Skip to content
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.

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

Fix various issues in comm replay including hashing, srcOrDst assignment, and trace file naming #148

Closed
wants to merge 15 commits into from

Conversation

TaekyungHeo
Copy link
Contributor

@TaekyungHeo TaekyungHeo commented Jul 31, 2024

Summary

  • Fixed hash function to correctly handle inSplit and outSplit by converting them to tuples.
  • Corrected srcOrDst assignment for "reduce", "broadcast", "gather", and "scatter" operations in runComms.
  • Updated trace file naming convention to rank{rank}.json when in directory mode.
  • Disabled warm-up by default and updated the help text for do_warm_up toggle.
  • Corrected a typo in the regenerateTensors description.

This PR relies on #147.

Test Plan

$ comm_replay --trace-type et --trace-path /home/sanshang/021_debug/000_code/param/trace/traces_megatronlm_gpt_43B_32ranks_pytnightly0703/execution_trace

image

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants