Skip to content

Conversation

@AndSonder
Copy link
Contributor

@AndSonder AndSonder commented Mar 8, 2025

PR Category

Auto Parallel

PR Types

Bug fixes

Description

  1. pp 场景下没有 init 的 tensor 在计算操作(比如 split,unsqueeze 等)后 device 会变为 cpu,进而导致程序 hang 在后续初始化通信 context 的地方(会错误地初始化 GlooCommContext)
  2. xpu send&recv 不支持 0-d tensor,修改 clip 的 value 的 shape 为 1

Pcard-76459

@paddle-bot
Copy link

paddle-bot bot commented Mar 8, 2025

你的PR提交成功,感谢你对开源项目的贡献!
请关注后续CI自动化测试结果,详情请参考Paddle-CI手册
Your PR has been submitted. Thanks for your contribution!
Please wait for the result of CI firstly. See Paddle CI Manual for details.

@From00 From00 merged commit 66349cc into PaddlePaddle:develop Mar 8, 2025
32 checks passed
YqGe585 pushed a commit to YqGe585/Paddle that referenced this pull request May 7, 2025
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants