Apache MXNet (incubating)

Monday March 04, 2019

Apache MXNet 1.4 is available!

Apache MXNet 1.4 is available!

Today the Apache MXNet community is excited to announce the 1.4.0 release of the Apache MXNet deep learning framework. We would like to thank the Apache MXNet community for all their contributions towards this power packed v1.4 release.

This release brings Java bindings for inference, Julia bindings, experimental control flow operators, JVM memory management, graph optimization and quantizations, and many more features and under-the-hood performance enhancements and usability improvements. Check out the full release highlights on MXNet v1.4 release notes.

mxnet-v-1-4-0.png

Java Inference APIs

This release includes new Java Inference APIs which offer easy to use, idiomatic high level APIs for performing predictions in Java with deep learning models trained using MXNet. This simplifies production deployment of Apache MXNet models for enterprise systems that run on Java. More details are available in this blog post which introduces how to use the new Java Inference APIs.

Julia API

The Julia package brings flexible and efficient GPU computing and state-of-the-art deep learning to Julia. Some highlights of the Julia package include efficient tensor computation across multiple devices including multiple CPUs, GPUs and distributed server nodes.

Control Flow Operators (experimental)

With the control flow operators variable dynamic neural network graphs can be turned into optimized static computation graphs. The optimized graph can greatly improve the speed for training and inference for dynamic models. To learn more details about these operators, check out the Control Flow Operators tutorial.

Automated JVM Memory Management

As the name suggests, the memory management feature provides an automated way for managing native memory when using the JVM language bindings of MXNet. Developers now get a seamless memory management system for managing both CPU and GPU memory footprint without any degradation in performance. More details on how to use can be found in this README file.

MXNet Horovod Integration

Apache MXNet now supports distributed training using Horovod framework. Horovod is an open source distributed framework created at Uber. It leverages efficient inter-GPU communication to distribute and aggregate model parameters across multiple workers thus allowing efficient use of network bandwidth and scaling of training of deep learning models. To learn more about MXNet-Horovod integration, check out this blog.

Subgraph API

Subgraph API empowers Apache MXNet to integrate different kinds of backend libraries such as TVM, MKLDNN, TensorRT, Intel nGraph and many more. Enhanced integration with different backend libraries provides MXNet with a significant performance boost, by optimizing the execution of graph by breaking it into smaller components. Check out integrating with external backend libraries to learn more.

MKLDNN : Graph Optimization and Quantization

MKLDNN takes advantage of the MXNet Subgraph API to implement graph optimizations in the form of operator fusion for inference. Fusions such as Convolution + ReLU and Batch Normalization Folding, provide a great boost in the speeds for inference.
Quantization allows the use of reduced precision (INT8), which reduces memory usage and improves inference time without a significant loss in accuracy. On models such as ResNet50, Inception-BN, MobileNet the observed accuracy loss was less than 0.5 %. More details on this project are available on the MXNet Graph Optimization and Quantization page.

Other Maintenance improvements :

  1. The community fixed 81 bugs, improving MXNet's stability and reliability.
  2. The community also addressed over 55 documentation issues, improving the user experience for using MXNet and earning trust of the users.

Getting started with MXNet

Getting started with MXNet is simple, visit the install page to get started. To learn more about MXNet Gluon interface and deep learning, you can follow our 60-minute crash course, and then later complete this comprehensive set of tutorials, which covers everything from an introduction to deep learning to how to implement cutting-edge neural network models. You can check out other related MXNet tutorials, MXNet blog posts and MXNet YouTube channel.

Have fun with MXNet 1.4.0!

Acknowledgements:

We would like to thank everyone from the Apache MXNet community for their contributions to the 1.4.0 release.

Comments:

As a trainer in educational and teaching centers, he conducts face-to-face and online training on active methodologies, tutorial action, emotional education, and ICT tools. For a few years, he combines teaching and training with the facet of blogger. Specifically, he is the author of the Justify your answer blog, an educational space for teachers.http://thestily.com/

Posted by pagotayat on March 18, 2019 at 02:58 PM UTC #

in the middle of the original facial appearance in description 9, we discover inhabitant editing of satisfied in HDR, sell abroad in HDR for broadcast and sharing on the Internet, which is a reply to the rising utilize of resources in HDR and complement the by now obtainable in the https://crackdj.com/edius-pro-crack/ agenda the opening to labor straight in 4K UHD.

Posted by yamoout on March 27, 2019 at 11:10 AM UTC #

https://cyberspc.com/avira-antivirus-crack/ has gained popularity as a no-nonsense, reliable, antivirus system from Germany. Unlike competitors including the likes of Kaspersky and Norton, Avira is often touted as being free of "bloatware" - meaning that it does what it's meant to without taking up too many system resources in the process.

Posted by Steve Wickham on April 09, 2019 at 01:34 PM UTC #

http://www.crackloverpc.com/tekken-3-game/ PS2 has fewer people than Tekken 3 Game and 5, and is particularly way more 'arcadey', meaning there's less talent needed to acquire during the game than there at any time was, and even more of successful is predicated on luck and various variables.

Posted by Ustin Robinson on April 09, 2019 at 02:53 PM UTC #

Road bikes are built for smooth surfaces like pavement which also enables high speed riding. Hybrid bikes are ideal for both paved roads and mountain trails. Comfort bikes are more for leisure riding or cruising which feature wider tires for a smoother ride. Mountain bikes are built specifically for tough trails with a sturdy frame to withstand bumpy. http://wisypick.com/best-road-bike-under-1000/

Posted by Christina Mog on April 09, 2019 at 05:48 PM UTC #

Simple controls and sub-screens make it easy to enter or search for destinations and get data about your trip. Simply touch the color screen to enter a destination, and nuvi takes you there with turn-by-turn voice directions, 2-D or 3-D maps and smooth map updates as you navigate. https://theoutdoorchamp.com/best-handheld-gps-navigators/

Posted by Radley Beale on April 09, 2019 at 06:33 PM UTC #

This is also a very good post which I really enjoyed reading. It is not every day that I have the possibility to see something like this.. <a href="https://precracks.com/anytrans-crack-license-code/">AnyTrans crack</a>

Posted by AnyTrans Crack on May 01, 2019 at 10:40 AM UTC #

Post a Comment:
  • HTML Syntax: NOT allowed

Calendar

Search

Hot Blogs (today's hits)

Tag Cloud

Categories

Feeds

Links

Navigation