Documentation Index
Fetch the complete documentation index at: /llms.txt
Use this file to discover all available pages before exploring further.
Concepts, launch scripts, customization hooks, and a complete CLI reference.
| Page | What it covers |
|---|---|
| Core Concepts | The four objects in the training loop and the four-knob invariant. |
| Launch Script | What python scripts/run_*.py does, how a launch script is structured, and how to override a recipe. |
| Argument Groups | Where model args, perf_args, grpo_args, and the other launch-script flag groups belong. |
| Fully Async RL | Continuous generation decoupled from training: the schedule, the data buffer, async eval, and the metrics to watch. |
| Training Backends | Megatron-LM and FSDP: what each one owns, how to choose, parallelism, checkpoints, and hooks. |
| Monitoring & Logging | wandb, structured logs, per-source breakdowns, profiling, router metrics. |
| Customization | The --*-path plug-points for custom Python — rollout, reward, filters, loss, hooks. |
| Generate Endpoint | Custom generate functions that own tokens and loss masks via the raw /generate endpoint. |
| Agentic Rollout (TITO) | Configure an OpenAI-compatible agent loop with TITO trajectory assembly. |
| Agentic Environments | Supplying an environment: dataset + reward, your own env via the plug points, or an external ecosystem. |
| CLI Reference | Every flag Miles accepts, grouped by subsystem. |
