-
- News
- Books
Featured Books
- smt007 Magazine
Latest Issues
Current IssueThe Butterfly Effect
The basis of chaos theory is a key concept known as the “butterfly effect.” It’s the idea that a small event in one place creates a cascading set of events elsewhere. So, how is the EMS landscape changing? We’re tracking the concerns and dynamics of this landscape, and there’s a lot to learn.
Coming to Terms With AI
In this issue, we examine the profound effect artificial intelligence and machine learning are having on manufacturing and business processes. We follow technology, innovation, and money as automation becomes the new key indicator of growth in our industry.
Box Build
One trend is to add box build and final assembly to your product offering. In this issue, we explore the opportunities and risks of adding system assembly to your service portfolio.
- Articles
- Columns
Search Console
- Links
- Events
||| MENU - smt007 Magazine
Intel Contributes AI Acceleration to PyTorch 2.0
March 17, 2023 | IntelEstimated reading time: 1 minute
![](https://iconnect007.com/application/files/9616/9087/1209/Intel_AI.jpg)
?In the release of Python 2.0, contributions from Intel using Intel® Extension for PyTorch , oneAPI Deep Neural Network Library (oneDNN) and additional support for Intel® CPUs enable developers to optimize inference and training performance for artificial intelligence (AI).
As part of the PyTorch 2.0 compilation stack, the TorchInductor CPU backend optimization by Intel Extension for PyTorch and PyTorch ATen CPU achieved up to 1.7 times faster FP32 inference performance when benchmarked with TorchBench, HuggingFace and timm.1 This update brings notable performance improvements to graph compilation over the PyTorch eager mode.
Other optimizations include:
- Improved message-passing between adjacent neural network nodes to support graph neural network in PyTorch Geometric (PyG) for enhanced inference and performance training on Intel CPUs.
- New x86 quantization backend – a combination of FBGEMM (Facebook General Matrix-Matrix Multiplication) and oneDNN backends – replaces FBGEMM as the default quantization backend for x86 CPU platforms to enable better end-to-end int8 inference performance.
- Extended use of oneDNN with oneDNN Graph API to maximize efficient code generation on AI hardware by automatically identifying the graph partitions to be accelerated through fusion. BFloat16 and Float32 data types are supported and only inference workloads can be optimized; BF16 is only optimized on machines with AVX512_BF16 ISA support.
Suggested Items
Alphawave Semi Tapes Out Industry-First, Multi-Protocol I/O Connectivity Chiplet for High-Performance Compute and AI Infrastructure
06/19/2024 | BUSINESS WIREAlphawave Semi, a global leader in high-speed connectivity and compute silicon for the world’s technology infrastructure, today announced the successful tape-out of the industry’s first off-the-shelf multi-protocol I/O connectivity chiplet on TSMC’s 7nm process.
Viant’s Next Generation AI Bid Optimizer Raises the Bar
06/18/2024 | BUSINESS WIREViant Technology Inc., a leading advertising technology company, announced the launch of the latest version of its AI Bid Optimizer, a significant upgrade to the AI-driven bidding solution offered within its demand-side platform (DSP).
Skunk Works Demonstrates AI For Air-To-Air Tactical Intercepts
06/11/2024 | Lockheed MartinLockheed Martin Skunk Works partnered with the University of Iowa's Operator Performance Laboratory (OPL) to demonstrate the use of artificial intelligence (AI) in air-to-air intercept scenarios.
Amphenol Honors Mouser Electronics with Best Performance High Service Digital Award for 5th Time
06/11/2024 | BUSINESS WIREMouser Electronics, Inc., the authorized global distributor with the newest semiconductors and electronic components, is pleased to announce that it has been honored with the 2023 Best Performance High Service Digital Award from its valued partner Amphenol Corporation, a global leader in the interconnect industry. With over 700,000 Amphenol parts available to order
Renesas Announces Execution of Facilities Agreement for Altium Acquisition
05/30/2024 | RenesasAs announced in a previous press release, Renesas Electronics Corporation, a premier supplier of advanced semiconductor solutions, reached an agreement with Altium Limited, a global leader in electronics design systems, in which Renesas will acquire all outstanding shares of Altium, making Altium a wholly-owned subsidiary of Renesas, subject to the satisfaction of certain conditions.