Working with models¶

$ om models -h
Usage:
  om models list [<pattern>] [--raw] [-E|--regexp] [options]
  om models put <module.callable> <name>
  om models drop <name>
  om models metadata <name>

Storing a model¶

Models can be stored directly from any of your project’s Python modules.

create a Python module and a callable that returns your model instance

# mymodule.py
from sklearn.linear_model import LinearRegression

def create_model():
    reg = LinearRegression()
    return reg

Use the cli to store the model
$ om models put mymodule.create_model regmodel Metadata(name=regmodel,bucket=omegaml,prefix=models/,kind=sklearn.joblib,created=2021-02-18 20:55:07.161042)
This will call the create_model() function and store the resulting model as regmodel. The model is stored in as returned. The create_model can return an unfitted, a fitted or a partially fitted model.

Use the Python shell

$ om shell
[] from mymodule import create_model

   mdl = create_model()
   om.models.put(mdl, 'regmodel')
   => Metadata(name=regmodel,bucket=omegaml,prefix=models/,kind=sklearn.joblib,created=2021-02-18 20:55:07.161042)

Listing models¶

Listing models follows the common pattern syntax. No pattern is equal to the generic pattern *, which finds all contents

$ om models list reg*
['regmodel']

Using the --raw option prints the corresponding Metadata entries:

$ om models list reg* --raw
[<Metadata: Metadata(name=regmodel,bucket=omegaml,prefix=models/,kind=sklearn.joblib,created=2020-06-18 14:38:30.627000)>]

Dropping models¶

Dropping models requires the specific model name:

$ om models drop regmodel
True

If the model does not exist the command will print an error message:

$ om models drop regmodel
*** ERROR

Showing a model’s metadata¶

A models’ metadata is shown as a JSON object.

Note

The content of the metadata is specific to the model’s type.

$ om models metadata regmodel
{
  "_id": {
    "$oid": "602ec61b0664b73eb299d1e5"
  },
  "name": "regmodel",
  "bucket": "omegaml",
  "prefix": "models/",
  "kind": "sklearn.joblib",
  "kind_meta": {
    "_om_backend_version": "1"
  },
  "attributes": {
    "versions": {
      "tags": {
        "latest": "b434c293f03a925a3932895ca1f11a7144093db7"
      },
      "commits": [
        {
          "name": "_versions/regmodel/b434c293f03a925a3932895ca1f11a7144093db7",
          "ref": "b434c293f03a925a3932895ca1f11a7144093db7"
        }
      ],
      "tree": {
        "b434c293f03a925a3932895ca1f11a7144093db7": "latest"
      }
    }
  },
  "s3file": {},
  "created": {
    "$date": 1613681707161
  },
  "modified": {
    "$date": 1613681707433
  },
  "gridfile": {
    "$oid": "602ec61b0664b73eb299d1e3"
  }
}