akka optimize training runs snapshots

List a run’s committed snapshots

Synopsis

List the snapshots a training run left in the catalog.

The list contains saves the catalog verified and committed. The configured interval does not show whether a save actually happened. Bytes left by an interrupted publication are not a snapshot.

ARTIFACTS shows how you can use each snapshot. Use weights to register a candidate model or start a branch. Use full training state to resume a run or start a branch where the backend supports it.

SIZE is what the snapshot holds in the bucket: the artifacts ARTIFACTS lists, summed. It falls when the catalog begins deleting one of them.

EPOCH shows which epoch a save ended, for a run that saves at the end of each epoch. Each of those saves is registered as a version of the run’s model, except the save at the run’s last update. That save is marked "run’s model". Its adapter is the model the run itself registers.

HELD BY identifies what retains a snapshot and prevents it from being deleted. The holders are a running branch, a candidate registration copying its weights, an operator pin, and the run itself for the weights it saved at the end of an epoch. An artifact kind with no holder is subject to the catalog’s retention limits and may be pruned.

Use "training snapshots get" to inspect one snapshot’s manifests, producer, and model shape, and "training snapshots pin" to retain one the catalog would otherwise sweep.

akka optimize training runs snapshots RUN [flags]

Examples

  # List the snapshots a run committed
  akka optimize training runs snapshots 7d496e82c3a15f04

  # Extract the snapshot IDs using jq
  akka optimize training runs snapshots 7d496e82c3a15f04 --jq '.snapshots[].snapshotId'

Options

  -h, --help         help for snapshots
      --limit int    Maximum number of entries to return. The service defaults to 50 and allows at most 100
      --offset int   Index of the first entry to return

Options inherited from parent commands

      --cache-file string    location of cache file (default "~/.akka/cache.yaml")
      --config string        location of config file (default "~/.akka/config.yaml")
      --context string       configuration context to use
      --disable-prompt       Disable all interactive prompts when running akka commands. If input is required, defaults will be used, or an error will be raised.
                             This is equivalent to setting the environment variable AKKA_DISABLE_PROMPTS to true.
      --force                Skip the confirmation prompt
      --jq string            Filter the JSON output with a jq expression
      --json string          Output JSON with only these fields; --json= lists the fields a command offers
      --no-color             Print without color, as NO_COLOR does
  -o, --output string        set output format to one of [text,json,json-compact,go-template=] (default "text")
      --page-mode string     the mode for paging, either paged, buffered or auto. (default "auto")
  -q, --quiet                set quiet output (helpful when used as part of a script)
      --region string        Region to use if the project has more than one region
      --timeout duration     client command timeout (default 10s)
      --trainer-url string   Trainer service URL (default TRAINER_URL, then AKKA_OPTIMIZE_URL, then http://localhost:9001)
      --url string           Optimize service URL (default OPTIMIZE_URL, then AKKA_OPTIMIZE_URL, then http://localhost:9010)
      --use-grpc-web         use grpc-web when talking to Akka APIs. This is useful when behind corporate firewalls that decrypt traffic but don't support HTTP/2.
      --verbose              set verbose output

SEE ALSO