verl-project/verl

[FR] Documentation about Resolving OOM

已关闭

#1,014 创建于 2025年4月10日

 (6 条评论) (7 个反应) (0 位负责人)Python (4,434 个派生)auto 404
call for contributiondocumentationgood first issue

仓库指标

星标
 (23,069 个星标)
PR 合并指标
 (平均合并 3天 6小时) (30 天内合并 125 个 PR)

描述

Motivation

There are many issues related to OOM, e.g. #328 . We might need a clear guide about how to resolve OOM.

Plan

A non-exclusive enumeration about related configurations:

  1. Rollout:gpu_memory_utilization
  2. Other Inference:
    1. Liger Kernel
    2. *_max_len_per_gpu / micro_batch_size_per_gpu
  3. Training:
    1. Liger Kernel
    2. Ulysses Sequence Parallelism
    3. gradient checkpointing
    4. offload

TODO

  • Complete the list of related configurations
  • Benchmark the effect & overhead of each configuration

贡献者指南