OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
AI Developer Tools & SDKs

Dks

Dks is a Python library implementing Deep Kernel Shaping and Tailored Activation Transformations for neural networks. It provides activation-function transformations, weight-initialization functions, and optional Per-Location Normalization for data preprocessing, with framework-specific subpackages for JAX, PyTorch, and TensorFlow.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Dks — OpenViST project profile in AI Developer Tools & SDKs.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use Dks when incorporating DKS or TAT methods into a supported framework's model code and when you can account for the methods' architecture-specific functions and requirements.

Who it is for

  • Machine-learning engineers adapting neural-network model code to use DKS or TAT.
  • Researchers working with JAX, PyTorch, or TensorFlow who need activation transformations or parameter initialization methods.
  • Practitioners who want to apply the package's optional Per-Location Normalization in dataset preprocessing.

Documented uses

  • Transform model activation functions using DKS or TAT procedures.
  • Sample model initial parameters using framework-specific functions.
  • Apply Per-Location Normalization during dataset preprocessing when appropriate.
  • Study a modified ResNet example implementing DKS and TAT.
Before you choose

Requirements & limits

The repository specifies an Apache-2.0 license. The README documents installation through pip from PyPI or GitHub.

  • The library supports JAX, PyTorch, and TensorFlow through separate subpackages; users must import the framework-specific package they intend to use.
  • Activation transformations require a model-specific function, with construction procedures described in the DKS and TAT papers.
  • Users are responsible for satisfying architectural requirements and converting applicable weighted sums into normalized sums.
  • ReLU is only partially supported by DKS and unsupported by TAT; the README recommends Leaky ReLUs.
  • Per-Location Normalization is optional and may be useful in cases such as datasets containing all-zero pixels.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Install Dks with a selected framework extra

Install the package and framework-specific dependencies for a Dks workflow.

Before you start

  • Python package installation environment with pip
  • Choose one or more supported framework extras: jax, pytorch, or tf

Steps

  1. Run the command below, replacing `` with the desired comma-separated framework names.

    Documented command
    pip install dks[]

  2. In model code, import the corresponding subpackage, such as dks.jax or dks.pytorch.

Expected outcome

The Dks package and selected framework dependencies are installed, and the relevant subpackage is available to import in model code.

Guide limitations

  • The package must be integrated and used within the user's own model code.
  • Importing dks alone does not import a framework subpackage.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
12 October 2026
Original website title
google-deepmind/dks · Official homepage title · retained in its source language
How we review sourcesSuggest a correction