nn.Graph reuse eager lbn without create duplicate variable op #6981
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
yuanms2
reviewed
Dec 9, 2021
yuanms2
reviewed
Dec 9, 2021
yuanms2
reviewed
Dec 9, 2021
|
这个pr是不是也能同时解决这个问题 https://github.com/Oneflow-Inc/OneTeam/issues/827 |
是的。一并解决 |
strint
reviewed
Dec 20, 2021
| @@ -31,6 +31,9 @@ class TensorNameScope { | |||
|
|
|||
| void Record(const std::shared_ptr<Tensor>& tensor, const std::string& name); | |||
|
|
|||
| // NOTE(chengcheng): TensorNameScope need to be cleared after current graph build. | |||
| if (!opt_lbn.empty()) { | ||
| // NOTE(chengcheng): This eager tensor has been fed as variable op before, so we just use the | ||
| // lbn, and will NOT create duplicate variable op again. | ||
| (*outputs)[0] = input_tensor; |
|
本 PR 关闭。后续的工作由：
分别支持。 |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
nn.Graph 捕获 nn.Module 中的 Eager Tensor（parameters） 时，没有考虑 Tensor 传入会重复的问题，导致同一个 tensor 指针（但是在 module 中有多个不同的 name ）会被重复创建 Variable Op，虽然不会影响正确性（多个 Variable Op 绑定相同的 Tensor 内存），但是会在多卡情形下造成额外的梯度同步操作，影响性能。
此问题由 issue：
反馈。
该 PR 还支持了 每次 Graph build 完，清空 TensorNameScope 的功能，解决：
中反应的问题