Model

Module: polygraphy.tools.args

class ModelArgs(model_opt_required: bool | None = None, required_model_type: str | None = None, input_shapes_opt_name: str | None = None, input_shapes_opt_doc: str | None = None, guess_model_type_from_runners: bool | None = None)[source]

Bases: BaseArgs

Model: the model

Depends on:

  • RunnerSelectArgs: if guess_model_type_from_runners == True

Parameters:
  • model_opt_required (bool) – Whether the model argument is required. Defaults to False.

  • required_model_type (str) – The required type of model. Use a value of None for tools that work with multiple model types. If provided, it causes the tool to support only one type of model and disables the --model-type option. Defaults to None.

  • input_shapes_opt_name (str) – The name of the option used to specify input shapes. A second option name will be automatically added by dropping the final s in the specified option and suffixing -shapes. For example, a value of “inputs” would generate an alias called “–input-shapes”. Defaults to “inputs”. Use a value of False to disable the option.

  • input_shapes_opt_doc (str) – Custom help text output for the input shapes option.

  • guess_model_type_from_runners (bool) – Whether to guess the model type based on which runners have been specified, if any. Defaults to False.

class ModelType(model_type)[source]

Bases: str

parse_impl(args)[source]

Parses command-line arguments and populates the following attributes:

input_shapes

Input names and their shapes.

Type:

TensorMetadata

path

Path to the model.

Type:

str

model_type

The type of model.

Type:

ModelArgs.ModelType

extra_model_info

Any extra model information specified after the model argument, separated by a colon. The meaning of this information may be specific to each model type. In most cases, no extra model information is provided.

Type:

str