2023 年的秋天，当全世界都在为 ChatGPT 和大语言模型疯狂的时候，远在澳大利亚悉尼的一对兄弟却在为一个看似简单的问题发愁：为什么微调一个开源模型要花这么长时间，还要用那么昂贵的 GPU？

Unsloth，一个由Daniel Han和Michael Han兄弟开发的开源项目，正在悄然改变着 AI 训练的格局。 2023年，当 ChatGPT 和大型语言模型（ LLM ）风靡全球时，这对来自澳大利亚的兄弟却专注于解决一个核心问题：如何加速 开源模型 的微调，降低对昂贵 GPU 的依赖。他们开发的Unsloth，通过一系列底层优化，实现了 AI 训练速度的显著提升，并为 LLM ...

I would like to serve the gpt-oss-20B model using Triton Inference Server on a setup with 4× Tesla T4 GPUs.

Tesla, Inc. has shut down its Dojo supercomputer, a project once billed as essential for enabling full self-driving, reflecting both technical challenges and a costly failure to deliver on autonomy ...

For years, Elon Musk has spoken of the promise of Dojo, the AI supercomputer that was supposed to be the cornerstone of Tesla’s AI ambitions. It was important enough to Musk that in July 2024, he said ...

I'm encountering a persistent runtime error, CUDA_ERROR_NOT_FOUND, when trying to run the stt-rs example on a Tesla T4 GPU. The program compiles successfully, but the ...

There’s no doubt that artificial intelligence (AI) is taking over the financial services industry more than most others. In fact, Gartner notes, “90% of finance functions will deploy at least one ...

Nvidia will make 5 million B200/B300 which are the leading edge AI GPU chips. They could make about 10 million advanced next generation chips each year in about 2028. Tesla Dojo 1 chips have small ...

A monthly overview of things you need to know as an architect or aspiring architect. Unlock the full InfoQ experience by logging in! Stay updated with your favorite authors and topics, engage with ...