Back
DDL4J logo

DL4J

Product information, use cases, and access for DL4J.

Developer platformsAI learningDeveloper platformsAI learningSee official site
Official website

Pricing information

Official product verified

No verified public pricing is available yet.

Official sources

VerifiedSeptember 7, 2026

DevPrice organizes public information and does not sell DL4J subscriptions. Prices and availability are determined by DL4J.

What is DL4J?

DL4J, or Eclipse Deeplearning4j, is an open-source distributed deep learning framework for the JVM. It is designed for developers who build, train, evaluate, and deploy neural networks in Java or Scala, with an ecosystem that covers tensor operations, data pipelines, model import, execution, and runtime configuration. This makes it a practical fit for teams that want their deep learning workflow to remain within a JVM-oriented development environment.

The framework exposes APIs for defining networks, running training loops, and performing inference, while SameDiff provides automatic differentiation and graph execution for lower-level computational workflows. ND4J supplies n-dimensional array operations, and DataVec handles data loading, transformation, and ETL before training. Official documentation also covers importing Keras, TensorFlow, and ONNX models, using Spark for distributed training, configuring CPU or CUDA-backed GPU execution, and working with supporting components such as the Model Zoo, Training UI, OmniHub, and Python4J.

Key features of DL4J

  • JVM neural network training and inference

    DL4J provides neural network training and inference in Java and Scala, with documentation covering network fundamentals, training loops, evaluation, and deployment-oriented workflows for JVM applications.

  • Tensor and graph computation tools

    ND4J provides n-dimensional array operations for numerical work, while SameDiff supports automatic differentiation and graph execution, giving developers complementary tools for tensor-based and graph-based model construction.

  • Data pipelines and model import

    DataVec covers data loading, transformation, and ETL operations, while the ecosystem documents import paths for Keras, TensorFlow, and ONNX models so existing model assets can be brought into JVM-based workflows.

  • Hardware and distributed execution

    The documentation covers CPU and CUDA GPU setup, memory and workspace configuration, and distributed training with Apache Spark, helping projects adapt execution to available hardware or larger training workloads.