Sigmoid Social

olеg lаvrоvskyIn response to my question on <a href="https://hachyderm.io/tags/openhardware" class="mention hashtag" rel="nofollow noopener" target="_blank">#openhardware</a> and alternatives to CUDA, he hopes that <a href="https://hachyderm.io/tags/OpenCL" class="mention hashtag" rel="nofollow noopener" target="_blank">#OpenCL</a> will rise to the challenge of demanding 64-bit+ models rather than the current focus on 4-bit quantized builds that are insufficient for HPC science.See also my notes and links on MLP like <a href="https://hachyderm.io/tags/Apertus" class="mention hashtag" rel="nofollow noopener" target="_blank">#Apertus</a> at <a href="https://hachyderm.io/tags/CSCS" class="mention hashtag" rel="nofollow noopener" target="_blank">#CSCS</a> at <a href="https://swissai.dribdat.cc/project/14" rel="nofollow noopener" translate="no" target="_blank">https://swissai.dribdat.cc/project/14</a>

olеg lаvrоvskyHow much power does 3 month of training <a href="https://hachyderm.io/tags/Apertus" class="mention hashtag" rel="nofollow noopener" target="_blank">#Apertus</a> on the most powerful open research cloud-native supercomputer in Europe? ⚡️About the same as raising 5 people to adulthood, or 50% peak (10MW) of a locomotive. Keep industrial power usage in perspective, and acknowledge the efforts into climate neutrality, asks Joost VandeVondele - who aims to scale up the infrastructure to support thousands of AI experiments and hundreds of SMEs in the future at <a href="https://hachyderm.io/tags/CSCS" class="mention hashtag" rel="nofollow noopener" target="_blank">#CSCS</a>

Neff :manjaro:<a href="https://mastodon.uno/tags/Apertus" class="mention hashtag" rel="nofollow noopener" target="_blank">#Apertus</a> : Un LLM svizzero, sviluppato dai due Politecnici federali <a href="https://mastodon.uno/tags/ETHZ" class="mention hashtag" rel="nofollow noopener" target="_blank">#ETHZ</a> e <a href="https://mastodon.uno/tags/EPFL" class="mention hashtag" rel="nofollow noopener" target="_blank">#EPFL</a> e il Centro Svizzero di Calcolo Scientifico <a href="https://mastodon.uno/tags/CSCS" class="mention hashtag" rel="nofollow noopener" target="_blank">#CSCS</a> promette applicazioni AI trasparenti e open source. Chissà come funziona! <a href="https://www.threads.com/@corrieredelticino" rel="nofollow noopener" translate="no" target="_blank">https://www.threads.com/@corrieredelticino</a>

Benjamin Carr, Ph.D. 👨🏻‍💻🧬<a href="https://hachyderm.io/tags/Switzerland" class="mention hashtag" rel="nofollow noopener" target="_blank">#Switzerland</a> launches transparent <a href="https://hachyderm.io/tags/ChatGPT" class="mention hashtag" rel="nofollow noopener" target="_blank">#ChatGPT</a> alternative <a href="https://hachyderm.io/tags/Apertus" class="mention hashtag" rel="nofollow noopener" target="_blank">#Apertus</a>: A fully <a href="https://hachyderm.io/tags/opensource" class="mention hashtag" rel="nofollow noopener" target="_blank">#opensource</a>, transparent, multilingual language model The Apertus <a href="https://hachyderm.io/tags/LLM" class="mention hashtag" rel="nofollow noopener" target="_blank">#LLM</a> is comparable to Meta’s <a href="https://hachyderm.io/tags/Llama3" class="mention hashtag" rel="nofollow noopener" target="_blank">#Llama3</a> model from 2024 The development of the <a href="https://hachyderm.io/tags/largelanguagemodel" class="mention hashtag" rel="nofollow noopener" target="_blank">#largelanguagemodel</a> (<a href="https://hachyderm.io/tags/LLM" class="mention hashtag" rel="nofollow noopener" target="_blank">#LLM</a>) and the research into other domain-specific foundation models is funded by an investment of over 10 million GPU hours on the “<a href="https://hachyderm.io/tags/Alps" class="mention hashtag" rel="nofollow noopener" target="_blank">#Alps</a>” <a href="https://hachyderm.io/tags/Supercomputer" class="mention hashtag" rel="nofollow noopener" target="_blank">#Supercomputer</a> by <a href="https://hachyderm.io/tags/CSCS" class="mention hashtag" rel="nofollow noopener" target="_blank">#CSCS</a>, and by the ETH Board <a href="https://www.cscs.ch/science/computer-science-hpc/2025/apertus-a-fully-open-transparent-multilingual-language-model" rel="nofollow noopener" translate="no" target="_blank">https://www.cscs.ch/science/computer-science-hpc/2025/apertus-a-fully-open-transparent-multilingual-language-model</a>

NiclasI recently visited the Swiss National Supercomputing Centre ( <a href="https://fosstodon.org/tags/CSCS" class="mention hashtag" rel="nofollow noopener" target="_blank">#CSCS</a> )in Lugano.Now <a href="https://fosstodon.org/tags/ETH" class="mention hashtag" rel="nofollow noopener" target="_blank">#ETH</a> and <a href="https://fosstodon.org/tags/EPFL" class="mention hashtag" rel="nofollow noopener" target="_blank">#EPFL</a> have publicly announced that they are training an <a href="https://fosstodon.org/tags/LLM" class="mention hashtag" rel="nofollow noopener" target="_blank">#LLM</a> on "Alps" supercomputing. The best part is that they are doing it <a href="https://fosstodon.org/tags/OpenSource" class="mention hashtag" rel="nofollow noopener" target="_blank">#OpenSource</a>"Alps" will compute at 0.435 * 10^18 FLOPS. This means that "Alps" can calculate more in one day than a modern laptop in 2025 could calculate in 40,000 years. Amazing!<a href="https://ethz.ch/en/news-and-events/eth-news/news/2025/07/a-language-model-built-for-the-public-good.html" rel="nofollow noopener" translate="no" target="_blank">https://ethz.ch/en/news-and-events/eth-news/news/2025/07/a-language-model-built-for-the-public-good.html</a><a href="https://fosstodon.org/tags/Switzerland" class="mention hashtag" rel="nofollow noopener" target="_blank">#Switzerland</a> <a href="https://fosstodon.org/tags/SwissAI" class="mention hashtag" rel="nofollow noopener" target="_blank">#SwissAI</a> <a href="https://fosstodon.org/tags/Europe" class="mention hashtag" rel="nofollow noopener" target="_blank">#Europe</a> <a href="https://fosstodon.org/tags/AI" class="mention hashtag" rel="nofollow noopener" target="_blank">#AI</a> <a href="https://fosstodon.org/tags/supercomputer" class="mention hashtag" rel="nofollow noopener" target="_blank">#supercomputer</a>

🧿🪬🍄🌈🎮💻🚲🥓🎃💀🏴🛻🇺🇸<a href="https://mastodon.social/tags/ETH" class="mention hashtag" rel="nofollow noopener" target="_blank">#ETH</a> Zurich & <a href="https://mastodon.social/tags/EPFL" class="mention hashtag" rel="nofollow noopener" target="_blank">#EPFL</a> (w/ <a href="https://mastodon.social/tags/CSCS" class="mention hashtag" rel="nofollow noopener" target="_blank">#CSCS</a>) will drop a fully open large-language model trained on the “Alps” supercomputer. Built from scratch on public infrastructure, fluent in 1000 + languages, Apache-2 licensed. Landing late-summer 2025> A language model built for the public good<a href="https://ethz.ch/en/news-and-events/eth-news/news/2025/07/a-language-model-built-for-the-public-good.html" rel="nofollow noopener" translate="no" target="_blank">https://ethz.ch/en/news-and-events/eth-news/news/2025/07/a-language-model-built-for-the-public-good.html</a><a href="https://mastodon.social/tags/OpenSource" class="mention hashtag" rel="nofollow noopener" target="_blank">#OpenSource</a> <a href="https://mastodon.social/tags/LLM" class="mention hashtag" rel="nofollow noopener" target="_blank">#LLM</a> <a href="https://mastodon.social/tags/AIForGood" class="mention hashtag" rel="nofollow noopener" target="_blank">#AIForGood</a> <a href="https://mastodon.social/tags/ai" class="mention hashtag" rel="nofollow noopener" target="_blank">#ai</a> <a href="https://mastodon.social/tags/llms" class="mention hashtag" rel="nofollow noopener" target="_blank">#llms</a> <a href="https://mastodon.social/tags/switzerland" class="mention hashtag" rel="nofollow noopener" target="_blank">#switzerland</a>

Recent searches

Search options

Administered by:

Server stats:

#cscs