Skip to main content
Ctrl+K

DNIKit 2.0.0 documentation

Getting Started

  • Installation
  • Example Jupyter Notebooks

How to Use

  • 1. DNIKit overview
  • 2. Load a model
  • 3. Load data
  • 4. Introspect

Algorithms

  • Data Introspectors
    • Dataset Report
      • Jupyter Notebook: Dataset Report
    • Familiarity
      • Jupyter Notebook: Familiarity for Rare Data Discovery
      • Jupyter Notebook: Familiarity for Comparison of Dataset Distribution
    • Duplicates
      • Jupyter Notebook: Find Near-Duplicate Data
    • Dimension Reduction
      • Jupyter Notebook: Dimension Reduction Strategies
  • Network Introspectors
    • Network Compression with PFA
      • Jupyter Notebook: Principal Filter Analysis (PFA)
    • Find Inactive Units with IUA
      • Jupyter Notebook: Inactive Unit Analysis (IUA)

Utilities

  • Data Producers
  • Batch Processors

Reference

  • DNIKit API
    • dnikit
      • Base API
      • Processors API
      • Introspectors API
      • Exceptions
      • Typing
    • dnikit_tensorflow
    • dnikit_torch
  • Citing DNIKit
  • Support
  • Contributor’s Guide
  • Changelog
  • .rst

Example Jupyter Notebooks

Contents

  • Data Introspection
  • Network Introspection

Example Jupyter Notebooks#

Each Introspectors page (see the sidebar) has at least one example Jupyter notebook in the “Example” section of the page. For ease of reference, all available notebooks are also listed here.

Data Introspection#

  • Generating a Dataset Report for CIFAR10

  • Finding dataset errors and rare samples with Familiarity

  • Analyzing dataset distribution

  • Finding near duplicate data samples

  • Dimension reduction

Network Introspection#

  • Principal Filter Analysis

  • Inactive Unit Analysis

previous

Installation

next

DNIKit Architecture Overview

Contents
  • Data Introspection
  • Network Introspection

By Apple, Inc.

© Copyright 2023 Apple Inc. All rights reserved..