ONNX Script

ONNX Script enables developers to naturally author ONNX functions and models using a subset of Python. ONNX Script is:

Expressive: enables the authoring of all ONNX functions.
Simple and concise: function code is natural and simple.
Debuggable: allows for eager-mode evaluation that provides for a more delightful ONNX model debugging experience.

This repo also covers:

ONNX IR: an in-memory IR that supports the full ONNX spec, designed for graph construction, analysis and transformation.
ONNX Script Optimizer: provides functionality to optimize an ONNX model by performing optimizations and clean-ups such as constant folding, dead code elimination, etc.
ONNX Rewriter: provides functionality to replace certain patterns in an ONNX graph with replacement patterns based on user-defined rewrite rules.

Note however that ONNX Script does not intend to support the entirety of the Python language.

Website: https://onnxscript.ai/

Design Overview

ONNX Script provides a few major capabilities for authoring and debugging ONNX models and functions:

A converter which translates a Python ONNX Script function into an ONNX graph, accomplished by traversing the Python Abstract Syntax Tree to build an ONNX graph equivalent of the function.
A converter that operates inversely, translating ONNX models and functions into ONNX Script. This capability can be used to fully round-trip ONNX Script ↔ ONNX graph.
A runtime shim that allows such functions to be evaluated (in an "eager mode"). This functionality currently relies on ONNX Runtime for executing every ONNX Operator, and there is a Python-only reference runtime for ONNX underway that will also be supported.

Note that the runtime is intended to help understand and debug function definitions. Performance is not a goal here.

Installing ONNX Script

pip install --upgrade onnxscript

Install for Development

git clone https://github.com/microsoft/onnxscript
cd onnxscript
pip install -r requirements-dev.txt
pip install -e .

Run Unit Tests

pytest .

Example

import onnx

# We use ONNX opset 15 to define the function below.
from onnxscript import FLOAT, script
from onnxscript import opset15 as op


# We use the script decorator to indicate that
# this is meant to be translated to ONNX.
@script()
def onnx_hardmax(X, axis: int):
    """Hardmax is similar to ArgMax, with the result being encoded OneHot style."""

    # The type annotation on X indicates that it is a float tensor of
    # unknown rank. The type annotation on axis indicates that it will
    # be treated as an int attribute in ONNX.
    #
    # Invoke ONNX opset 15 op ArgMax.
    # Use unnamed arguments for ONNX input parameters, and named
    # arguments for ONNX attribute parameters.
    argmax = op.ArgMax(X, axis=axis, keepdims=False)
    xshape = op.Shape(X, start=axis)
    # use the Constant operator to create constant tensors
    zero = op.Constant(value_ints=[0])
    depth = op.GatherElements(xshape, zero)
    empty_shape = op.Constant(value_ints=[0])
    depth = op.Reshape(depth, empty_shape)
    values = op.Constant(value_ints=[0, 1])
    cast_values = op.CastLike(values, X)
    return op.OneHot(argmax, depth, cast_values, axis=axis)


# We use the script decorator to indicate that
# this is meant to be translated to ONNX.
@script()
def sample_model(X: FLOAT[64, 128], Wt: FLOAT[128, 10], Bias: FLOAT[10]) -> FLOAT[64, 10]:
    matmul = op.MatMul(X, Wt) + Bias
    return onnx_hardmax(matmul, axis=1)


# onnx_model is an in-memory ModelProto
onnx_model = sample_model.to_model_proto()

# Save the ONNX model at a given path
onnx.save(onnx_model, "sample_model.onnx")

# Check the model
try:
    onnx.checker.check_model(onnx_model)
except onnx.checker.ValidationError as e:
    print(f"The model is invalid: {e}")
else:
    print("The model is valid!")

The decorator parses the code of the function, converting it into an intermediate representation. If it fails, it produces an error message indicating the line where the error was detected. If it succeeds, the intermediate representation can be converted into an ONNX graph structure of type FunctionProto:

Hardmax.to_function_proto() returns a FunctionProto

Eager Mode Evaluation

Eager mode is mostly used to debug and validate that intermediate results are as expected. The function defined above can be called as below, executing in an eager-evaluation mode:

import numpy as np

v = np.array([[0, 1], [2, 3]], dtype=np.float32)
result = Hardmax(v)

More examples can be found in the docs/examples directory.

ONNX IR

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

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.

ONNX Script Tools

ONNX Optimizer

The ONNX Script Optimizer tool provides the user with the functionality to optimize an ONNX model by performing optimizations and clean-ups such as constant folding, dead code elimination, etc. In order to utilize the optimizer tool:

import onnxscript

onnxscript.optimizer.optimize(onnx_model)

For a detailed summary of all the optimizations applied by the optimizer call, refer to the tutorial Optimizing a Model using the Optimizer

ONNX Rewriter

The ONNX Rewriter tool provides the user with the functionality to replace certain patterns in an ONNX graph with another pattern based on user-defined rewrite rules. The rewriter tools allows two different methods in which patterns in the graph can be rewritten.

Pattern-based rewriting

For this style of rewriting, the user provides a target_pattern that is to be replaced, a replacement_pattern and a match_condition (pattern rewrite will occur only if the match condition is satisfied). A simple example on how to use the pattern-based rewriting tool is as follows:

from onnxscript.rewriter import pattern

# The target pattern
def erf_gelu_pattern(op, x):
    return 0.5 * (x * (op.Erf(x / math.sqrt(2)) + 1.0))

def erf_gelu_pattern_2(op, x):
    return (x * (op.Erf(x / math.sqrt(2)) + 1.0)) * 0.5

# The replacement pattern
def gelu(op, x: ir.Value):
    return op.Gelu(x, domain="com.microsoft")

# Create multiple rules
rule1 = pattern.RewriteRule(
    erf_gelu_pattern,  # Target Pattern
    gelu,  # Replacement
)
rule2 = pattern.RewriteRule(
    erf_gelu_pattern_2,  # Target Pattern
    gelu,  # Replacement
)
# Create a Rewrite Rule Set with multiple rules.
rewrite_rule_set = pattern.RewriteRuleSet([rule1, rule2])
# Apply rewrites
model_with_rewrite_applied = onnxscript.rewriter.rewrite(
    model,  # Original ONNX Model
    pattern_rewrite_rules=rewrite_rule_set,
)
return model_with_rewrite_applied

For a detailed tutorial on how to create target_pattern, replacement_pattern and match_condition blocks in order to utilize the pattern-based rewriter, refer to the tutorial Pattern-based Rewrite Using Rules

Function-based rewriting

This style of rewriting matches a FUNCTION_KEYWORD and PACKAGE_NAME provided by the user to an existing function within the graph and replaces it with a new function provided by the user.

Development Guidelines

Every change impacting the converter or the eager evaluation must be unit tested with class OnnxScriptTestCase to ensure both systems do return the same results with the same inputs.

Coding Style

We use ruff, black, isort, and mypy etc. to check code formatting and use lintrunner to run all linters. You can install the dependencies and initialize with

pip install lintrunner lintrunner-adapters
lintrunner init

This will install lintrunner on your system and download all the necessary dependencies to run linters locally. If you want to see what lintrunner init will install, run lintrunner init --dry-run.

To lint local changes:

lintrunner

To format files:

lintrunner f

To lint all files:

lintrunner --all-files

Use --output oneline to produce a compact list of lint errors, useful when there are many errors to fix.

See all available options with lintrunner -h.

To read more about lintrunner, see wiki. To update an existing linting rule or create a new one, modify .lintrunner.toml or create a new adapter following examples in https://github.com/justinchuby/lintrunner-adapters.

Contributing

We're always looking for your help to improve the product (bug fixes, new features, documentation, etc). Currently ONNX Script is under early and heavy development, so we encourage proposing any major changes by filing an issue to discuss your idea with the team first.

Report a Security Issue

Please do not report security vulnerabilities through public GitHub issues.

Please refer to our guidance on filing Security Issues.

Licensing Guidelines

This project welcomes contributions and suggestions. Most contributions require you to agree to a Contributor License Agreement (CLA) declaring that you have the right to, and actually do, grant us the rights to use your contribution. For details, visit https://cla.microsoft.com.

When you submit a pull request, a CLA-bot will automatically determine whether you need to provide a CLA and decorate the PR appropriately (e.g., label, comment). Simply follow the instructions provided by the bot. You will only need to do this once across all repositories using our CLA.

Code of Conduct

This project has adopted the Microsoft Open Source Code of Conduct. For more information see the Code of Conduct FAQ or contact opencode@microsoft.com with any additional questions or comments.

Trademarks

This project may contain trademarks or logos for projects, products, or services. Authorized use of Microsoft trademarks or logos is subject to and must follow Microsoft's Trademark & Brand Guidelines. Use of Microsoft trademarks or logos in modified versions of this project must not cause confusion or imply Microsoft sponsorship. Any use of third-party trademarks or logos is subject to those third-party's policies.