Repository navigation
Conversation
jlnav
marked this pull request as ready for review
October 27, 2026 19:28
Codecov Report❌ Patch coverage is Additional details and impacted files@@ Coverage Diff @@
## develop #1798 +/- ##
===========================================
- Coverage 88.95% 88.70% -0.26%
===========================================
Files 74 74
Lines 8795 9015 +220
Branches 1421 1470 +49
===========================================
+ Hits 7824 7997 +173
- Misses 723 754 +31
- Partials 248 264 +16 ☔ View full report in Codecov by Harness. 🚀 New features to boost your workflow:
|
…gned logical "shares" of physical gpus). libE_specs["gpu_sharing_factor"] requests number of logical shares to make out of each physical gpu. gpu_sharing_factor of 1 (default) produces traditional behavior. shares assigned round-robin. workers convert assigned shares into physical GPU ids. associated tests.
…PU was assigned. new "pack" policy for scheduling fractional gpu shares (multiple jobs get packed onto a GPU by default). "pack" or "spread" options for gpu_packing. scheduler builds "gpu_bins" for allocating into. jobs can still request a whole GPU. tests.
…lculating available GPUs
…two GPUs. unit tests for gpu_sharing_factor, gpus_per_group, building gpu bins, capacity handling for gpus, "pack" and "spread" policies, and fractional-gpu support in alloc_support
Use local Resources instance instead of Resources.resources (None when run via run_tests.py -u isolation). Return Resources not just manager, pass directly to AllocSupport, and override scheduler for pack policy.
Add test_select_rsets_fragmented_remainder for sf=4 where only partial GPUs are free. No single GPU satisfies rem, so _select_rsets_for_gpu_group must fragment across multiple GPUs (pack vs spread ordering) and raise InsufficientFreeResources when fragments insufficient. Covers previously uncovered else block scheduler.py:256-269.
jlnav
force-pushed
the
feature/gpu-sharing-factor
branch
from
October 29, 2026 20:59
0f8b4f5 to
f5d73e0
Compare
This file contains hidden or 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
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.
Addresses #1394
libE_specs["gpu_sharing_factor"](shares per physical GPU) with physical-exclusive, packing-aware scheduling.num_gpusnow supports fractional values (e.g. 0.5 → 1 share when sharing_factor is 2, 2 → 2*sharing_factor shares)scheduler_opts["gpu_packing"]="pack"|"spread"controls co-location.