viable/strict/1791021071: [fsdp2][mp] per-param mixed_precision policy (#196626)
- PyTorch 近 90 天出现 1850 次
- PyTorch 近 90 天第 1826 次发版
- 上一次:同一天稍早 · trunk/47df693ee1bd31c15dc89fbc4884c99ddd82a0a4
发生了什么
Summary: We want to enable per-param mixed precision policy in FSDP2 for several reasons including KDA, etc. Key Points All-gather: During all-gather, existing homogenous path remains unchanged. When parameters require different compute dtypes, FSDP groups staging copies by source/target dtype and invokes the existing uint8 communication buffer, keeping only one AG per fully_shard. As seen in benchmarking results be…
摘要按规则整理自下方来源原文