nillion aivm Download - nillion aivm Source code download

nillion aivm

AI Source Code

1.0.0

Download

Nillion AIVM

AIVM is a cutting-edge framework designed for privacy-preserving inference using advanced cryptographic protocols. With AIVM, you can deploy a local development network (devnet) to explore private inference using provided examples or custom models.

Installing AIVM

Create a virtual environment:
```
python3 -m venv .venv
```
Activate the virtual environment:

On Linux/macOS:
```
source .venv/bin/activate
```
On Windows:
```
.venvScriptsactivate
```
Install the package:

If you are going to execute the examples execute:
```
pip install "nillion-aivm[examples]"
```
Otherwise, if you are going to produce your own code, you can just:
```
pip install nillion-aivm
```

Running AIVM

Start the AIVM devnet:
```
aivm-devnet
```
Open the provided Jupyter notebook examples/getting-started.ipynb to run private inference examples on AIVM.
After completing your tasks, terminate the devnet process by pressing CTRL+C.

Usage

For additional usage, refer to the examples folder, which demonstrates how to set up private inference workflows using AIVM.

Performing Secure Inference

Basic Usage

First, import the AIVM client and check available models:

import aivm_client as aic

# List all supported models
available_models = aic.get_supported_models()
print(available_models)

Prepare your input data. Here's an example using PyTorch to generate a random input:

import torch

# Create a sample input (e.g., for LeNet5 MNIST)
random_input = torch.randn((1, 1, 28, 28))  # Batch size 1, 1 channel, 28x28 pixels

Encrypt your input using the appropriate Cryptensor:

# Encrypt the input
encrypted_input = aic.LeNet5Cryptensor(random_input)

Perform secure inference:

# Get prediction while maintaining privacy
result = aic.get_prediction(encrypted_input, "LeNet5MNIST")

The get_prediction function automatically handles the secure computation protocol with the aivm-devnet nodes, ensuring that your input data remains private throughout the inference process.

Custom Model Upload

You can deploy your own trained models to AIVM, provided they follow the supported architectures (BertTiny or LeNet5).

Uploading Custom Models

Import the AIVM client:

import aivm_client as aic

Upload your custom model:

# For BertTiny models
aic.upload_bert_tiny_model(model_path, "MyCustomBertTiny")

# For LeNet5 models
aic.upload_lenet5_model(model_path, "MyCustomLeNet5")

Perform inference with your custom model:

# For BertTiny models
result = aic.get_prediction(private_berttiny_input, "MyCustomBertTiny")

# For LeNet5 models
result = aic.get_prediction(private_lenet5_input, "MyCustomLeNet5")