DistFlow decouples control and data flows in RL post-training, achieving near-linear scaling to 512 GPUs and up to 2.63x throughput vs. verl in tested workloads.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.DC 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
DistFlow: A Fully Distributed RL Framework for Scalable and Efficient LLM Post-Training
DistFlow decouples control and data flows in RL post-training, achieving near-linear scaling to 512 GPUs and up to 2.63x throughput vs. verl in tested workloads.