Fact-checked by Grok 2 weeks ago

Apache MXNet

Apache MXNet is an open-source framework designed to enable efficient development, training, and deployment of models, particularly deep neural networks, across heterogeneous distributed systems ranging from mobile devices to multi-GPU clusters. It features a front-end that seamlessly blends (similar to ) with via the , allowing for rapid prototyping and optimized performance in production environments. The framework supports distributed training with near-linear scalability, multiple language bindings including , C++, , , , , and , and an ecosystem of libraries for , , and time series forecasting. MXNet originated from the integration of earlier libraries such as CXXNet, , and Purine2, and was formally introduced in a 2015 by Tianqi Chen, Mu Li, Yutian Li, Min Lin, Nairanjana Das, Nadathur Satish, and Zheng Zhang, emphasizing its support for both symbolic expressions and tensor computations with . The project was donated to in December 2016, entering the program, and graduated to become a top-level Apache project on September 21, 2022. Adopted by as its preferred framework in 2016, MXNet was optimized for cloud-scale applications and demonstrated significant speedups, such as up to 109 times faster training on 128 GPUs compared to a single GPU. Following its peak adoption, Apache MXNet saw declining contributions and maintenance, leading to its retirement by the Apache community in September 2023, with the project archived on GitHub on November 17, 2023, and officially moved to the Apache Attic in February 2024. Although no longer actively developed, the framework's codebase, documentation, and historical releases remain accessible for legacy use and study, preserving its contributions to scalable deep learning architectures.

Overview

Description

Apache MXNet is an open-source framework designed for efficient training and deployment of neural networks across various scales, from prototyping to environments. It enables developers to define, train, and deploy deep neural networks on a wide range of devices, including single GPUs and large distributed clusters. A key attribute of MXNet is its hybrid front-end, which allows seamless mixing of symbolic and paradigms to balance flexibility and performance. This design emphasizes efficiency and , making it suitable for both rapid experimentation and high-throughput workloads. MXNet is released under the Apache License 2.0 and supports multiple platforms, including Windows, macOS, and Linux. The latest stable release is version 1.9.1, issued on May 10, 2022. It offers multi-language bindings and distributed training capabilities for broader accessibility and large-scale applications.

Development Status

As of November 2025, Apache MXNet has been officially retired by the Apache Software Foundation, with the project termination approved in September 2023 due to prolonged inactivity, and its codebase moved to the Apache Attic in February 2024. The retirement vote by committers highlighted a lack of significant contributions and community engagement, as code development had effectively halted by late 2022. The decline was influenced by intense competition from more actively maintained deep learning frameworks such as and , which captured greater adoption in research and industry amid the rapid evolution of AI technologies. Additionally, initial backing from , which had integrated MXNet into services like AWS Deep Learning Containers, waned as the company shifted focus to , culminating in the end of MXNet support in those containers starting October 2023. The final major release, version 1.9.1, occurred in May 2022, incorporating bug fixes and performance tweaks, after which community efforts largely dissipated by 2023. Post-retirement, MXNet receives no active development, security updates, or official support, though existing installations continue to function for legacy applications. Users are advised against adopting it for new projects due to potential vulnerabilities and lack of with modern or libraries. For those maintaining MXNet-based workflows, paths to active frameworks like exist, often facilitated by model conversion tools such as MMdnn.

History

Origins and Early Development

Apache MXNet was initiated in 2015 by a team of researchers led by Tianqi Chen from the and Mu Li from (CMU), with advisory contributions from Carlos Guestrin, also at the . This collaboration brought together experts from multiple institutions, including and , to develop a new framework. The project emerged from the Distributed (Deep) Machine Learning Community (DMLC), a group focused on scalable tools. The primary motivations for creating MXNet stemmed from the shortcomings of contemporary frameworks like Theano, which emphasized but struggled with imperative flexibility, and , which offered imperative control yet limited scalability for distributed environments. Developers sought to enable efficient training of large-scale deep neural networks on heterogeneous systems, including multi-GPU setups and cloud clusters, to handle the demands of datasets like comprising millions of samples. This focus addressed the need for frameworks that could scale computations involving billions of operations per training example without sacrificing ease of use for researchers. MXNet originated within the broader context of the GraphLab project, an open-source framework for graph-based initiated by Guestrin at CMU in , which emphasized distributed computation for irregular data structures. As DMLC evolved from GraphLab's foundations, MXNet adapted these principles to support workflows, extending graph computation ideas to training. The prototype saw its first public release in 2015 as an open-source project, providing tools for constructing efficient computation graphs that integrated symbolic expression definition with tensor-based imperative execution. A key early innovation was its lightweight and portable architecture, designed to run seamlessly from research prototypes on laptops to deployments across distributed GPU clusters, minimizing overhead while maximizing on diverse . This dual-programming allowed users to prototype dynamically and optimize statically, bridging gaps in prior systems.

Apache Incubation and Growth

In late 2016, the original developers from academia and industry, along with (AWS), donated MXNet to to foster its growth as an open-source project under the . This move aligned with AWS's commitment to contribute code, documentation, and resources to evolve MXNet into a scalable framework. The project officially entered the Apache Incubator in January 2017, marking the beginning of its formal incubation phase where it underwent rigorous , establishment, and code maturation to meet Apache standards. During incubation, MXNet achieved several key milestones that solidified its stability and appeal. The release of version 1.0 in December 2017 introduced a stable , enabling more reliable development and deployment of models, while incorporating contributions like the new model serving capability from AWS. This version also featured the Gluon , launched as part of the 1.0 milestone, which provided an interface to simplify prototyping and training, enhancing usability for researchers and developers. By early 2018, MXNet integrated seamlessly with AWS SageMaker, allowing users to train and deploy models at scale using managed infrastructure, which accelerated its adoption in cloud-based workflows. The project's growth accelerated through expanding community involvement and strategic partnerships. By 2019, contributions came from a diverse group of developers, including those from AWS, , and other organizations, supporting optimizations for like NVIDIA GPUs and integration with standards such as ONNX for . Partnerships with enabled efficient GPU acceleration, while collaborations with advanced cross-framework compatibility, and Huawei contributed to support in the ONNX ecosystem. After meeting Apache's requirements for active community, inclusive governance, and , MXNet graduated from incubation to become a top-level Apache project in September 2022. From 2018 to 2020, MXNet reached peak adoption in industry applications, particularly for and tasks. The API's ease of use facilitated rapid experimentation, leading to specialized libraries like GluonCV for models and GluonNLP for text processing, which were widely applied in real-world scenarios such as classification and . This period saw MXNet powering production systems at companies leveraging its scalability for distributed training, though later shifts in focus contributed to a gradual decline.

Decline and Retirement

The decline of Apache MXNet began around 2021, marked by a noticeable reduction in community contributions and development activity, as the deep learning ecosystem increasingly shifted toward dominant frameworks like and . This slowdown was exacerbated by Amazon's reduced investment following its initial strong backing, with the company redirecting resources to integration in services like . By late 2022, code development had effectively halted. An effort to develop , initiated in 2020 to modernize the framework and address legacy issues, ultimately failed to gain sufficient community traction. leaving the project struggling to keep pace with rapid advancements in generative and other technologies. Key events underscored the project's fading momentum, including the release of MXNet 1.9.1 in May as the last significant update incorporating bug fixes and performance improvements. Community discussions on intensified in , with a pivotal GitHub in June highlighting the lack of active engagement and proposing options like or . These talks revealed a historical peak of 875 contributors and 51 members, but recent years saw a sharp drop, placing an unsustainable burden on a small group of volunteers amid fierce competition from better-supported alternatives. The retirement timeline unfolded methodically within . An announcement of project inactivity was issued in early 2023, leading to a formal retirement vote by the MXNet committers due to prolonged inactivity. The ASF Board unanimously approved the termination of the MXNet on September 20, 2023, retiring the project effective that month. The transfer to the Apache Attic—a repository for discontinued projects—was completed in February 2024, rendering the repository read-only and archived for historical preservation. Contributing factors to the retirement included intense market competition, where and captured the majority of adoption in and by 2022, leaving MXNet with diminishing relevance. The maintenance burden fell heavily on volunteers without sustained corporate support, as Amazon's pivot away from MXNet diminished the resources needed for ongoing development and updates. Despite these challenges, the project's legacy was preserved through full archival of its code, documentation, and artifacts in the Apache Attic, with encouragement from the for users to the or migrate to active frameworks like GluonTS, a successor library for time-series .

Architecture

Core Components

Apache MXNet's backend engine is implemented in C++ to deliver high-performance computation, serving as the core that handles tensor operations and enables optimizations such as dependency-driven scheduling across heterogeneous devices. This engine processes operations by resolving read/write dependencies, serializing those involving shared variables while allowing parallel execution for independent ones, thereby maximizing resource utilization through multi-threading. At the heart of MXNet's modular design are two key modules: NDArray and . NDArray provides dynamic, multi-dimensional arrays that support imperative-style programming, allowing immediate execution of tensor operations like matrix multiplications directly on CPU or GPU hardware. In contrast, enables the construction of static computation graphs through declarative symbolic expressions, facilitating graph-level optimizations such as operator fusion and auto-differentiation before execution. These modules together underpin MXNet's hybrid approach to computation paradigms, blending imperative flexibility with symbolic efficiency. Data loading in MXNet integrates with data iterators to create efficient input pipelines, employing multi-threaded pre-fetching and augmentation to process and pack examples into compact formats without blocking the main thread. This design ensures seamless flow during model by handling preprocessing tasks asynchronously. The engine's asynchronous execution model further enhances by overlapping and communication, using an internal scheduler to push operations via like PushSync and AsyncFn, which manage non-blocking tasks across and devices. This allows the backend to execute functions only after prerequisites are met, minimizing idle time in pipelines involving tensor manipulations. Memory management relies on a unified allocator that optimizes for both GPU and CPU, incorporating strategies like "inplace" updates—where output tensors input —and "co-share" mechanisms to share among compatible arrays, potentially reducing peak usage by up to four times during execution. By tracking mutations and blocks efficiently, this allocator minimizes overhead and supports scalable workflows on limited .

Computation Model

Apache MXNet employs a computation model that integrates imperative and symbolic programming paradigms through its , enabling developers to mix dynamic execution for flexibility with static optimization for efficiency. The front-end uses HybridBlock and HybridSequential classes to define models that default to imperative style but can be converted to via the hybridize() function. This approach allows seamless transitions, where imperative code—resembling operations—facilitates debugging and rapid prototyping, while symbolic mode compiles the into an optimized for deployment. In the execution flow, MXNet's NDArray module handles imperative computations by executing operations sequentially on tensors, providing Python-like interactivity for tasks such as data manipulation and model building. For symbolic execution, developers define computations using Symbol objects, which construct a directed acyclic graph (DAG) representing the neural network; this graph is then compiled into an executable form by the backend executor. During compilation, MXNet applies optimizations such as operator fusion—merging multiple small operations (e.g., element-wise addition and multiplication) into a single kernel to reduce overhead—and graph-level rewrites to eliminate redundant computations, improving runtime performance by up to 20-30% in typical benchmarks. A key component in MXNet's computation model is the KVStore, a key-value that facilitates parameter synchronization during by allowing devices to push updates (e.g., gradients) and pull synchronized values across the model. This mechanism integrates with the execution to ensure consistent parameter states without delving into distributed specifics here. The trade-offs in MXNet's model balance research and production needs: imperative execution offers high flexibility for experimentation and but incurs higher computational costs due to immediate , whereas symbolic mode prioritizes speed and portability through pre-optimized graphs, making it suitable for large-scale .

Features

Scalability and Distributed Training

Apache MXNet employs a for distributed , utilizing the KVStore (key-value ) to manage parameter across multiple devices and machines. The KVStore supports both synchronous and asynchronous update modes: in synchronous mode (dist_sync), workers compute gradients, push them to servers for aggregation, and pull updated parameters before proceeding to the next iteration, ensuring consistency; in asynchronous mode (dist_async), updates occur independently, allowing faster but potentially less stable . This design facilitates efficient communication and in heterogeneous environments. For multi-GPU training, MXNet provides built-in support for , where the model is replicated across GPUs and data batches are split for parallel computation, with gradients aggregated via KVStore; model parallelism is also available, partitioning the model layers across GPUs for handling large models that exceed single-GPU limits. These mechanisms leverage MXNet's computation graph model, which enables seamless distribution of operators. Integration with Horovod allows MPI-based distributed training, enabling all-reduce operations for gradient synchronization and scaling across clusters, often achieving better performance than the native parameter server for certain workloads. MXNet has demonstrated to hundreds of GPUs in production settings, with Horovod extending support for larger clusters. Benchmarks on image classification tasks, such as ResNet-50 training, show near-linear speedup with increasing GPU count; a TuSimple benchmark found MXNet faster, more memory-efficient, and more accurate than TensorFlow with eight GPUs. To address challenges in dynamic environments, MXNet incorporates fault tolerance through parameter server replication and worker redundancy, allowing recovery from node failures without restarting training. Elastic training capabilities further support varying cluster sizes by dynamically adding or removing workers during sessions, with minimal impact on convergence accuracy, as validated in cloud-based experiments.

Flexibility in Programming Paradigms

Apache MXNet provides flexibility in programming paradigms through its API, which supports both and symbolic approaches to model development. in MXNet, facilitated by , allows users to define and execute computations dynamically, similar to operations on NDArrays, enabling the creation of dynamic computation graphs that are easy to debug and iterate upon during development. This define-by-run style executes code statement by statement, making it intuitive for rapid prototyping of complex models. In contrast, symbolic programming in MXNet employs a define-and-run , where the computation graph is first defined and then compiled for execution, optimizing performance through and portability across devices. integrates in both imperative and symbolic modes, allowing gradients to be computed seamlessly regardless of the chosen style. The further enhances usability with high-level building blocks, such as HybridSequential and nn.Dense layers, which enable modular network construction akin to , streamlining the assembly of neural architectures. MXNet's hybrid programming capability allows seamless switching between paradigms within the same model, particularly useful for custom layers via the HybridBlock class. For instance, a can implement a in imperative mode for flexibility during training and hybridize specific components to symbolic mode for acceleration, as shown in the hybrid_forward method that dispatches operations based on the execution context. This approach yields benefits such as faster prototyping in imperative mode for experimentation and optimized in symbolic mode, which can reduce computation time significantly— for example, hybridizing a simple network can improve performance in repeated executions by compiling the graph once. Overall, these paradigms empower users to balance ease of use with efficiency tailored to different stages of the workflow.

Multi-Language Support

Apache MXNet provides bindings for multiple programming languages, enabling developers to access its core functionality across diverse ecosystems. The supported languages include , , , , , C++, , and . These bindings allow users to define, train, and deploy deep learning models while leveraging language-specific strengths, such as 's extensive ecosystem for rapid prototyping and research. serves as the primary interface, featuring deep integration with the high-level , which simplifies model development through imperative and symbolic programming paradigms. bindings facilitate seamless integration with the (JVM), making it suitable for enterprise applications and production environments that require interoperability with existing Java-based systems. bindings emphasize high-performance numerical computing, bringing efficient GPU acceleration and capabilities to scientific workflows. Other languages like offer perspectives on the JVM, provides robust object-oriented support for inference, C++ enables low-level optimizations, caters to statistical modeling, and supports scripting tasks. At the core, MXNet employs a unified C++ backend for computation, with language bindings implemented via a C API that acts as a (FFI) to ensure consistency and performance across interfaces. This design allows the same computational graph and operators to be executed uniformly, minimizing discrepancies in behavior or efficiency between languages. Adoption has been highest for , particularly among researchers and for initial model training due to its accessibility and rich tooling. Scala has seen notable use in production settings, especially for deploying models in JVM-integrated services. Bindings for other languages, such as for web-based applications and for statistical analysis, have found niche applications but lower overall usage. Following the project's retirement in and entry into the Apache Attic, MXNet's language bindings are no longer actively maintained, leading to partial operator support in some interfaces, particularly those reliant on deprecated C APIs planned for revision in the unreleased version 2.0. For instance, bindings for , , , , and were removed in the master branch of version 2.x due to these deprecations, though version 1.x remains functional for legacy use.

Portability and Deployment

Apache MXNet demonstrates strong portability across diverse hardware platforms, supporting execution on CPUs through standard installations that enable efficient computation without specialized accelerators. For GPU acceleration, MXNet natively integrates with , allowing models to leverage multiple GPUs for and , provided is properly installed on systems with compatible hardware. Additionally, through its integration with the Apache TVM compiler stack, MXNet extends support to GPUs via , enabling compilation and deployment of MXNet models on hardware by generating optimized code for backends. This TVM integration also facilitates optimization and deployment on edge devices, such as embedded systems and hardware, by compiling models into lightweight, hardware-specific code that targets various accelerators including FPGAs and specialized processors. A key aspect of MXNet's portability lies in its lightweight runtime, designed for resource-constrained environments like mobile and devices. The framework's NNVM compiler, evolved into TVM, allows models trained in MXNet to be exported and optimized for platforms such as and , producing compact executables that run with minimal overhead. This enables seamless portability from development on high-end servers to deployment on edge hardware, where and power efficiency are critical, without requiring framework-specific modifications. For production deployment, MXNet provides dedicated tools to facilitate scalable and interoperable model serving. The MXNet Model Server offers a robust platform for hosting trained models, supporting high-throughput predictions via RESTful APIs and integrating with monitoring services like Amazon CloudWatch for operational metrics. Complementing this, MXNet supports export to the ONNX format starting from version 1.3, allowing models to be converted into a standardized representation for use across different frameworks and runtimes, with features like dynamic input shapes for flexible inference. To further enhance deployment in low-resource settings, MXNet incorporates optimization techniques such as quantization, which reduces of weights and activations to lower bit depths, and , which eliminates redundant connections, thereby shrinking model size and accelerating inference on constrained devices. Practical examples illustrate MXNet's deployment versatility; for instance, models can be embedded directly into functions for serverless inference, leveraging MXNet's lightweight nature to handle scalable predictions without managing infrastructure. Similarly, optimized MXNet models deploy effectively on embedded systems in scenarios, such as invoking predictions via AWS IoT services on resource-limited , combining quantization and to fit within tight constraints while maintaining accuracy.

Ecosystem and Integrations

Libraries and Tools

Apache MXNet's ecosystem included specialized libraries built on its core framework to facilitate tasks in various domains. These libraries leveraged MXNet's interface for imperative and symbolic programming, providing pre-built components, models, and utilities that streamlined development. was a dedicated toolkit for applications, offering implementations of state-of-the-art algorithms such as , semantic segmentation, and pose estimation. It included a rich model zoo with pre-trained weights for models like , Mask R-CNN, and ResNet variants, along with pipelines and evaluation metrics to accelerate prototyping and . For instance, developers could build a vision model using GluonCV by loading a pre-trained classifier, applying transformations to input images, and performing inference in just a few lines of , as demonstrated in its documentation examples for image classification. It now supports in addition to MXNet and remains actively maintained, recommending AutoGluon for image classification and tasks. GluonNLP extended MXNet's capabilities to , supplying tools like tokenizers, word embeddings (e.g., and ), and pre-trained models for tasks including , , and . It supported modular components for building pipelines, such as sequence encoders and attention mechanisms, enabling efficient experimentation with transformer-based architectures. The project was archived in January 2024 and is no longer actively maintained. For time series analysis, GluonTS provided a probabilistic modeling focused on methods for and . It included models like DeepAR and architectures, along with datasets and evaluators for handling multivariate , allowing users to train global models across multiple series for scalable predictions. Following MXNet's retirement, GluonTS primarily uses and continues development independently under AWS Labs. The MXNet contrib module served as a repository for experimental operators and utilities, offering advanced features like custom NDArray operations and integration tools that were not yet part of the stable core. This enabled developers to extend MXNet with cutting-edge or domain-specific functionality during research phases. Additionally, the ecosystem featured model zoos across libraries, hosting downloadable pre-trained weights for transfer learning, and contrib modules for seamless custom extensions. Development was supported by tools such as the MXNet profiler, which captured execution traces to analyze , memory usage, and operator bottlenecks, aiding in optimization. The autograd engine, integral to , automated gradient computation for , simplifying the differentiation process in dynamic neural networks. These utilities integrated with MXNet's core components for efficient debugging and training workflows. Following MXNet's retirement in September 2023 and its move to the Apache Attic in February 2024, the libraries and tools are legacy components, with varying levels of independent maintenance as noted above.

Cloud and Hardware Support

Apache MXNet offered native support for training and hosting models within AWS SageMaker, enabling users to build, train, and deploy models using MXNet's containers and estimators directly in the platform. It was also compatible with through custom estimators and scripts, allowing integration for distributed training workflows. Additionally, MXNet integrated with Google Cloud AI Platform, particularly via Deep Learning VMs that pre-installed MXNet for GPU-accelerated compute instances. On the hardware side, MXNet was optimized for GPUs through integration with the cuDNN library, which accelerated operations and improved training efficiency on CUDA-enabled devices. It provided support for GPUs via , facilitating GPU acceleration for deep learning tasks on compatible hardware. For CPU-based computations, MXNet incorporated optimizations, including oneDNN for enhanced performance and Intel DL Boost for vector instructions on processors. Cloud deployment options for MXNet included serverless inference on , where models could be packaged and invoked for low-latency predictions without managing servers. It also supported orchestration for distributed inference, leveraging Amazon Elastic Kubernetes Service (EKS) to scale MXNet models across clusters for high-throughput applications. Specific tools like AWS Deep Learning AMIs came pre-configured with MXNet, streamlining setup on EC2 instances for rapid prototyping and production. Similarly, 's MXNet estimator facilitated job submission and resource allocation within Azure ML workspaces. Following MXNet's retirement in September 2023 and its move to the Attic in February 2024, these cloud and hardware integrations remain functional for existing deployments but are no longer actively maintained by the community.

References

  1. [1]
    MXNet: A Flexible and Efficient Machine Learning Library for ... - arXiv
    Dec 3, 2015 · This paper describes both the API design and the system implementation of MXNet, and explains how embedding of both symbolic expression and ...Missing: original | Show results with:original
  2. [2]
    What Is Apache MXNet and Why Does It Matter? | NVIDIA Glossary
    MXNet is an open-source deep learning framework that allows you to define, train, and deploy deep neural networks on a wide array of devices.
  3. [3]
    Apache MXNet | A flexible and efficient library for deep learning.
    A truly open source deep learning framework suited for flexible research prototyping and production.PyTorch vs Apache MXNetMXNet | The Apache AtticGet StartedFeaturesPython API
  4. [4]
    Board Meeting Minutes - MXNet - Apache Whimsy
    ... graduation Date of last release: 2019-03-04 MXNet 1.4.0 When were the ... Date of last release: A maintenance release Apache MXNet-incubating 0.11.0 ...
  5. [5]
    What Is Apache MXNet? - AWS Documentation
    Apache MXNet (MXNet) is an open source deep learning framework that allows you to define, train, and deploy deep neural networks on a wide array of platforms.
  6. [6]
    MXNet - The Apache Attic
    MXNet retired in September 2023 and the move to the Attic was completed in February 2024. MXNet was a flexible and efficient library for Deep Learning.
  7. [7]
    GitHub - apache/mxnet
    Nov 17, 2023 · Apache MXNet is a deep learning framework designed for both efficiency and flexibility. It allows you to mix symbolic and imperative programming.
  8. [8]
    Features | Apache MXNet
    A flexible library to quickly develop cutting-edge deep learning research or a robust framework to push production workload, Apache MXNet caters to all needs.<|control11|><|separator|>
  9. [9]
    mxnet - PyPI
    1. pip install mxnet. Copy PIP instructions. Latest version. Released: May 17, 2022. Apache MXNet is an ultra-scalable deep learning framework. This version ...
  10. [10]
    board_minutes_2023_09_20.txt - The Apache Software Foundation
    Sep 20, 2023 · Committee reports approved as submitted by General Consent. 7. Special Orders A. Terminate the Apache MXNet Project WHEREAS, the Board of ...
  11. [11]
    Apache MXNet is now retired-Apache Mail Archives
    The Apache MXNet committers have voted to retire the project due to inactivity. MXNet was a flexible and efficient library for Deep Learning.Missing: Software September 2023
  12. [12]
    [RFC] Future of Apache MXNet · Issue #21206 - GitHub
    Jun 20, 2023 · Apache MXNet is an open-source deep learning framework used to train and deploy deep learning models developed by contributors from multiple organizations.
  13. [13]
    Framework Support Policy Table - AWS Deep Learning Containers
    2023-10-01 End of support for MXNet and Elastic Inference. Unsupported Frameworks. From 1st Oct 2023, MXNet, Elastic Inference frameworks are not supported ...
  14. [14]
    Convert Full ImageNet Pre-trained Model from MXNet to PyTorch
    In this tutorial, I want to convert the Full ImageNet pre-trained model from MXNet to PyTorch via MMdnn convertor.
  15. [15]
    [PDF] MXNet: Flexible and Efficient Library for Deep Learning
    Simon Fraser University ... Carlos Guestrin. U Washington. Major Developers. Advisors. Hardware and software supports. Page 80. Go mxnet.dmlc.ml to Get Started.Missing: 2015 | Show results with:2015
  16. [16]
    Major Open-Source Projects - Carlos Guestrin
    MXNet: lightweight, portable, flexible distributed/mobile deep learning library (Apache incubated) ... GraphLab and PowerGraph: framework for large-scale ...Missing: origins founders
  17. [17]
    MXNet - Deep Learning Framework of Choice at AWS
    Nov 22, 2016 · MXNet will be the deep learning framework of choice at AWS. AWS will contribute code and improved documentation as well as invest in the ...<|separator|>
  18. [18]
    board_minutes_2019_05_15.txt - The Apache Software Foundation
    May 15, 2019 · The project entered the incubator in October 2016 and in April 2019 graduated to become a top level Apache project. Three releases have been ...
  19. [19]
    Milestone 1.0.0 Release for Apache MXNet - Blogs Archive
    This release also includes a full-featured and performance optimized version of the Gluon ... Milestone 1.0.0 Release for Apache MXNet. Dec 5, 2017.
  20. [20]
    AWS Contributes to Milestone 1.0 Release of Apache MXNet ...
    Dec 4, 2017 · Today AWS announced contributions to the milestone 1.0 release of the Apache MXNet deep learning engine and the introduction of a new model ...
  21. [21]
    Document History for Amazon SageMaker AI
    TensorFlow 1.5 and MXNet 1.0 support. Amazon SageMaker AI Deep Learning containers now support TensorFlow 1.5 and Apache MXNet 1.0. February 27, 2018.
  22. [22]
    ONNX V1 released - Engineering at Meta - Facebook
    Dec 8, 2017 · ... Microsoft's Cognitive Toolkit, Apache MXNet, PyTorch and NVIDIA's TensorRT. We also have community contributed converters for other projects ...
  23. [23]
    Deep Learning in Computer Vision and Natural Language Processing
    We present GluonCV and GluonNLP, the deep learning toolkits for computer vision and natural language processing based on Apache MXNet (incubating).Missing: NLP | Show results with:NLP
  24. [24]
    Why Did Apache MXNet Shut Down? The Rise and Fall of an AI Giant
    MXNet failed due to a combination of declining community contributions, outdated builds, and Amazon shifting its focus to PyTorch.Lessons Learned From Apache... · Faqs · How Much Does Apache.Org...<|control11|><|separator|>
  25. [25]
    [RFC] Future of Apache MXNet-Apache Mail Archives
    Dear MXNet community, I would like to start a conversation about the future of Apache MXNet and where we should head next. # What we built Apache MXNet is ...<|control11|><|separator|>
  26. [26]
    MXNet: A Growing Deep Learning Framework
    Nov 14, 2022 · MXNet (pronounced mix-net) is Apache's open-source spin on a deep-learning framework that supports building and training models in multiple ...
  27. [27]
    MXNet System Architecture
    NDArray: Dynamic, asynchronous n-dimensional arrays, which provide flexible imperative programs for MXNet. Symbolic Execution: Static symbolic graph executor, ...Missing: backend C++
  28. [28]
    Hybridize — Apache MXNet documentation
    ### Summary of MXNet's Hybrid Front-End and Gluon API
  29. [29]
    Deep Learning Programming Paradigm | Apache MXNet
    A flexible and efficient library for deep learning. Apache MXNet is an effort undergoing incubation at The Apache Software Foundation (ASF), sponsored by the ...Architecture · Symbolic Vs. Imperative... · Case Study: Backprop And...<|separator|>
  30. [30]
    mxnet.kvstore — Apache MXNet documentation
    kvstore. Key value store interface of MXNet for parameter synchronization. A key-value store for synchronization of values, over multiple devices.
  31. [31]
    Distributed Training in MXNet
    dist_sync : In synchronous distributed training, all workers use the same synchronized set of model parameters at the start of every batch.
  32. [32]
    Data Parallelism with Multiple CPU/GPUs on MXNet
    Run MXNet on Multiple CPU/GPUs with Data Parallelism. MXNet supports training with multiple CPUs and GPUs, which may be located on different physical ...
  33. [33]
    Model Parallel | Apache MXNet
    Training with Multiple GPUs Using Model Parallelism. Training deep learning models can be resource intensive. Even with a powerful GPU, some models can take ...
  34. [34]
    Horovod with MXNet
    Horovod supports Apache MXNet and regular TensorFlow in similar ways. ... Distributed training framework for TensorFlow, Keras, PyTorch, and Apache MXNet.Missing: integration | Show results with:integration
  35. [35]
    How to run distributed training using Horovod and MXNet on AWS ...
    Sep 1, 2020 · To learn more about the performance comparison between Horovod and Parameter Server, this blog post illustrates the difference as ResNet50 ...Training An Mxnet Model With... · Training A Mxnet Model With... · Horovod With Mxnet Training...
  36. [36]
    Extend MXNet Distributed Training by MPI AllReduce
    Mar 26, 2018 · MXNet inherent distributed training mechanism, parameter server, provides efficient communication in ASGD and fault tolerance especially in ...
  37. [37]
  38. [38]
    awslabs/dynamic-training-with-apache-mxnet-on-aws - GitHub
    Jul 16, 2024 · Dynamic training with Apache MXNet reduces cost and time for training deep neural networks by leveraging AWS cloud elasticity and scale.Missing: fault tolerance
  39. [39]
    Debugging and performance optimization tips - Apache MXNet
    The difference between imperative style (Gluon non-hybridized) and symbolic style (Gluon hybridized) is: imperative style is define-by-run; symbolic style is ...
  40. [40]
  41. [41]
    Introducing Gluon: a new library for machine learning from AWS and ...
    Oct 12, 2017 · Gluon provides a clear, concise API for defining machine learning models using a collection of pre-built, optimized neural network components.<|control11|><|separator|>
  42. [42]
    Julia Guide | Apache MXNet
    MXNet supports the Julia programming language. The MXNet Julia package brings flexible and efficient GPU computing and the state-of-art deep learning to Julia.Missing: high- performance numerics
  43. [43]
    MXNet FFI for Operator Imperative Invocation
    Jan 24, 2020 · MXNet FFI passes its arguments mainly via ctypes. All keyword arguments are serialized to string at frontend and deserialized from string at ...
  44. [44]
    Running Java-based deep learning with MXNet and Amazon Elastic ...
    Apr 29, 2019 · The Java/Scala support for MXNet on Amazon EI enables Java applications to add cost-effective deep learning acceleration to existing production ...Missing: integration | Show results with:integration
  45. [45]
    Docs | Apache MXNet
    Call for Contribution. The Clojure, Java, Julia, R, and Scala language bindings of MXNet v1. x were removed in v2. x due to some C APIs being deprecated and ...Python Api · Python-First Api · Apache Mxnet Architecture
  46. [46]
    Windows Setup - Apache MXNet
    Recommended System Requirements · Windows 10, Server 2012 R2, or Server 2016 · Visual Studio 2017 (any type) · At least one NVIDIA CUDA-enabled GPU · MKL-enabled ...Installing Mxnet On Windows · Install Mxnet With Python · Build From Source
  47. [47]
    Use GPUs — Apache MXNet documentation
    First, make sure you have at least one Nvidia GPU in your machine and CUDA properly installed. Other GPUs such as AMD and Intel GPUs are not supported yet.Missing: hardware ROCm
  48. [48]
    Bringing AMDGPUs to TVM Stack and NNVM Compiler with ROCm
    Oct 30, 2017 · We have put together working examples of compiling models from MXNet and PyTorch with NNVM, and running them on AMD GPUs with ROCm backend. More ...
  49. [49]
  50. [50]
    NNVM Compiler: Open Compiler for AI Frameworks - Apache TVM
    Oct 6, 2017 · The NNVM compiler can directly take models from deep learning frameworks such as Apache MXNet. It also support model exchange formats such as ...Missing: portability | Show results with:portability
  51. [51]
    Model Server for Apache MXNet introduces ONNX support and ...
    Feb 5, 2018 · Now you can serve models in Open Neural Network Exchange (ONNX) format and publish operational metrics directly to Amazon CloudWatch, where you can create ...
  52. [52]
    Exporting to ONNX format - Apache MXNet
    Exports the MXNet model file, passed as a parameter, into ONNX model. Accepts both symbol,parameter objects as well as json and params filepaths as input.Missing: Server | Show results with:Server
  53. [53]
    [PDF] Enabling Deep Learning in IoT Applications with Apache MXNet
    Host it in S3 (or embed it in a Lambda function). • Write a Lambda function performing prediction. • Invoke it through AWS IoT. Best when. Devices ...
  54. [54]
    Seamlessly Scale Predictions with AWS Lambda and MXNet
    Jan 23, 2017 · In this post, we show how to use MXNet and AWS Lambda to deploy models at scale for predictions. What is MXNet? MXNet is a full-featured ...Missing: embedded | Show results with:embedded
  55. [55]
    GluonCV Toolkit
    GluonCV provides implementations of state-of-the-art (SOTA) deep learning algorithms in computer vision. It aims to help engineers, researchers, and students.
  56. [56]
    Model Zoo — gluoncv 0.11.0 documentation
    The Model Zoo includes models for Classification, Detection, Segmentation, Pose Estimation, Action Recognition, and Depth Prediction.
  57. [57]
    GluonNLP: NLP made easy — gluonnlp 0.10.0 documentation
    GluonNLP provides SOTA deep learning models for NLP, designed for engineers, researchers, and students to prototype research ideas and products.
  58. [58]
    GluonTS documentation
    GluonTS is a Python package for probabilistic time series modeling, focusing on deep learning based models, based on PyTorch and MXNet.Gluonts.time_feature... · Gluonts.time_feature.holiday... · Gluonts.dataset...
  59. [59]
    mxnet.contrib — Apache MXNet documentation
    This project has retired. For details please refer to its Attic page.. Python APInavigate_next mxnet.contrib.
  60. [60]
    Profiling MXNet Models
    In this tutorial, we will learn how to profile MXNet models to measure their running time and memory consumption using the MXNet profiler.
  61. [61]
    Automatic Differentiation — Apache MXNet documentation
    MXNet Gluon uses Reverse Mode Automatic Differentiation (autograd) to backprogate gradients from the loss metric to the network parameters.
  62. [62]
    Resources for using Apache MXNet with Amazon SageMaker AI
    The following section provides reference material you can use to learn how to use SageMaker AI to train and deploy a model using custom MXNet code.
  63. [63]
    GPUs are now available for Google Compute Engine and Cloud ...
    Feb 22, 2017 · These instances support popular machine learning and deep learning frameworks such as TensorFlow, Theano, Torch, MXNet and Caffe, as well as ...
  64. [64]
    GPU-Accelerated MXNet - NVIDIA
    MXNet is an open-source deep learning framework that allows you to define, train, and deploy deep neural networks on a wide array of devices.
  65. [65]
    Deep learning frameworks for ROCm
    The AMD ROCm ... It includes details on ROCm compatibility and third-party tool support, installation steps and options, and links to GitHub resources.Missing: Apache MXNet<|separator|>
  66. [66]
    Optimizing inference on CPU in the Upcoming Apache MXNet 2.0
    Feb 22, 2021 · Operator fusion allows to chain operations together to speed up memory-bound operations by reducing memory IO operations. Quantization speeds up ...
  67. [67]
    Apache* MXNet* (Incubating) Gets a Lift with Intel® DL Boost
    Dec 20, 2019 · Both inference throughput and latency performance are significantly improved by leveraging the operator fusion and model quantization on Apache ...
  68. [68]
    aws-samples/deploy-apache-mxnet-on-aws-lambda - GitHub
    This is a example of how to deploy your Apache MXNet model on Lambda, which you can benefit from Lambda pricing model.Deploying Apache Mxnet On... · Use Apache Mxnet For... · Aws Lambda Function SettingsMissing: embedded | Show results with:embedded
  69. [69]
    Optimizing TensorFlow model serving with Kubernetes and Amazon ...
    Sep 6, 2019 · This post offers a dive deep into how to use Amazon Elastic Inference with Amazon Elastic Kubernetes Service.Elastic Inference Overview · Integrating Elastic... · Inference Node Container...