simple-onnx-processing-tools

What it is and what it does

simple-onnx-processing-tools is a metapackage that bundles 25 specialized command-line utilities for ONNX model manipulation. Each tool addresses a specific transformation task: merging multiple models, extracting subgraphs, deleting unused nodes, shrinking file size by deduplicating constants, modifying operation attributes, changing opsets, converting between NCHW and NHWC layouts, adjusting batch dimensions, renaming operations, and bidirectional JSON serialization.

The package is designed for model engineers and researchers who need to preprocess or postprocess ONNX graphs before deployment or inference. Rather than a single monolithic API, it provides a suite of focused CLI tools that can be chained or used independently. Installation includes optional dependencies (via the [full] extra) for heavier operations like model simplification and runtime testing. The tools assume familiarity with ONNX graph structure and are most useful in workflows where model optimization, format conversion, or structural modification is required.

Use it for:

• Merge multiple trained ONNX models into a single graph for ensemble inference.
• Split a large ONNX model to stay under the 2GB Protocol Buffers file size limit.
• Remove unused nodes and operations to reduce model file size and inference latency.
• Convert model channel layout from RGB to BGR or NCHW to NHWC for different inference frameworks.
• Adjust batch size dimensions in pre-trained models to match deployment requirements.
• Export ONNX graphs to JSON for inspection, debugging, or programmatic analysis.

Worth the install?

AI-flagged interpretation of the facts on this page — verify before relying

A collection of command-line tools for transforming ONNX models: merging, splitting, deleting nodes, compressing size, modifying attributes and constants, changing opsets, converting channel formats, adjusting batch sizes, and converting to/from JSON.

Yes, if you work with ONNX models and need CLI-based transformation tools. The package is stable (low install friction, no known vulnerabilities, permissive license) but dormant—last updated 844 days ago. Install it for specific model preprocessing tasks, but do not expect active maintenance or new features. Verify tool compatibility with your ONNX opset version before relying on it in production pipelines.

Install

simple-onnx-processing-tools on PyPI

pip install simple-onnx-processing-tools

uv add simple-onnx-processing-tools

poetry add simple-onnx-processing-tools

Installing simple-onnx-processing-tools

Before you install

Low install friction with a pure-Python wheel. Maintenance is dormant—last release was 844 days ago (April 2024)—but the repository remains active and unarchived with 305 stars. No recent updates suggest the tools are stable but not under active development.

License in practice

MIT License permits commercial and private use with minimal restrictions, making it safe to adopt for most projects.

Quickstart

pip install -U simple-onnx-processing-tools
pip install -U onnx

# Then use individual tools, e.g.:
# snc4onnx --help  # merge models
# snd4onnx --help  # delete nodes
# scs4onnx --help  # compress constants

Requires onnx and optionally onnx_graphsurgeon (from NVIDIA's index). Full installation with [full] extra pulls additional dependencies like onnxruntime and onnx-simplifier.

Verify before relying

• Whether individual tools (snc4onnx, sne4onnx, etc.) are installed as separate executables or accessed programmatically via imports.
• Whether the package works with modern ONNX opset versions or has compatibility constraints.
• Performance characteristics when processing large models or performing batch operations.

Package facts

Evidence: simple_onnx_processing_tools-1.1.32-py3-none-any.whl

Tags

More Artificial Intelligence packages