deeplearning4j/deeplearning4j
Deeplearning4j
Deeplearning4j is a JVM-based deep learning framework and tensor computing library. It provides a computational graph engine for defining and executing deep learning workflows and mathematical operations within the Java Virtual Machine.
The project includes a dedicated importer for loading and running pretrained models exported from Keras, TensorFlow, and ONNX formats. Its tensor computing capabilities are driven by a modular native C++ math core to execute high-performance linear algebra operations.
The framework covers neural network training, deep learning model inference, and the construction of declarative computational graphs to manage data flow. These tools enable the deployment of machine learning models into enterprise environments.
Features
- JVM Deep Learning Runtimes - Provides a dedicated runtime to execute deep learning workloads within the Java Virtual Machine for enterprise deployment.
- Computational Graph Definitions - Provides a declarative system for defining intermediate operations and variables as a computational graph for execution.
- Computational Graphs - Implements a declarative system to represent mathematical operations as directed graphs for efficient data flow and automatic differentiation.
- Deep Learning Frameworks - Serves as a complete JVM-based framework for training and deploying deep learning models.
- Tensor Computing Libraries - Provides low-level libraries for tensor manipulation and hardware-accelerated mathematical operations.
- Neural Network Training - Implements a comprehensive suite of tools for developing and training deep learning models within the JVM.
- JVM Implementations - Enables the building and training of neural networks directly within the Java Virtual Machine.
- High-Performance Tensor Libraries - Offers high-performance tensor libraries for efficient multidimensional array math and linear algebra with low-level hardware control.
- Compiled Core Wrappers - Employs an architectural pattern that wraps a performance-critical compiled C++ math core with a high-level Java interface.
- Native Library Integrations - Utilizes JNI to link high-level Java code with an optimized C++ library for high-performance mathematical operations.
- Computational Graphs - Provides a framework for defining and executing complex deep learning workflows as directed graphs of data flow.
- Low-Level Tensor Libraries - Executes high-performance tensor operations using a low-level native library wrapped in a high-level API.
- Linear Algebra - Integrates high-performance native linear algebra routines for optimized vector and matrix operations.
- Cross-Format Model Importers - Ships a dedicated importer for loading pretrained models exported from Keras, TensorFlow, and ONNX.
- Pretrained Model Integrations - Includes utilities for loading and integrating pretrained models from TensorFlow, Keras, and PyTorch into the JVM.
- Enterprise AI Infrastructure - Provides a production-grade platform for deploying AI models into scalable enterprise environments using the JVM.
- Model Inference - Enables the execution of trained neural networks within production environments to provide model predictions at scale.
- Cross-Framework Translation - Provides the ability to translate external model definitions from Keras and TensorFlow into an internal JVM-compatible representation.
- AI and Machine Learning - Trains and deploys deep learning models on the JVM.
- Machine Learning - Deep learning library for the JVM.
- Machine-Learning-Frameworks - Scalable deep learning library optimized for industrial GPU usage.
- Machine Learning Libraries - Distributed neural network library for deep learning applications.
