.GRPO algorithm.
  • eval: Documentation and setup for evaluation environments using NeMo-Skills.
  • eval_multi_task: Example for supporting OOD evaluation tasks, e.g., GPQA, IFBench.
  • formal_math: Examples related to formal math reasoning tasks, including a single round demo.
  • multi_agent: Example of running multi-agent RL with miles.
  • nemo-gym: SWE-agent training with NVIDIA NeMo Gym as the environment ecosystem.
  • openenv: Rollouts against OpenEnv-hosted environments.
  • reproducibility: Guides on achieving bitwise experiment reproduction using deterministic modes.
  • search-r1: A minimal reproduction of Search-R1, featuring multi-turn conversation and tool-calling.
  • strands_sglang: Integration example with the Strands-Agents scaffolding framework.
  • swe-agent-harbor-daytona: The swe-agent-harbor-docker pipeline with task sandboxes hosted on Daytona instead of local Docker.
  • tau-bench: Training in an agentic multi-turn tool use environment (Tau-bench).
  • verifiers: Training on a Prime Intellect Verifiers environment instead of a Miles prompt dataset.