afe.apis.model

Source: afe/apis/model.py

Back to index

Imports

• afe.apis.compilation_job_base.GroundTruth
• afe.apis.defines.ExceptionFuncType
• afe.apis.defines.InputValues
• afe.apis.defines.gen1_target
• afe.apis.release_v1.compose_awesomenets
• afe.apis.release_v1.create_auxiliary_processing_network
• afe.apis.statistic.Statistic
• afe.apis.transform.Transform
• afe.backends.mpk.interface.L2CachingMode
• afe.backends.mpk.interface.TessellateParameters
• afe.core.graph_analyzer.analyzed_results.AnalyzedResultDict
• afe.core.graph_analyzer.graph_analyzer.QuantizedGraphAnalyzer
• afe.core.graph_analyzer.utils.Metric
• afe.core.graph_analyzer.utils.QuantizedGraphAnalyzerMode
• afe.core.utils.LengthHintedIterable
• afe.devkit.devkit_context_manager.SetupConnection
• afe.ir.attributes as afe_attrs
• afe.ir.defines.InputName
• afe.ir.defines.LayerStats
• afe.ir.defines.NodeName
• afe.ir.defines.Status
• afe.ir.net.AwesomeNet
• afe.ir.node.node_is_awesomenet
• afe.ir.node.node_is_ev
• afe.ir.node.node_is_external
• afe.ir.node.node_is_subgraph
• afe.ir.node.node_is_tuple
• afe.ir.node.node_is_tuple_get_item
• afe.ir.sima_ir.SiMaIR
• afe.ir.tensor_type.TensorType
• copy
• devkit_inference_models.utils.connection_params as cp
• logging
• numpy as np
• os
• os.path
• pickle
• re
• sima_utils.common.Platform
• sima_utils.common.print_progressbar
• sima_utils.logging.sima_logger
• tempfile
• typing.Any
• typing.Dict
• typing.Iterable
• typing.List
• typing.Optional
• typing.Tuple

Classes

• Model (line 200)

  • execute(inputs: InputValues, *, fast_mode: bool = False, use_jax: bool = False, log_level: int | None = logging.NOTSET, keep_layer_outputs: list[NodeName] | str | None = None, output_file_path: str | None = None) -> List[np.ndarray] (line 212): Run input data through the quantized model. Decorators: _sanitize_exceptions(ExceptionFuncType.MODEL_EXECUTE).

    Parameters:

    • inputs: type InputValues
    • fast_mode: type bool, default False
    • use_jax: type bool, default False
    • log_level: type int | None, default logging.NOTSET
    • keep_layer_outputs: type list[NodeName] | str | None, default None
    • output_file_path: type str | None, default None

    Returns: List[np.ndarray]

  • save(model_name: str, output_directory: str = '', *, log_level: Optional[int] = logging.NOTSET, include_unquantized_net: bool = True) -> None (line 282) Decorators: _sanitize_exceptions(ExceptionFuncType.MODEL_SAVE).

    Parameters:

    • model_name: type str
    • output_directory: type str, default ''
    • log_level: type Optional[int], default logging.NOTSET
    • include_unquantized_net: type bool, default True

    Returns: None

  • load(model_name: str, network_directory: str = '', *, log_level: Optional[int] = logging.NOTSET, include_unquantized_net: bool = True) -> Model (line 299) Decorators: staticmethod, _sanitize_exceptions(ExceptionFuncType.MODEL_LOAD).

    Parameters:

    • model_name: type str
    • network_directory: type str, default ''
    • log_level: type Optional[int], default logging.NOTSET
    • include_unquantized_net: type bool, default True

    Returns: Model

  • compile(output_path: str, batch_size: int = 1, compress: bool = True, log_level: Optional[int] = logging.NOTSET, tessellate_parameters: Optional[TessellateParameters] = None, l2_caching_mode: L2CachingMode = L2CachingMode.NONE, preserve: bool = True, **kwargs) -> None (line 316): Compile the model and generate MPK JSON file. Decorators: _sanitize_exceptions(ExceptionFuncType.MODEL_COMPILE).

    Parameters:

    • output_path: type str
    • batch_size: type int, default 1
    • compress: type bool, default True
    • log_level: type Optional[int], default logging.NOTSET
    • tessellate_parameters: type Optional[TessellateParameters], default None
    • l2_caching_mode: type L2CachingMode, default L2CachingMode.NONE
    • preserve: type bool, default True
    • kwargs: default {}

    Returns: None

  • create_auxiliary_network(transforms: List[Transform], input_types: Dict[InputName, TensorType], *, target: Platform = gen1_target, log_level: Optional[int] = logging.NOTSET) -> Model (line 370) Decorators: staticmethod, _sanitize_exceptions(ExceptionFuncType.MODEL_CREATE_AUXILIARY).

    Parameters:

    • transforms: type List[Transform]
    • input_types: type Dict[InputName, TensorType]
    • target: type Platform, default gen1_target
    • log_level: type Optional[int], default logging.NOTSET

    Returns: Model

  • compose(nets: List[Model], combined_model_name: str = 'main', log_level: Optional[int] = logging.NOTSET) -> Model (line 392) Decorators: staticmethod, _sanitize_exceptions(ExceptionFuncType.MODEL_COMPOSE).

    Parameters:

    • nets: type List[Model]
    • combined_model_name: type str, default 'main'
    • log_level: type Optional[int], default logging.NOTSET

    Returns: Model

  • evaluate(evaluation_data: Iterable[Tuple[InputValues, GroundTruth]], criterion: Statistic[Tuple[List[np.ndarray], GroundTruth], str], *, fast_mode: bool = False, log_level: Optional[int] = logging.NOTSET) -> str (line 410) Decorators: _sanitize_exceptions(ExceptionFuncType.MODEL_EVALUATE).

    Parameters:

    • evaluation_data: type Iterable[Tuple[InputValues, GroundTruth]]
    • criterion: type Statistic[Tuple[List[np.ndarray], GroundTruth], str]
    • fast_mode: type bool, default False
    • log_level: type Optional[int], default logging.NOTSET

    Returns: str

  • analyze_quantization_error(evaluation_data: Iterable[InputValues], error_metric: Metric, *, local_feed: bool, log_level: Optional[int] = logging.NOTSET) (line 428) Decorators: _sanitize_exceptions(ExceptionFuncType.QUANTIZATION_ERROR_ANALYSIS).

    Parameters:

    • evaluation_data: type Iterable[InputValues]
    • error_metric: type Metric
    • local_feed: type bool
    • log_level: type Optional[int], default logging.NOTSET

  • get_performance_metrics(output_kpi_path: str, *, log_level: Optional[int] = logging.NOTSET) (line 453) Decorators: _sanitize_exceptions(ExceptionFuncType.MODEL_PERFORMANCE).

    Parameters:

    • output_kpi_path: type str
    • log_level: type Optional[int], default logging.NOTSET

  • generate_elf_and_reference_files(input_data: Iterable[InputValues], output_dir: str, *, batch_size: int = 1, compress: bool = True, tessellate_parameters: Optional[TessellateParameters] = None, log_level: Optional[int] = logging.NOTSET, l2_caching_mode: L2CachingMode = L2CachingMode.NONE) -> None (line 473) Decorators: _sanitize_exceptions(ExceptionFuncType.GENERATE_ELF_FILES).

    Parameters:

    • input_data: type Iterable[InputValues]
    • output_dir: type str
    • batch_size: type int, default 1
    • compress: type bool, default True
    • tessellate_parameters: type Optional[TessellateParameters], default None
    • log_level: type Optional[int], default logging.NOTSET
    • l2_caching_mode: type L2CachingMode, default L2CachingMode.NONE

    Returns: None

  • execute_in_accelerator_mode(input_data: Iterable[InputValues], devkit: str, *, username: str = cp.DEFAULT_USERNAME, password: str = '', batch_size: int = 1, compress: bool = True, tessellate_parameters: Optional[TessellateParameters] = None, log_level: Optional[int] = logging.NOTSET, l2_caching_mode: L2CachingMode = L2CachingMode.NONE) -> List[np.ndarray] (line 511) Decorators: _sanitize_exceptions(ExceptionFuncType.GENERATE_ELF_FILES).

    Parameters:

    • input_data: type Iterable[InputValues]
    • devkit: type str
    • username: type str, default cp.DEFAULT_USERNAME
    • password: type str, default ''
    • batch_size: type int, default 1
    • compress: type bool, default True
    • tessellate_parameters: type Optional[TessellateParameters], default None
    • log_level: type Optional[int], default logging.NOTSET
    • l2_caching_mode: type L2CachingMode, default L2CachingMode.NONE

    Returns: List[np.ndarray]