verl-project/verl

[FR] Documentation about Resolving OOM

Open

#1,014 创建于 2025年4月10日

在 GitHub 查看
 (6 评论) (7 反应) (0 负责人)Python (21,533 star) (3,940 fork)auto 404
call for contributiondocumentationgood first issue

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

技术栈
pythonpytorch
领域
documentationperformance
议题类型
documentation
难度面向新贡献者的预计实现难度，1 表示很小改动，5 表示专家级工作。
3
预计时间有经验贡献者完成调查、实现、测试并准备 pull request 的粗略时间范围。
1-2 days
活动状态议题当前的可参与程度：新鲜、活跃、陈旧、阻塞或等待维护者输入。
active
清晰度议题是否清楚说明期望改动、验收标准和下一步。
clear
前置要求
basic Pythonfamiliarity with PyTorchunderstanding of GPU memory
新手友好度1-100 的估计分数，表示该议题对首次贡献者的友好程度。
70
研究方向
Investigate existing OOM related issues, e.g., #328, to identify common scenarios. For each configuration listed (rollout's gpu memory utilization, Liger Kernel, max len/gpu, gradient checkpointing, etc.), benchmark their effect on memory usage and training performance. Document findings in a structured guide, referencing relevant source files like configuration scripts and training loops.