akka optimize training endpoints create

Make a model answerable at an endpoint

Synopsis

Bring up an endpoint for a model, or attach the model to the endpoint already serving its base model.

--model takes a selected base model, by repository or registry version, or a candidate, by registry reference NS/NAME[:VERSION]. --project is what the endpoint’s devices count against, and defaults to the training project the model names. A candidate names the training project it was trained for, so it needs no --project; a base model names none, so serving one takes the flag. Given for a candidate, it must be the training project the candidate already names.

Where nothing is serving the base model yet, this claims devices and the endpoint reports STARTING until the process answers. Where something is, the model attaches to it and is answerable at once.

The command returns straight away. Use --wait to return once the endpoint answers, which is what a script needs before it sends anything.

--timeout bounds that wait and leaves the endpoint alone when it expires, and zero waits without a deadline. It means nothing without --wait, so giving it alone is refused.

akka optimize training endpoints create --model MODEL [--project PROJECT] [flags]

Examples

  # Serve a base model, which belongs to no training project
  akka optimize training endpoints create --model Qwen/Qwen3-8B --project triage

  # Serve a candidate over the same base model, under the training project it was trained for
  akka optimize training endpoints create --model triage/candidate:v3

  # Return once it answers
  akka optimize training endpoints create --model triage/candidate:v3 --wait

Options

  -h, --help               help for create
      --model string       Selected base model or candidate to serve, as a registry reference NS/NAME[:VERSION]. With --project, NAME[:VERSION] names a model in that project's namespace
      --project string     Training project whose GPU quota the endpoint's devices count against. Defaults to the training project the model names, and must agree with it when given
      --timeout duration   Maximum time to wait for the endpoint; zero waits without a deadline; requires --wait (default 15m0s)
      --wait               Wait until the endpoint answers

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
      --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