Microsoft Boosts AI Efficiency With A ‘Heavy Metal Quartet’ Of Compilers - Decrypt

Ponovno objavil Platon

Spremljevalci: 0

Microsoft je predstavil zbirko štirih novih prevajalnikov umetne inteligence, zasnovanih za optimizacijo delovanja različnih modelov AI. »Heavy metal kvartet« najsodobnejših orodij za zbiranje nosi imena Nabijač, Valjar, Varilec in Brusilnik.

Orodja je razvil Microsoft Research v sodelovanju s številnimi akademskimi ustanovami. Zagotavljajo napredne rešitve za prevajanje — v bistvu pretvorbo iz izvorne kode (človeku berljive) v strojno kodo (šop enic in ničel, zaradi katerih je računalnik izvedljiv) — običajnih modelov umetne inteligence in njihovo učinkovitejše izvajanje na pospeševalnikih strojne opreme, kot so grafični procesorji.

In a Microsoft Research blog post highlighting their capabilities, the company says the compilers build on Microsoft’s extensive research and development in artificial intelligence.

"Prevajalniki umetne inteligence, ki smo jih razvili, so pokazali znatno izboljšanje učinkovitosti prevajanja umetne inteligence in s tem olajšali usposabljanje in uvajanje modelov umetne inteligence," je zapisal Jilong Xue, glavni raziskovalec pri MSR Asia. "V prihodnosti lahko ti obsežni modeli sami po sebi pomagajo pri doseganju optimizacije in kompilacije."

Vsak od štirih novih prevajalnikov se spopada z različnimi izzivi pri optimizaciji delovnih obremenitev umetne inteligence.

Rammer focuses on maximizing hardware parallelism—the capacity of hardware to do different things simoultaneously. This is a key factor in performance, and Rammer minimizes runtime scheduling overhead through improved utilization of parallel resources.

Roller takes a different approach to accelerate compilation, using a fast construction algorithm to find solutions, ultimately generating optimized kernels in seconds rather than hours. In other words, Roller helps create efficient computer programs for AI faster by simplifying the design process.

Varilec reduces expensive memory access traffic by connecting operators in a concentrated pipeline. It unifies memory optimizations into a single framework for greater efficiency.

Končno, Grinder enables control-flow execution on accelerators by integrating it with data flow. This allows optimization across control flow boundaries. Think of it like an expert guiding the steps of an apprentice, telling them what to do to get the job done faster.

As one of the leading technology giants, Microsoft has been at the forefront of AI advancement. The company has partnered closely with AI research firm OpenAI on large language models like GPT-3.5 and GPT-4, which powers ChatGPT and Bing Chat. More recently, Microsoft sodeloval z Meto to integrate LLaMA-2 in its cloud computing solution and introduced a technique called the Algoritem misli to enhance reasoning in models like ChatGPT.

Testiranje je pokazalo, da so prevajalniki bistveno boljši od obstoječih rešitev na merilih uspešnosti. Rammer je na grafičnih procesorjih presegel druge prevajalnike do 20x. Roller je dosegel ali presegel najsodobnejšo zmogljivost, medtem ko je čas prevajanja zmanjšal za velikostne rede. Welder je na grafičnih procesorjih do 21x presegel okvire, kot je PyTorch. Modeli s pospešenim brušenjem z nadzorom pretoka do 8x.

The heavy metal quartet demonstrates Microsoft’s continued leadership in designing breakthrough AI systems —and coming up with fun names for its products. While big partnerships in the AI space like the one with OpenAI grab headlines, the company also actively develops vital software infrastructure to empower AI behind the scenes.

S precejšnjim povečanjem zmogljivosti v primerjavi z obstoječimi rešitvami bi lahko nabijač, valjar, varilec in brusilnik zagotovili ključne konkurenčne prednosti, ko se pojavijo bolj zapletene delovne obremenitve z umetno inteligenco.