Module converter 

APR Converter Module - Import Pipeline

Implements Section 13 of APR-SPEC.md: Import/Convert Pipeline

Supports:

• HuggingFace Hub downloads (hf://org/repo)
• SafeTensors conversion
• Inline validation during conversion
• Quantization and compression

Re-exports

pub use crate::format::converter_types::detect_sharded_model;

pub use crate::format::converter_types::Architecture;

pub use crate::format::converter_types::DequantizedTensors;

pub use crate::format::converter_types::ImportError;

pub use crate::format::converter_types::ImportOptions;

pub use crate::format::converter_types::NativeF32Tensors;

pub use crate::format::converter_types::QuantizationType;

pub use crate::format::converter_types::ShardedIndex;

pub use crate::format::converter_types::Source;

pub use crate::format::converter_types::TensorExpectation;

pub use crate::format::converter_types::TensorProvenance;

pub use crate::format::converter_types::ValidationConfig;

Structs

AprConverter

APR Converter with builder pattern

ConvertOptions

Options for model conversion

ConvertReport

Report from model conversion

ExportOptions

Options for model export

ExportReport

Report from export operation

MergeOptions

Options for model merging

MergeReport

Report from merge operation

Enums

ExportFormat

Export format options

MergeStrategy

Merge strategy options

Constants

F16_MIN_NORMAL

Minimum valid f16 normal value (~6.1e-5) Prevents NaN on round-trip through f16 encoding

Functions

apr_convert

Convert a model with quantization and/or compression

apr_export

Export APR/SafeTensors model to another format

apr_import

apr_merge

Merge multiple models into one