๊ฐ€์ž… ํ›„ ์ดˆ๋Œ€ ๋งํฌ๋ฅผ ๊ณต์œ ํ•˜๋ฉด ๋™์˜์ƒ ์žฌ์ƒ ๋ฐ ์ดˆ๋Œ€ ๋ณด์ƒ์„ ๋ฐ›์„ ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.

SGLang
@sgl_project
Run LLMs fast at any scale ๐Ÿ”— Join our community For AI tech blogs & deep-dives ๐Ÿ‘‰ @lmsysorg
๊ฐ€์ž… May 2025
53 ํŒ”๋กœ์ž‰ ์ค‘    9.5K ํŒฌ
Congrats to the Halo team (@whitecircle) on the launch! ๐ŸŽ‰ SGLang powers Halo rollouts as the primary engine. It runs in an isolated serving environment, returning token IDs, logprobs, and MoE routing to the trainer, with weights synced over NCCL and generation overlapped across servers. Excited to partner with the Halo team!
๋” ๋ณด๊ธฐ
Halo is not limited to supervised fine-tuning. It supports async RL and training with external environments. We are excited to partner with the @sgl_project team to make it the primary engine for rollouts.
๋” ๋ณด๊ธฐ