PR #7908 Add Feature Universal Checkpoint for AutoTP

Initial plan

Copilot committed 47 days ago

Revert "fix: update 1 file reformatted."

Copilot committed 47 days ago

Merge pull request #5 from nathon-lee/copilot/git-revert-ff886701

nathon-lee committed 47 days ago

Merge branch 'deepspeedai:master' into master

nathon-lee committed 40 days ago

Initial plan

Copilot committed 40 days ago

Reapply "fix: update 1 file reformatted."

Copilot committed 40 days ago

Merge pull request #6 from nathon-lee/copilot/remove-commits-from-master

nathon-lee committed 40 days ago

feat: Refactor AutoTP universal checkpoint metadata schema handling

nathon-lee committed 35 days ago

fix: update unit test file test_autotp_universal_checkpoint.py

nathon-lee committed 30 days ago

fix: update unit test file test_autotp_universal_checkpoint.py

nathon-lee committed 30 days ago

Add constant for AutoTP universal-checkpoint metadata key

nathon-lee committed 29 days ago

Remove redundant callable() guard for _mark_uc_metadata hook

nathon-lee committed 29 days ago

tests: cover uneven sub_param_sizes in AutoTP UC restore

nathon-lee committed 28 days ago

fix: update some logic for _resolve_autotp_partition

nathon-lee committed 28 days ago

docs: update universal checkpointing and AutoTP checkpoint docs

nathon-lee committed 28 days ago

Merge branch 'master' into feat_uc_autotp

delock committed 27 days ago

tests: avoid pytest import file mismatch by renaming AutoTP UC test

nathon-lee committed 23 days ago

fix: Automatically fix file end line breaks and formatting issues

nathon-lee committed 23 days ago

Merge branch 'master' into feat_uc_autotp

delock committed 23 days ago

DeepSpeed Add Feature Universal Checkpoint for AutoTP #7908 Merged