Overview
This document describes the SoftNeuro SDK.
Getting Started
Please start at the Getting Started document.
This will proceed with the SDK setup, execute the Command Line Interface (CLI) tool and perform Python inference.
Components
SoftNeuro SDK consists of three components: C API, Python API and CLI tool.
Documentation for individual components are provided below:
- CLI Tool Manual : The manual of SoftNeuro CLI tool for using various features of SoftNeuro from the command prompt.
- C API Specification : Describes the SoftNeuro API that can be used within C programs.
- Python API Specification : Describes the SoftNeuro API that can be used within Python programs.
Supported Environments
SoftNeuro SDK supports environmets as below:
OS(Architecture) | OpenCL | CUDA+cudnn | Python |
---|---|---|---|
Linux(x86_x64) | 2.0~ | CUDA10.2+cudnn7.6, CUDA10.2+cudnn8.2 | 3.6~3.8 |
Linux(armv8/armv8.2) | 2.0~ | CUDA10.0+cudnn7.3, CUDA10.2+cudnn8.0 | 3.6~3.8 |
Windows(x86_x64) | 2.0~ | Unsupported | 3.6~3.8 |
TensorFlow model converter is tested with TensorFlow 2.2~2.4.
ONNX model converter is tested with ONNX 1.6.0~1.8.0.