ferricml 0.2.1

A lean, pure-Rust toolkit for classical machine learning
Documentation
//! Stable, bounded model-artifact errors and format identity.

mod contract;
mod cursor;
mod envelope;
mod error;
mod logical_tree;

pub use contract::{ModelArtifact, StageArtifact};
pub(crate) use cursor::ArtifactCursor;
pub(crate) use envelope::{
    ANY_CLASSIFIER_ARTIFACT_KIND, ANY_REGRESSOR_ARTIFACT_KIND, ArtifactPayloadWriter,
    DECISION_TREE_CLASSIFIER_ARTIFACT_KIND, DECISION_TREE_REGRESSOR_ARTIFACT_KIND,
    ELASTIC_NET_ARTIFACT_KIND, EXTRA_TREES_CLASSIFIER_ARTIFACT_KIND,
    EXTRA_TREES_REGRESSOR_ARTIFACT_KIND, HIST_GRADIENT_BOOSTING_CLASSIFIER_ARTIFACT_KIND,
    HIST_GRADIENT_BOOSTING_REGRESSOR_ARTIFACT_KIND, LASSO_ARTIFACT_KIND,
    LINEAR_REGRESSION_ARTIFACT_KIND, LOGISTIC_ARTIFACT_KIND, MAX_ABS_SCALER_ARTIFACT_KIND,
    MIN_MAX_SCALER_ARTIFACT_KIND, PAIRWISE_LINEAR_RANKER_ARTIFACT_KIND,
    POLYNOMIAL_FEATURES_ARTIFACT_KIND, RANDOM_FOREST_CLASSIFIER_ARTIFACT_KIND,
    RANDOM_FOREST_REGRESSOR_ARTIFACT_KIND, RIDGE_ARTIFACT_KIND, ROBUST_SCALER_ARTIFACT_KIND,
    STAGED_PIPELINE_ARTIFACT_KIND, STANDARD_SCALER_ARTIFACT_KIND,
    STANDARD_SCALER_LINEAR_PIPELINE_ARTIFACT_KIND, STANDARD_SCALER_LOGISTIC_PIPELINE_ARTIFACT_KIND,
    STANDARD_SCALER_RIDGE_PIPELINE_ARTIFACT_KIND, SchemaRole, artifact_payload_version,
    artifact_version, decode_component, decode_legacy_envelope, decode_v2_envelope,
    encode_component, encode_v2_envelope,
};
pub use error::ArtifactError;
pub(crate) use logical_tree::{
    LogicalTreeNode, MIN_ENCODED_TREE_BYTES, decode_logical_tree, encode_logical_tree,
};

/// Current FerricML binary model artifact version.
pub const MODEL_ARTIFACT_VERSION: u16 = 2;

/// Maximum accepted size of one encoded model artifact.
pub const MAX_MODEL_ARTIFACT_BYTES: usize = 32 * 1024 * 1024;