ailia SDK Tutorial (Predict API)

The Predict API is the core function of the ailia SDK, an AI inference engine that allows you to perform fast deep learning inference on GPU.

Predict API overview

The Predict API is the fundamental API to perform inference, it can handle not only single-input /single-output models, but also models with multiple inputs and outputs.

Inference for models with single input and output

The C API is defined in library/include/ailia.h, the Python API in python/ailia/__init__.py, and the C# API in unity/Assets/AILIA/Scripts/Models/AiliaModel.cs. AiliaModel.cs

Below is the flow for inferring a single-input, single-output model using the Predict API.

(1) Create an instance with ailiaCreate
(2) Load the model with ailiaOpenStreamFile and ailiaOpenWeightFile
(3) Get the input shape with ailiaGetInputShape
(4) Get the output shape with ailiaGetOutputShape
(5) Pass on a float array of input and output tensors to ailiaPredict
(6) Release the instance with ailiaDestroy

Here is an example of code in C. The API for C is defined in C99, because the ABI is not standardized for C++, and compatibility issues arise when using different compilers. In the future, we are planning to provide the wrapper for C++ in a header file.

Here is an example code in C#, using the AiliaModel class. ailiaCreate, ailiaOpenStreamFile, and ailiaOpenWeightFile are integrated into the call to AiliaModel.Open

Finally, here is the code in Python using the class Ailia.Net. ailiaCreate, ailiaOpenStreamFile and ailiaOpenWeightFile have been merged to the ailia.Net constructor.

Inference for models with multiple inputs and outputs

Below is the flow for running inference on a model with multiple inputs and outputs.

(1) Create an instance with ailiaCreate
(2) Load the model with ailiaOpenStreamFile and ailiaOpenWeightFile
(3) Get the input shape with ailiaFindBlobIndexByName and ailiaGetBlobShape
(4) Get the output shape with ailiaFindBlobIndexByName and ailiaGetBlobShape
(5) Give a float array of input tensors with ailiaSetInputBlobData
(6) Run the inference with ailiaUpdate
(7) Give a float array (output tensors) toailiaGetBlobData and get the inference result
(8) Release the instance with ailiaDestroy

Below is an example written in C.

Below is an example of code in C#. Multiple input/output in C# is supported since ailia SDK 1.2.1.

And finally an example written in Python.

Choose your inference backend

Calling ailiaCreate with the constant AILIA_ENVIRONMENT_ID_AUTO will run the inference on the CPU. In order to use the GPU, change this value by the target GPU env_id returned by ailiaGetEnvironment.

ax Inc. has developed ailia SDK, which enables cross-platform, GPU-based rapid inference.

ax Inc. provides a wide range of services from consulting and model creation, to the development of AI-based applications and SDKs. Feel free to contact us for any inquiry.