It’s time to move beyond Artificial Intelligence frameworks. Recently, a joined effort from the Digital giants Microsoft and Facebook has paved the pathway for developers to move beyond traditional AI frameworks. The Open Neural Network Exchange (ONNX) format announced the other day that Facebook and Microsoft are on a lookout to boost AI interoperability and innovation. This piece of information was published in their own blog posts, and from there it got viral.
In Facebook’s blog post, the Social Media behemoth clearly defined its new effort is “toward an open ecosystem where AI developers can easily move between state-of-the-art tools and choose the combination that is best for them.”
— DexLab Analytics (@Dexlabanalytics) August 7, 2017
Now, delving into the technicalities, the process of developing AI tools and technologies is not a piece of cake. The tech gurus, pundits, engineers and researchers need to focus on a specific framework constructed on a unique set of capabilities – all this was stated in the Facebook post. In the end, all of this means that the developers get stuck into a single framework throughout the process, which at times can get really frustrating..
The Facebook post also added, a big issue pops up when the research and development phase asks for different features and capabilities instead of shipping to production. It leads to a set of inefficiencies, like translating models manually to fit into different frameworks as and when required.
Skills required during Interviews for a Data Scientist @ Facebook, Intel, Ebay. Square etc. – @Dexlabanalytics.
So, here comes the ONNX – it’s expected to bridge the gap between frameworks and let the research maestros select whichever framework suits their project type at various stages across the way of work. This month, ONNX support is said to be coming to Caffe2, PyTorch, and Cognitive Toolkit, indicating “models trained in one of these frameworks to be exported to another for inference.”
Another noteworthy feature is called shared optimization – it helps in advance AI development. Frequently, vendors and third-parties feel the urge to integrate optimizations for particular AI frameworks individually, and with ONNX they could specifically target multiple frameworks all at the same time – Microsoft noted this.
Facebook has for long maintained the difference between FAIR and AML Machine Learning groups, and for good. While Facebook AI Research (FAIR) deals with bleeding edge research, Applied Machine Learning brings forth intelligence into products. The former is accustomed to tally with PyTorch – an effective deep learning framework optimized to produce cutting edge results, irrespective of resource constraints.
The Open Neural Network Exchange (ONNX) is best described as a standard to allow developers shift their neural networks from one framework to another, provided both the frameworks comes under ONNX standard. It was released as an Open Source Project, meaning the format can go through additional changes, developments and advancements from the expansive open source community.
“Enabling interoperability between different frameworks and streamlining the path from research to production will help increase the speed of innovation in the AI community,” quoted the Facebook post.
In a nutshell,
- AI developers can switch between AI frameworks.
- This September, Caffe2, PyTorch and Cognitive Toolkit will be empowered by ONNX support.
- ONNX gives way to shared optimizations – ensuring better flexibility and ease of work.
For more such interesting blogs and updates, follow us on DexLab Analytics. We are a thriving Big Data Hadoop institute in Pune waiting to enamor you with our expansive, in-detailed Hadoop training courses in Pune.
Interested in a career in Data Analyst?
To learn more about Machine Learning Using Python and Spark – click here.
To learn more about Data Analyst with Advanced excel course – click here.
To learn more about Data Analyst with SAS Course – click here.
To learn more about Data Analyst with R Course – click here.
To learn more about Big Data Course – click here.