RSS Reader

The Beijing Academy of Artificial Intelligence (BAAI) researchers announced this week a natural language processing model called WuDao 2.0 that, per the South China Morning Post , is more advanced than similar models developed by OpenAI and Google.

The report said WuDao 2.0 uses 1.75 trillion parameters to "simulate conversational speech, write poems, understand pictures and even generate recipes." The models developed by OpenAI and Google are supposed to do similar things, but they use fewer parameters to do so, which means

Open Source LLM

POCL 1.7 Released With Better Support For SPIR-V Binaries On CPUs

Phoronix | May 19, 2021, 5:15 a.m. | Original

POCL 1.7 is out as the newest version of this "Portable Computing Language" that aims to effectively allow OpenCL to run well on various CPU architectures as well as other targets like OpenCL over NVIDIA CUDA and AMD HSA.

POCL continues to support OpenCL 1.2 with various OpenCL 2.x features thanks to LLVM's Clang doing much of the heavy lifting for supporting OpenCL across targets. POCL 1.7 as the open-source project's first new feature release since December brings a few key improvements

Open Source LLM

Google發表可談論無限主題的對話AI模型LaMDA

ITHome | May 19, 2021, 2:05 a.m. | Original

Google在開發者大會Google I/O 2021上，發表了最新的對話AI模型 LaMDA（Language Model for Dialogue Applications），官方提到，這是一個非常複雜的語言模型，比當前所有模型，在理解對話上下文方面的能力都還要好。

Google在語言處理耕耘已久，從早期的網頁翻

Open Source LLM

LibreOffice Merges Initial Support For Compiling To WebAssembly

Phoronix | May 6, 2021, 1:52 p.m. | Original

Merged into LibreOffice yesterday is initial support for an EmScripten-based cross-build and compiling to WebAssembly (WASM) for in-browser execution or potentially running on the desktop in a portable manner with the likes of Wasmer .

Merged to LibreOffice were a number of WASM commits yesterday. This documentation outlines the WASM build steps involved.

There is also a simple demo application that was added which ends up being a Mandelbrot with the Qt5 tool-kit.

This port of LibreOffice to WebAssembly using the EmScripten toolchain has been ongoing

Open Source LLM

Microsoft Posts WSLg Preview - GUI App Support With Windows Subsystem For Linux

Phoronix | April 21, 2021, 2:30 p.m. | Original

After announcing last year that they would be supporting GUI applications with Windows Subsystem for Linux , today Microsoft published their "WSLg" preview for this feature to run Linux GUI applications atop Windows 10.

WSLg is the first public preview available of this GUI application support for WSL2. This expands upon Microsoft's long-running work on Windows Subsystem for Linux to now allow GUI applications/workloads to be used easily. This also includes the ability to make use of 3D acceleration within applications.

WSLg is comprised of

Open Source LLM

MLPerf Inference v1.0: 2000 Suite Results, New Power Measurements

AnandTech | April 21, 2021, 9 a.m. | Original

There has been a strong desire for a series of industry standard machine learning benchmarks, akin to the SPEC benchmarks for CPUs, in order to compare relative solutions. Over the past two years, MLCommons, an open engineering consortium, have been discussing and disclosing its MLPerf benchmarks for training and inference, with key consortium members releasing benchmark numbers as the series of tests gets refined. Today we see the full launch of MLPerf Inference v1.0, along with ~2000 results into the database. Alongside this

Open Source LLM

GraalVM 21.1 Released With Experimental Java 16 Support, Many Performance Improvements

Phoronix | April 21, 2021, midnight | Original

Succeeding January's release of GraalVM 21.0 is now GraalVM 21.1 with many performance improvements and other new features.

GraalVM as the Oracle-led Java VM/JDK that also supports other languages/run-times continues to be quite interesting with its ongoing work. With GraalVM 21.1 there is a lot of additions throughout its large code-base, including many performance optimizations.

Some of the GraalVM 21.1 changes that jumped out at us include:

- Experimental support for

Open Source LLM

POCL 1.7-RC1 Up For Testing, Now Exposes OpenCL 3.0

Phoronix | April 19, 2021, 6:14 a.m. | Original

The first release candidate is up for version 1.7 of the Portable Computing Language, the portable OpenCL implementation that can run on CPUs and other accelerators. With POCL 1.7, OpenCL 3.0 is now being exposed and there is also improved support for SPIR-V binaries on CPUs.

POCL 1.7-RC1 was just published and includes the following highlights:

Support for Clang/LLVM 12.0
Improved support for cross-compiling
Improved support for SPIR-V binaries when using CPU device

Open Source LLM

LLVM 12.0 Released With Alder Lake + Sapphire Rapids Support, More C++20

Phoronix | April 15, 2021, 4 a.m. | Original

After the release cycle dragged on an extra month due to blocker bugs, LLVM 12 was officially tagged on Wednesday night as the latest half-year update to this open-source compiler stack.

LLVM 12 is a big feature release with support for x86-64 micro-architecture feature levels (matching the behavior of the GNU/GCC toolchain), adds support for Intel Alder Lake and Sapphire Rapids processors, provides initial support for AMD Zen 3 with "znver3" (though further tuning is

Open Source LLM

Multigenerational LRU Code Updated For Enhancing Linux Kernel Performance

Phoronix | April 13, 2021, 8:44 a.m. | Original

Last month Google engineers proposed multi-generational LRU for Linux to enhance the kernel performance and today the work has advanced to a second version.

Google's Yu Zhao announced the "v2" of the multigenerational LRU framework today for the Linux kernel. The key takeaway from this work remains: " The current page reclaim is too expensive in terms of CPU usage and often making poor choices about what to evict. We would like to offer an alternative framework that is performant, versatile and straightforward. "

As noted in

China Says WuDao 2.0 AI Is an Even Better Conversationalist than OpenAI, Google

POCL 1.7 Released With Better Support For SPIR-V Binaries On CPUs

Google發表可談論無限主題的對話AI模型LaMDA

LibreOffice Merges Initial Support For Compiling To WebAssembly

Microsoft Posts WSLg Preview - GUI App Support With Windows Subsystem For Linux

MLPerf Inference v1.0: 2000 Suite Results, New Power Measurements

GraalVM 21.1 Released With Experimental Java 16 Support, Many Performance Improvements

POCL 1.7-RC1 Up For Testing, Now Exposes OpenCL 3.0

LLVM 12.0 Released With Alder Lake + Sapphire Rapids Support, More C++20

Multigenerational LRU Code Updated For Enhancing Linux Kernel Performance

Menu