X · @huggingface
· X / Twitter
RT Sergio Paniego: TRL v1.7.0 is out‼️ + continuous batching makes GRPO and RLOO 1.25x faster at -16 GB + proper MoE post-training across GRPO/RLOO/…
RT Sergio PaniegoTRL v1.7.0 is out‼️+ continuous batching makes GRPO and RLOO 1.25x faster at -16 GB+ proper MoE post-training across GRPO/RLOO/AsyncGRPO+ new GMPO trainer+ AsyncGRPO weight sync + padding-free+ morehttps://github.com/huggingface/trl/releases/tag/v1.7.0