HF RL Explorer

Feature Request: Support mp Distributed Executor Backend for Single-Node vLLM Inference

Feature Request: Support mp Distributed Executor Backend for Single-Node vLLM Inference: a task in LegoFlow-SWE (Harbor dataset). Users of SkyRL’s vLLM inference engine cannot select the mp (multiprocessing) distributed executor backend for single-node, multi‑GPU serving, despite Ray being…

The task

Users of SkyRL’s vLLM inference engine cannot select the `mp` (multiprocessing) distributed executor backend for single-node, multi‑GPU serving, despite Ray being unnecessary for intra‑engine worker management in that scenario.

Part of Lego-X/LegoFlow-SWE.