Deterministic Inference across Tensor Parallel Sizes That Eliminates Training-Inference Mismatch 文章

ArXiv CS.CL2026-06-01NEWSen作者: Ziyang Zhang, Xinheng Ding, Jiayi Yuan, Rixin Liu, Huizi Mao, Jiarong Xing, Zirui Liu

查看原文 →

Deterministic Inference across Tensor Parallel Sizes That Eliminates Training-Inference Mismatch · 相关技术

相关技术

batch-invariant kernels multi-agent systems LLM-as-a-judge evaluation Greedy Decoding reinforcement learning large language models Fully Sharded Data Parallel tensor parallelism