create-version

Description

Create a version using the current directory as the source code.

Synopsis

$ abeja training create-version [--help]
Usage: abeja training create-version [OPTIONS]

Options:
  -j, --job_definition_name, --job-definition-name TEXT
                                  Training job definition name
  -h, --handler TEXT              Training handler
  -i, --image TEXT                Specify base image name and tag in the
                                  "name:tag" format. ex) abeja-inc/all-gpu:19.10
  -d, --description TEXT          Description for the training job, which must be less
                                  than or equal to 256 characters.
  -e, --environment ENVIRONMENTSTRING
                                  Environment variables, ex. BATCH_SIZE:32
  --exclude TEXT                  Specify the file/dir to exclude from create-version.
  --dataset, --datasets DATASETPARAMSTRING
                                  Datasets name
  --help                          Show this message and exit.

Argument

Get the parameters from the training configuration file (training.yaml). Upload the current directory as source code.

Options

-j, --job_definition_name, --job-definition-name

Training job definition name

With training.yaml, values defined as name in training.yaml is set by default.
This option can overwrite name in training.yaml.

-h, --handler

Path for call function

With training.yaml, values defined as handler in training.yaml is set by default.
This option can overwrite handler in training.yaml.

-i, --image

Image for training model

With training.yaml, values defined as image in training.yaml is set by default.
This option can overwrite image in training.yaml.

-d, --description

Explanation of training definition version

-e, --environment

Specify an environment variable. Registered environment variables can be referenced from the code. e.g.）IMAGE_WIDTH:100
For more information on user-specifiable environment variables, see here.

( version 0.14.0 or later ) With training.yaml, values defined as environment ( params ) in training.yaml are set by default.
This option can overwrite environment ( params ) in training.yaml.

--exclude

Specify files / directories to be excluded from the version to be created.
If you set ignores in the learning configuration file (training.yaml) , both are valid.
If you want to specify more than one, repeat --exclude <value1> --exclude <value2> ....

--dataset, --datasets

Using for training model

With training.yaml, values defined as datasets in training.yaml is set by default.
This option can overwrite datasets in training.yaml.

-e option is used as an abbreviation for --environment option in version 0.14.0 or later

Example

To make a training version

Create a training version in this example

Default training configuration file (training.yaml):

name: training1
handler: train:handler
image: abeja-inc/all-gpu:19.04
datasets:
  "mnist": "1111111111111"

Command:

$ abeja training create-version

Output:

{
    "created_at": "2018-02-13T10:13:10.956198Z",
    "datasets": {
        "mnist": "1111111111111"
    },
    "handler": "train:handler",
    "image": "abeja-inc/all-gpu:19.04",
    "job_definition_id": "1368416720943",
    "job_definition_version": 1,
    "modified_at": "2018-02-13T10:13:11.000461Z",
    "user_parameters": {}
}