maxtext icon indicating copy to clipboard operation
maxtext copied to clipboard

Optimal Device Assignment for 2-D mesh

Open Lumosis opened this issue 9 months ago • 1 comments

Description

Introduces an algorithm to derive the optimal device assignment that maps a 1D ring of logical devices to a 2D mesh of physical devices.

Tests

* Baseline with optimize_mesh_for_tpu_v6e=false

================================================
MLPerf Results Summary
================================================
SUT name : PySUT
Scenario : Offline
Mode     : PerformanceOnly
Samples per second: 17.2537
Tokens per second: 4869.9130
Result is : VALID
  Min duration satisfied : Yes
  Min queries satisfied : Yes
  Early stopping satisfied: Yes



* Baseline with optimize_mesh_for_tpu_v6e=true

================================================
MLPerf Results Summary
================================================
SUT name : PySUT
Scenario : Offline
Mode     : PerformanceOnly
Samples per second: 19.0753
Tokens per second: 5414.4059
Result is : VALID
  Min duration satisfied : Yes
  Min queries satisfied : Yes
  Early stopping satisfied: Yes


* New code

================================================
MLPerf Results Summary
================================================
SUT name : PySUT
Scenario : Offline
Mode     : PerformanceOnly
Samples per second: 19.5143
Tokens per second: 5528.4437
Result is : VALID
  Min duration satisfied : Yes
  Min queries satisfied : Yes
  Early stopping satisfied: Yes

Checklist

Before submitting this PR, please make sure (put X in square brackets):

  • [x] I have performed a self-review of my code.
  • [x] I have necessary comments in my code, particularly in hard-to-understand areas.
  • [x] I have run end-to-end tests tests and provided workload links above if applicable.
  • [x] I have made or will make corresponding changes to the doc if needed.

Lumosis avatar Mar 20 '25 04:03 Lumosis

This PR has been automatically marked as stale because it has not had recent activity. It will be closed soon if no further activity occurs. Thank you for your contributions.

github-actions[bot] avatar Nov 21 '25 16:11 github-actions[bot]

This PR was closed because it has been inactive for a while. Please reopen it if you are still working on it.

github-actions[bot] avatar Nov 29 '25 16:11 github-actions[bot]