Crate onnx_runtime_loader 

onnx-runtime-loader

Loads ONNX models from disk into the onnx_runtime_ir::Graph IR (see docs/ORT2.md §19).

Pipeline (load_model / load_model_with_weights):

1. proto — decode the ONNX protobuf (prost types generated from the vendored onnx.proto3) into a ModelProto.
2. [graph_builder] — build an onnx_runtime_ir::Graph (nodes, values, symbolic dim interning, opset imports), upholding the §3.5 invariants.
3. weights — resolve inline and external initializer data (external files are memory-mapped into a WeightStore).
4. Static/symbolic shape inference via onnx-runtime-shape-inference: the loader owns the “loader = shape-inference” seam, so after the Graph is built (with initializers applied) it runs the extensible per-op registry to populate every value’s shape and dtype. Values that cannot be resolved statically (genuinely data-dependent extents) are left symbolic for the session to resolve just-in-time.

Obtaining weight bytes at session time

Use load_model_with_weights (or load_model_bytes_with_weights) to receive both the Graph and an Arc<WeightStore>. Then, given any onnx_runtime_ir::WeightRef stored in graph.initializers, call WeightStore::bytes to get the raw little-endian byte slice:

let (graph, store) = load_model_with_weights("model.onnx")?;
for (vid, weight_ref) in &graph.initializers {
    let bytes: &[u8] = store.bytes(weight_ref).expect("weight bytes live");
    // ... hand bytes to a kernel
}

The Arc keeps all memory maps alive as long as any clone of it exists, so kernel dispatch can store Arc<WeightStore> alongside the Graph without lifetime coupling.

Re-exports

pub use encoder::DEFAULT_IR_VERSION;

pub use encoder::Model;

pub use encoder::ModelMetadata;

pub use encoder::encode_model;

pub use encoder::encode_model_proto;

pub use encoder::write_model;

pub use epcontext::EmbedMode;

pub use epcontext::EpContextBlob;

pub use epcontext::EpContextNode;

pub use epcontext::ep_context_node_ids;

pub use epcontext::ep_context_nodes;

pub use epcontext::is_ep_context_op;

pub use epcontext::resolve_ep_context;

pub use weights::ExpertQuantization;

pub use weights::ExpertStorageOrder;

pub use weights::ExpertTensorLayout;

pub use weights::ExpertWeightRegion;

pub use weights::NonPageableReason;

pub use weights::Pageability;

pub use weights::WeightRegionCatalog;

pub use weights::WeightStore;

pub use writer::EpContextDumpConfig;

pub use writer::EpContextPartition;

pub use writer::dump_ep_context;

Modules

encoder

ONNX protobuf encoding — the inverse of graph_builder and weights (§19.1, §55.4 dump path).

epcontext

com.microsoft::EPContext node support — the load path (§55.3).

function_inline

Model-local function inlining (ONNX function expansion) at load time.

proto

ONNX protobuf decoding (§19.1).

weights

Initializer weight resolution: inline data and external mmap (§19.2, §12).

writer

com.microsoft::EPContext dump / writer path (§55.4) — the inverse of the load path in crate::epcontext.

Enums

LoaderError

Errors produced while loading an ONNX model.

Functions

load_model

Load a model from a filesystem path, producing a fully-built Graph.

load_model_bytes

Load a model from an in-memory protobuf buffer, producing a Graph.

load_model_bytes_with_weights

Load a model from an in-memory protobuf buffer, returning both the Graph and the live WeightStore that backs all initializer data.

load_model_with_weights

Load a model from a filesystem path, returning both the Graph and the live WeightStore that backs all initializer data.

validate_model

Fail-fast, load-time validation of everything statically knowable to be illegal or unsupported (RULES #1: fail at load, never via a silent sentinel at run time).

validate_model_proto

Validate ONNX model metadata and protobuf-level graph invariants that are intentionally not preserved by the runtime IR (notably ir_version and AttributeProto::ref_attr_name).

validate_no_control_flow

Reject subgraph-bearing (control-flow) ops the runtime cannot execute.

validate_no_dangling_refs

Reject graphs with a node input that has no source.

validate_no_initializer_producer

Reject graphs where an initializer value is also produced by a node.

validate_opset_imports

ONNX treats "" and "ai.onnx" as equivalent spellings of the default domain. Model-level imports also govern nodes nested in subgraphs.