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

Deeplift

Deeplift is an open-source Python library for interpreting deep-learning models by calculating feature-contribution and sensitivity scores. It implements DeepLIFT alongside gradients, gradient-times-input, guided backpropagation and integrated gradients, and includes functions to convert saved Keras models into its format.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Deeplift — 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 Deeplift when you need a Python-based attribution workflow around a trained deep-learning model, particularly when the model is saved in a supported Keras format or can be recreated with Deeplift layers.

Who it is for

  • Python developers analyzing predictions from trained deep-learning models.
  • Researchers comparing feature-attribution approaches or inspecting model inputs and intermediate layers.
  • Keras users who want to convert saved models for attribution workflows.

Documented uses

  • Calculate feature-importance scores for a model prediction.
  • Compare DeepLIFT attribution with gradients, guided backpropagation or integrated gradients.
  • Inspect contributions at input pixels or, for some vision tasks, at a higher convolutional layer.
  • Convert a saved Keras model into Deeplift’s model format.
Before you choose

Requirements & limits

The repository is public and identifies the project as open source under the MIT license. The README provides installation through pip.

  • The README reports this version tested with Keras 2.2.4 and TensorFlow 1.14.0; its installation section also refers to earlier TensorFlow and Keras test versions.
  • Keras models have documented autoconversion functions. Other models require recreation with Deeplift layers, except non-Keras models saved in Keras 2.0 format, which the README says can be converted.
  • Attribution depends on selecting a method, target layer and reference; the documentation identifies optimal reference selection as an open problem.
  • For some computer-vision tasks, scoring a higher convolutional layer rather than input pixels may produce better results, according to the README.

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

Convert a saved Keras model for attribution

Convert a saved Keras model into Deeplift’s format as the documented preparation for calculating importance scores.

Before you start

  • Deeplift installed in a Python environment.
  • A saved model in the HDF5 format used by the README example.

Steps

  1. Import deeplift and kerasapi_conversion from deeplift.conversion.

  2. Call convert_model_from_saved_files with the saved model file path.

  3. Choose the documented nonlinear multiplier mode, such as DeepLIFT_GenomicsDefault.

Expected outcome

The saved Keras model is converted into a Deeplift model object for the subsequent scoring workflow.

Guide limitations

  • The repository describes autoconversion for Keras models; models from other libraries must be recreated using Deeplift layers unless saved in Keras 2.0 format.
Documentation 1 ↗
Evidence and transparency

Sources & review

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