Release 0.1.0

Release Notes: v0.1.0

Overview

• Initial release of minisom2onnx for converting MiniSom models to ONNX format.

Features

• Model Conversion: Export trained MiniSom models to ONNX with support for:
  - Distance computation.
  - Quantization and quantization error.
  - BMU coordinates.
• Optional Features:
  - Thresholding based on a specified quantization error threshold.
  - Label mapping for class prediction.
• Output Filtering: Choose specific outputs to include in the ONNX model.

Validation & Error Handling

• Distance Function Check: Ensures use of supported distance functions; errors raised for custom functions.
• Clear Guidance on Invalid Inputs: Provides detailed error messages for invalid inputs.

Known Limitations

• Distance Compatibility: Only euclidean distance is supported by ONNX Runtime.
• Custom Distance Functions: Unsupported in ONNX conversion.

Documentation

Includes usage guide, examples, and notes on limitations.

Installation

Available on PyPI: pip install minisom2onnx.

Full Changelog: https://github.com/Chiragasourabh/minisom-onnx/commits/v0.1.0