OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
AI Developer Tools & SDKs

Taso

TASO is an open-source C++ tool that optimizes deep neural network computation graphs by generating verified equivalent transformations and using cost-based search to select an optimized graph.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Taso — OpenViST project profile in AI Developer Tools & SDKs.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use TASO when graph-level optimization is part of a DNN workflow and the model can be provided as ONNX or constructed through TASO’s Python interface.

Who it is for

  • Developers working with pre-trained DNN models in ONNX format.
  • Developers who want to construct DNN architectures through TASO’s Python interface.
  • Teams converting TensorFlow or PyTorch models to ONNX for graph optimization.

Documented uses

  • Search equivalent computation graphs for an optimized alternative to a DNN model.
  • Optimize an ONNX model and export the result as ONNX.
  • Build a DNN graph through the Python interface, optimize it, and export the resulting graph.
Before you choose

Requirements & limits

The public repository lists an Apache-2.0 license and provides source-installation instructions and prebuilt Docker images.

  • TensorFlow models require conversion to ONNX with tf2onnx before TASO optimization.
  • PyTorch models are handled in ONNX format.
  • The project exports optimized graphs to ONNX.
  • Installation is available from source or through provided prebuilt Docker images.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Optimize and export an ONNX model

Load a pre-trained ONNX DNN model, optimize its computation graph with TASO, and save the optimized graph as ONNX.

Before you start

  • TASO installed from source or through a provided prebuilt Docker image.
  • A pre-trained DNN model in ONNX format.
  • Python packages and dependencies required by the TASO installation.

Steps

  1. Load the ONNX model with taso.load_onnx.

  2. Pass the loaded model to taso.optimize.

  3. Export the optimized graph with taso.export_onnx.

  4. Save the exported model as an ONNX file.

Expected outcome

An optimized computation graph saved in ONNX format for use as an input to deep-learning frameworks.

Guide limitations

  • The guide documents the ONNX workflow; TensorFlow models must first be converted to ONNX with tf2onnx.
  • The project’s performance examples do not guarantee a particular speedup for a given model or environment.
Documentation 1 ↗Documentation 2 ↗
Evidence and transparency

Sources & review

Official repository ↗Guide documentation ↗
Source record details
Source availability
Active
Sources checked
7 November 2026
Original website title
jiazhihao/TASO · Official homepage title · retained in its source language
How we review sourcesSuggest a correction