Skip to content

Latest commit

 

History

History
48 lines (31 loc) · 3.66 KB

README.md

File metadata and controls

48 lines (31 loc) · 3.66 KB
Raw

AIVM Examples

This directory contains a series of example Jupyter notebooks and pre-trained model files that demonstrate how to use AIVM for privacy-preserving inference, fine-tuning, and custom model deployment. Below is a brief overview of each example:

Notebooks

1. 1-getting-started.ipynb

This notebook is your introduction to AIVM. It walks you through setting up the AIVM devnet and performing a basic private inference task using a pre-trained model (e.g., LeNet5 for MNIST data).

2. 2a-fine-tuning-bert-tiny.ipynb Open In Colab

This example demonstrates how to fine-tune the BertTiny model on a new dataset. It includes steps for loading the pre-trained BertTiny model and fine-tuning it on a different classification task.

3. 2b-fine-tuning-lenet5.ipynb Open In Colab

Similar to 2a, this notebook covers the fine-tuning process for the LeNet5 model, which is designed for image classification tasks. The example uses a sample dataset to show how to fine-tune LeNet5 to new image data.

4. 2c-fine-tuning-bert-tiny-tweet-dataset.ipynb Open In Colab

This notebook extends the fine-tuning process specifically for BertTiny using a tweet sentiment classification dataset. It showcases how to adapt the model for a real-world text-based task (positive, neutral, negative sentiment classification).

5. 3a-upload-your-bert-tiny-model.ipynb

In this notebook, you’ll learn how to upload a custom BertTiny model to AIVM. It covers how to prepare your model, encrypt the input, and upload it to AIVM for secure inference.

6. 3b-upload-your-lenet5-model.ipynb

This example explains how to upload a custom LeNet5 model to AIVM. The notebook provides step-by-step instructions to deploy your trained model for privacy-preserving inference.

7. 3c-upload-your-bert-tiny-for-tweet-sentiment.ipynb

This notebook focuses on deploying a BertTiny model that has been fine-tuned for tweet sentiment classification. It demonstrates how to upload the model and perform secure sentiment analysis on tweet data.

Model Files

1. cats_dogs_lenet5.pth

A pre-trained LeNet5 model designed for the Cats vs. Dogs dataset. This file can be used in the fine-tuning and upload notebooks for LeNet5.

2. imdb_bert_tiny.onnx

A pre-trained BertTiny model for sentiment classification on the IMDB dataset. This file can be used for both fine-tuning and deploying custom models using AIVM.

3. twitter_bert_tiny.onnx

A pre-trained BertTiny model for sentiment classification on Twitter data. This file is specifically used in notebooks related to tweet sentiment classification and can be uploaded to AIVM for inference.

Each notebook is designed to be run sequentially or independently based on your specific use case. Refer to the main README for more information on how to set up and run AIVM.

Project Ideas