Skip to content

PyTannic

Python bindings for the Tannic framework.

PyTannic is a Python library for interacting with the Tannic-NN framework. It provides utilities for serializing PyTorch models, handling metadata, and sending/receiving tensor data over TCP in a format compatible with Tannic-NN.

Features

  • Serialize torch.nn.Module parameters.
  • Network client for sending/receiving serialized tensors.

Installation

pip install pytannic

Quick Start

Write trained modules to files to load them from the C++ Tannic framework. 

from pytannic.torch.modules import write
from torch.nn import Linear

module = Linear(10, 5)
write(module, "linear_model")

PyTannic allows you to send serialized PyTorch tensors to a Tannic server and receive responses.

tensor = torch.randn(3, 3)

with Client("127.0.0.1", 8080) as client:
    client.send(serialize(tensor)
    data = client.receive()
    result = deserialize(data)

print("Sent tensor:")
print(tensor)
print("\nReceived tensor:")
print(response)

This will allow you to easily create inference servers in C++. For a complete MNIST server example, see the mnist-server repository.

License

License

Tannic is licensed under the Apache License, Version 2.0. See the LICENSE file for details.