Skip to content

v0.7.0

Choose a tag to compare

View all tags
@martinez-hub martinez-hub released this 19 Jul 03:30
v0.7.0
93eff47
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

Ships the hardware-validated HPC features, plus docs/CI hardening.

Added

  • Multi-node DDP — submitit_slurm_config + seed_everything_per_rank, a fail-fast world-size guard, rank-0-only metrics (#50)
  • HydraFSDP — Fully-Sharded Data Parallel training under Hydra --multirun (#58)
  • pin_gpu_round_robin — pack several small sweep jobs onto each GPU (#59)

All three were validated on real SLURM/GPU hardware (2× L40S single-node, 2× A100 across 2 nodes) before this release.

Fixed

  • HydraDDP/HydraFSDP docs: the working launcher-provides-ranks pattern (#95)
  • Actionable error when a bare list/tuple is passed as a sweep arg (#93)

Misc

Docs/examples wiring, cluster-validation runbook, least-privilege CI workflow permissions, and CITATION.cff.

Full changelog: https://github.com/martinez-hub/mushin/blob/main/CHANGELOG.md

Assets 2
Loading