Skip to content

[fleet exe] Update comm init #39603

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Merged
merged 2 commits into from
Feb 16, 2022
Merged

[fleet exe] Update comm init #39603

merged 2 commits into from
Feb 16, 2022

Conversation

@FeixLiu
Copy link
Contributor

@FeixLiu FeixLiu commented Feb 16, 2022 •
edited
Loading

PR types

Others

PR changes

Others

Describe

  • Update the way to init the nccl comm.
  • Delete some useless check since auto cut model don't have concept of mp and pp.
@FeixLiu FeixLiu requested a review from wangxicoding February 16, 2022 03:30
wangxicoding
wangxicoding approved these changes Feb 16, 2022
View reviewed changes
Copy link
Contributor

@wangxicoding wangxicoding left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM
paddle/fluid/distributed/fleet_executor/dist_model.cc
} else if (input_data.dtype == DistModelDataType::INT32) {
input_tensor_ptr = input_tensor->mutable_data<int32_t>(dims, place);
} else {
// Q(fleet exe dev): for input/output, should we support fp16
Copy link
Contributor

@wangxicoding wangxicoding Feb 16, 2022

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

后续支持一下吧
@FeixLiu FeixLiu merged commit 7d53a28 into PaddlePaddle:develop Feb 16, 2022
@FeixLiu FeixLiu deleted the update_pp_comm_init branch February 16, 2022 09:30
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

2 participants

@FeixLiu @wangxicoding