---
id: ssi4onnx
version: "1.0.4"
license: MIT License
license_treatment: permissive
maintenance: dormant
---

# ssi4onnx — Simple Shape Inference tool for ONNX.

License: permissive · Maintenance: dormant · Downloads: 82.7K/mo

## What it is and what it does

ssi4onnx is a shape inference utility for ONNX models that automatically computes and fills in tensor dimension information that may be missing or incomplete. It works by analyzing the model graph structure and operator definitions to deduce output shapes from input specifications. The tool is useful when working with ONNX models that have been simplified, converted from other frameworks, or manually edited—situations where shape metadata may be incomplete or incorrect.

You can use it either as a command-line tool (passing ONNX files in and out) or as a Python library function that accepts an ONNX ModelProto object directly. The library has no external runtime dependencies beyond ONNX itself, making it lightweight to integrate into model processing pipelines. It's part of a broader ONNX processing toolkit maintained by PINTO0309.

Use it for:

- Repair ONNX models with incomplete shape information before deploying to inference engines like TensorRT or ONNX Runtime.
- Debug shape mismatches in converted models (e.g., from PyTorch or TensorFlow) by inferring correct tensor dimensions.
- Preprocess ONNX graphs in a model optimization pipeline to ensure shape metadata is complete before further transformations.
- Validate ONNX model structure by checking whether shape inference succeeds without errors.
- Integrate shape inference into automated model conversion workflows for consistent shape annotation.

## Worth the install?

AI-flagged interpretation of the facts above — verify before relying.

Infers and fills missing tensor shape information in ONNX model files, available as both a CLI tool and a Python library function.

Yes, if you work with ONNX models and need to repair or validate shape information. The package is lightweight (no runtime dependencies), permissively licensed, and solves a real problem in ONNX workflows. However, note that the project is dormant (last update 808 days ago), so expect no active maintenance or updates for recent ONNX features.

## Install

pip install ssi4onnx
uv add ssi4onnx
poetry add ssi4onnx

## Installing ssi4onnx

Before you install: No runtime dependencies and a pure-Python wheel distribution make installation straightforward. The project is dormant (last update 808 days ago) but carries no known vulnerabilities.

License in practice: MIT License permits unrestricted use, modification, and distribution with minimal restrictions, making it safe to integrate into proprietary or open-source projects.

Quickstart:

pip install ssi4onnx

from ssi4onnx import shape_inference

estimated_graph = shape_inference(
    input_onnx_file_path="model.onnx"
)

Requires ONNX to be installed separately; the package itself has no runtime dependencies but expects an ONNX model file as input.

Verify before relying:

- Whether shape inference handles all ONNX operator types or has known limitations on specific model architectures.
- Performance characteristics on large or complex ONNX graphs.
- Compatibility with recent ONNX opset versions beyond the package's last update.

## Package facts

- License: MIT License (permissive)
- Python support: supports_current
- Install friction: low
- Maintenance: dormant
- Downloads: 82.7K/month (top 15,000 on PyPI)
- Known vulnerabilities: none known

## Tags

onnx shape inference, tensor shape estimation, onnx model processing, infer missing shapes onnx, onnx graph shape completion, neural network model shape analysis, onnx model debugging, onnx-tools, model-processing, shape-inference

[View on SkillFed](https://skillfed.io/packages/ssi4onnx) · [View on PyPI](https://pypi.org/project/ssi4onnx/)