onnx-ir 0.2.1

Efficient in-memory representation for ONNX

An in-memory IR that supports the full ONNX spec, designed for graph construction, analysis and transformation.

Getting Started

onnx-ir documentation

Installation

Via pip:

pip install onnx-ir

Or from source:

pip install git+https://github.com/onnx/ir-py.git

Features ✨

• Full ONNX spec support: all valid models representable by ONNX protobuf, and a subset of invalid models (so you can load and fix them).
• Low memory footprint: mmap'ed external tensors; unified interface for ONNX TensorProto, Numpy arrays and PyTorch Tensors etc. No tensor size limitation. Zero copies.
• Straightforward access patterns: Access value information and traverse the graph topology at ease.
• Robust mutation: Create as many iterators as you like on the graph while mutating it.
• Speed: Performant graph manipulation, serialization/deserialization to Protobuf.
• Pythonic and familiar APIs: Classes define Pythonic apis and still map to ONNX protobuf concepts in an intuitive way.
• No protobuf dependency: The IR does not require protobuf once the model is converted to the IR representation, decoupling from the serialization format.

Concept Diagram

Code Organization 🗺️

• _protocols.py: Interfaces defined for all entities in the IR.
• _core.py: Implementation of the core entities in the IR, including Model, Graph, Node, Value, and others.
• _enums.py: Definition of the type enums that correspond to the DataType and AttributeType in onnx.proto.
• _name_authority.py: The authority for giving names to entities in the graph, used internally.
• _linked_list.py: The data structure as the node container in the graph that supports robust iteration and mutation. Internal.
• _metadata.py: Metadata store for all entities in the IR.