hummingbird.ml.operator_converters.constants

Constants used in the Hummingbird converters are defined here.

hummingbird.ml.operator_converters.constants.BASE_PREDICTION = 'base_prediction'

base prediction for ensemble models requiring it.

hummingbird.ml.operator_converters.constants.FEATURE_NAMES = 'feature_names'

Names of the features used for training.

hummingbird.ml.operator_converters.constants.GET_PARAMETERS_FOR_TREE_TRAVERSAL = 'get_parameters_for_tree_trav'

Which function to use to extract the parameters for the tree traversal strategy.

hummingbird.ml.operator_converters.constants.IFOREST_THRESHOLD = 'iforest_threshold'

threshold of the sklearn isolation forest implementation, backward compatibility for sklearn <= 0.21.

hummingbird.ml.operator_converters.constants.LEARNING_RATE = 'learning_rate'

Learning Rate.

hummingbird.ml.operator_converters.constants.MAX_SAMPLES = 'max_samples'

max_samples of sklearn isolation forest implementation.

hummingbird.ml.operator_converters.constants.NUM_TREES = 'n_trees'

Number of trees composing an ensemble.

hummingbird.ml.operator_converters.constants.N_FEATURES = 'n_features'

Number of features expected in the input data.

hummingbird.ml.operator_converters.constants.N_INPUTS = 'n_inputs'

Number of inputs expected by the model.

hummingbird.ml.operator_converters.constants.OFFSET = 'offset'

offset of the sklearn anomaly detection implementation.

hummingbird.ml.operator_converters.constants.ONNX_INITIALIZERS = 'onnx_initializers'

The initializers of the onnx model.

hummingbird.ml.operator_converters.constants.POST_TRANSFORM = 'post_transform'

Post transform for tree inference.

hummingbird.ml.operator_converters.constants.REORDER_TREES = 'reorder_trees'

Whether to reorder trees in multiclass tasks.

hummingbird.ml.operator_converters.constants.SAVE_LOAD_CONTAINER_PATH = 'container.pkl'

Path where to find the container when saving or loading.

hummingbird.ml.operator_converters.constants.SAVE_LOAD_MODEL_CONFIGURATION_PATH = 'model_configuration.txt'

Path where to find the modules versions used when saving or loading.

hummingbird.ml.operator_converters.constants.SAVE_LOAD_MODEL_TYPE_PATH = 'model_type.txt'

Path where to find the model type when saving or loading.

hummingbird.ml.operator_converters.constants.SAVE_LOAD_ONNX_PATH = 'deploy_model.onnx'

Path where to find the onnx model when saving or loading.

hummingbird.ml.operator_converters.constants.SAVE_LOAD_TORCH_JIT_PATH = 'deploy_model.zip'

Path where to find the torchscript model when saving or loading.

hummingbird.ml.operator_converters.constants.SAVE_LOAD_TVM_GRAPH_PATH = 'deploy_graph.json'

Path where to find the TVM graph when saving or loading.

hummingbird.ml.operator_converters.constants.SAVE_LOAD_TVM_LIB_PATH = 'deploy_lib.tar'

Path where to find the TVM lib when saving or loading.

hummingbird.ml.operator_converters.constants.SAVE_LOAD_TVM_PARAMS_PATH = 'deploy_param.params'

Path where to find the TVM params when saving or loading.

hummingbird.ml.operator_converters.constants.SIGMOID = 'LOGISTIC'

Sigmoid post transform.

hummingbird.ml.operator_converters.constants.SOFTMAX = 'SOFTMAX'

Softmax post transform.

hummingbird.ml.operator_converters.constants.SUPPORTED_STRING_TYPES = {'S', 'U'}

Numpy string types suppoted by Humingbird.

hummingbird.ml.operator_converters.constants.TEST_INPUT = 'test_input'

The test input data for models that need to be traced.

hummingbird.ml.operator_converters.constants.TVM_CONTEXT = 'tvm_context'

The context for TVM containing information on the target.

hummingbird.ml.operator_converters.constants.TVM_GRAPH = 'tvm_graph'

The graph defining the TVM model. This parameter is used for saving and loading a TVM model.

hummingbird.ml.operator_converters.constants.TVM_INPUT_NAMES = 'tvm_input_names'

TVM expects named inputs. This is used to set the names for the inputs.

hummingbird.ml.operator_converters.constants.TVM_LIB = 'tvm_lib'

The lib for the TVM model. This parameter is used for saving and loading a TVM model.

hummingbird.ml.operator_converters.constants.TVM_PARAMS = 'tvm_params'

The params for the TVM model. This parameter is used for saving and loading a TVM model.

hummingbird.ml.operator_converters.constants.TWEEDIE = 'TWEEDIE'

Tweedie post transform.