title: where the flag value came from, and the one pair that both sides accept
source: manuals/sglang-101/sections/7-2-sglang-as-an-rl-rollout-engine.md
lines: 211-213
lang: text
note: The first paragraph explains why the name reached the trainer unchanged.
---
The flag value came from the wrong example. In the FSDP actor, the SGLang Triton attention bridge turns on only when `torch.version.hip` is set, which means AMD GPUs {{code}}. On an NVIDIA GPU the value goes straight to transformers, which has no attention named `triton`.

The only true on-policy contract in Miles pairs the SGLang `fa3` backend with `flash_attention_3` in FSDP {{code}}. We ran the job again with the contract flags to see where it stops.
