mindspore-ai / mindspore-ai/hyper-parallel
[fully_shard]注册PostBackwardHook时，输入requires_grad属性跳变
Nobody has claimed this yet.
- Dominant language
- Python
- Stars
- 53
- Forks
- 63
- Avg merge
- 23h 45m
- Merged PRs (30d)
- 63
Description
该问题是怎么引起的？
在注册PostBackwardHook时，调用Autograd.Function输入中，既有需要求导的非叶子节点，也有不需要求导的非叶子节点，生成的输出中所有非叶子节点的requires_grad全变成True，不符合预期
重现步骤
报错信息
schema_version: 1
source: gitcode
gitcode_repo: mindspore/hyper-parallel
gitcode_issue: 66
source_url: https://gitcode.com/mindspore/hyper-parallel/issues/66
Contributor guide
No contributing guide indexed for this repository
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Start by locating the fully_shard PostBackwardHook and Autograd.Function handling, then reproduce the described case with both gradient-requiring and non-gradient-requiring non-leaf inputs. Verify how requires_grad is propagated to outputs; done means the two input categories retain the expected behavior. The issue provides no file or test path, so project navigation and a focused reproduction are needed first.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python
- Domain
- distributed-systems
- Issue type
- Bug
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Active
- Clarity
- Needs clarification
- Newbie friendliness
- 35/100