Reference
crab export
Export one Crab snapshot as ordinary files into raw object storage. This is the
inverse of crab import and does not create
Crab metadata in the target.
Synopsis
crab export [OPTIONS] --to <URL> [REPO_OR_PATH]...Use a repository followed by optional exact paths or trailing-slash subtrees,
or provide the repository with --from and use positional values only as path
selectors.
Options
| Option | Description |
|---|---|
--from <REPO> | Source repository when positional values are path selectors |
--to <URL> | Required raw target: s3://, gs://, az://, azure://, or file:// |
--revision <REVISION> | Branch, tag, ref, or full commit SHA |
--include <GLOB> | Include a glob; repeat as needed |
--exclude <GLOB> | Exclude a glob; repeat as needed |
--cache-dir <DIR> | Override the Crab cache root |
-j, --jobs <N> | File-level export concurrency |
--dry-run | Plan without writing target objects |
--force | Overwrite existing target objects |
--quiet | Suppress progress and summary while printing paths |
--json / --jsonl | Emit structured output |
Examples
crab export crab://bucket/models --to s3://raw-bucket/models --revision main
crab export --from crab://bucket/models --to file:///tmp/models models/
crab export crab://bucket/models --to gs://archive/models --include '*.json' --dry-runReview the target and use --force only when overwriting existing objects is
intentional.