NVIDIA / NVIDIA/Megatron-LM

[QUESTION] why batch_p2p_sync is still forced?

Open
#1,844 0 comments 1 reaction 0 assignees View on GitHub
question
Dominant language
Python
Stars
17.9k
Forks
4.5k
Avg merge
4d 6h
Merged PRs (30d)
271

Description

Is there still a bug in pytorch related to batch_p2p_sync? (Comment says it is a bug in older version of pytorch.)
batch_p2p_sync=True is forced now as there is no argument to set batch_p2p_sync and the default of it is "True" in the following code (https://github.com/NVIDIA/Megatron-LM/blob/main/megatron/core/model_parallel_config.py#L270)
Thanks

Contributor guide

Open the contributing guide

Assessment

This issue has not been assessed yet.

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.