Metadata manipulation

The CatBoost models contain metadata (for example, the list of training parameters or user-defined data) in key-value format. Several operations are provided to manipulate the model's metadata.

Get values of the selected keys

Execution format

catboost metadata get -m <path to the model> --key '<>' .. --key '<>' [--dump-format <output format>]

Options

Option Description Default value

-m

--model-path

The name of the input file with the description of the model obtained as the result of training.

Required key
--model-format

The format of the input model.

Possible values:
  • CatboostBinary.
  • AppleCoreML (only datasets without categorical features are currently supported).
  • json (multiclassification models are not currently supported). Refer to the CatBoost JSON model tutorial for format details.
CatboostBinary
--key

The name of the key to retrieve data from.

The key can be set several times to output multiple values.

Required key
--dump-format

The output format.

Possible values:
  • Plain
  • JSON
Plain
Option Description Default value

-m

--model-path

The name of the input file with the description of the model obtained as the result of training.

Required key
--model-format

The format of the input model.

Possible values:
  • CatboostBinary.
  • AppleCoreML (only datasets without categorical features are currently supported).
  • json (multiclassification models are not currently supported). Refer to the CatBoost JSON model tutorial for format details.
CatboostBinary
--key

The name of the key to retrieve data from.

The key can be set several times to output multiple values.

Required key
--dump-format

The output format.

Possible values:
  • Plain
  • JSON
Plain

Output format

The format of the data that is output to stdout depends on the selected mode:

Each line of the output contains tab-separated data on the key and its' value:

key_1</t>value_1
key_2</t>value_2
...
key_N</t>value_N
For example:
model_name</t>catboost model
params</t>{"detailed_profile":true,"used_ram_limit":18446744073709551615}

Get values of all keys

Execution format

catboost metadata dump -m <path to the model> [--dump-format <output format>]

Options

Option Description Default value

-m

--model-path

The name of the input file with the description of the model obtained as the result of training.

Required key
--dump-format

The output format.

Possible values:
  • Plain
  • JSON
Plain
Option Description Default value

-m

--model-path

The name of the input file with the description of the model obtained as the result of training.

Required key
--dump-format

The output format.

Possible values:
  • Plain
  • JSON
Plain

Output

The format of the data that is output to stdout depends on the selected mode:

Each line of the output contains tab-separated data on the key and its' value:

key_1</t>value_1
key_2</t>value_2
...
key_N</t>value_N
For example:
model_name</t>catboost model
params</t>{"detailed_profile":true,"used_ram_limit":18446744073709551615}

Set a key-value pair

Execution format

catboost metadata set -m <path to the model> --key '<>' --value 'VALUE' []

Options

Option Description Default value

-m

--model-path

The name of the input file with the description of the model obtained as the result of training.

Required key
--key

The name of the key.

Required key
--value

The value that the key should be set to.

The given value overwrites existing data if the specified key matches an existing one.

Required key
-o

The path to the output model.

If defined, the key-value pair and the initial model's key-value pairs are written to the specified model and the input model is not affected by the operation.

Input model (the values of existing keys are overridden)
Option Description Default value

-m

--model-path

The name of the input file with the description of the model obtained as the result of training.

Required key
--key

The name of the key.

Required key
--value

The value that the key should be set to.

The given value overwrites existing data if the specified key matches an existing one.

Required key
-o

The path to the output model.

If defined, the key-value pair and the initial model's key-value pairs are written to the specified model and the input model is not affected by the operation.

Input model (the values of existing keys are overridden)