Repository navigation

feat(pt_expt): resume checkpointed training data progress - #6093

Merged

OutisLi merged 11 commits into

deepmodeling:masterfrom

OutisLi:fix/resume-training-data-progress

Oct 26, 2026

Merged

feat(pt_expt): resume checkpointed training data progress#6093
OutisLi merged 11 commits into
deepmodeling:masterfrom
OutisLi:fix/resume-training-data-progress

Commits on Oct 21, 2026
- feat(pt_expt): resume checkpointed training data progress
  
  OutisLi
  and
  OutisLi-Bot
  committed
  16654d6
- fix(data): preserve saved probabilities across restart roundoff
  OutisLi
  committed
  ea8f990
- fix(pt_expt): preserve restart across changed world sizes
  OutisLi
  committed
  3d7e08d
Commits on Oct 22, 2026
- Merge branch 'master' into fix/resume-training-data-progress
  OutisLi
  authored
  e793038
- fix(pt_expt): allow explicit restart data progress reset
  OutisLi
  committed
  20a9107
Commits on Oct 23, 2026
- test(pt_expt): close LMDB pipelines before fixture cleanup
  OutisLi
  committed
  f5758d6
- test(pt_expt): match checkpoint optimizer restore devices
  OutisLi
  committed
  5237ad0
Commits on Oct 24, 2026
- chore: merge master into training data restart
  OutisLi
  committed
  4277388
- fix(pt-expt): preserve independent restart sampling streams
  OutisLi
  committed
  6d1ee3f
Commits on Oct 25, 2026
- test(train): select multitask gradient cases explicitly
  OutisLi
  committed
  7c745dd
Commits on Oct 26, 2026
- fix(data): integrate grouped sampling with restart progress
  OutisLi
  committed
  ef510a0