Skip to content

docs: clarify distributed DiskSaver construction - #3823

Open
nightcityblade wants to merge 1 commit into
pytorch:masterfrom
nightcityblade:docs/issue-2035
Open

docs: clarify distributed DiskSaver construction#3823
nightcityblade wants to merge 1 commit into
pytorch:masterfrom
nightcityblade:docs/issue-2035

Conversation

@nightcityblade

Copy link
Copy Markdown
Contributor

Fixes #2035

Description:

Clarify that Checkpoint and DiskSaver must be instantiated on every process in distributed configurations. Document that save_on_rank controls which process writes the checkpoint and that rank-guarded construction can block while Ignite discovers an externally initialized process group.

Tests:

  • pre-commit run ruff-check --files ignite/handlers/checkpoint.py
  • pre-commit run black --files ignite/handlers/checkpoint.py
  • python -m pytest tests/ignite/handlers/test_checkpoint.py -q (66 passed, 4 skipped)

Check list:

  • New tests are added (if a new feature is added) — N/A, documentation-only clarification
  • New doc strings: description and/or example code are in RST format
  • Documentation is updated (if required)

@github-actions github-actions Bot added the module: handlers Core Handlers module label Aug 12, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

module: handlers Core Handlers module

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Unable to create DiskSaver when program launched with torch.distributed.launcher

1 participant