# forck.live — 11–17 September 2023 > AI lab announcements, news reports and podcasts. forck.live follows what companies announce about their own products. Press may corroborate an official story; it never leads and never appears as an item of its own. A missing date stays missing. A title is the publisher's words — where a company does not publish in English, an English rendering is listed separately as `Title in English` and named as this site's translation, never as theirs. This file is one issue of the weekly archive: every confirmed first-party announcement of ISO week 2023-W37, 11–17 September 2023, newest first. The window never moves, so this URL always names the same seven days — but a post published inside them and detected later is added when it arrives, so a closed week can still gain entries. Every entry is one first-party announcement, and carries the same eight fields in this order before any others: Company · Announced · Category · Coverage · Announcement · Group · Source · Record `Source` is the company's own page the announcement was traced to and `Record` is this site's page for it. Both are safe to cite. `Coverage` is a number of outlets, or `not counted` — which means this record reached no count of its own, never that nobody wrote about it. `Category` is this tracker's own analysis, and is authoritative wherever it is stated. `not stated` means either that nothing has analysed the post yet or that the analysis placed it outside this taxonomy; neither is a verdict about the announcement. `Announcement` is whether the post is news of something the company did, rather than a how-to, a customer story, a changelog listing or a staffing note. `Group` is which of this record's groups the entry falls in, carried on the entry itself so it holds where the groups are headings and where they are not. One of: `models`, `covered`, `announcements`, `routine`. Anything after `Record` is held detail that only some announcements carry. - This week's page: https://forck.live/week/2023-W37 - The current issue, and the catalogue of the other weeks: https://forck.live/briefing.md - Catalogue: https://forck.live/llms.txt ## Announcements ### Replit — Introducing the Replit Desktop App - Company: Replit (replit.com) - Announced: 2023-09-14 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://replit.com/blog/desktop-app - Record: https://forck.live/items/18153-introducing-the-replit-desktop-app - Subject: Replit Agent Replit announced the official Replit Desktop App, a native application for macOS, Windows, and Linux that provides a browser-free coding experience with features like multiple windows, native keyboard shortcuts, and native Chromium devtools. The app is built using Electron and is available immediately. ### OpenAI — Introducing OpenAI Dublin - Company: OpenAI (openai.com) - Announced: 2023-09-13T07:00:00+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://openai.com/index/introducing-openai-dublin - Record: https://forck.live/items/839-introducing-openai-dublin - Subject: GPT / ChatGPT / API OpenAI is opening an office in Dublin, Ireland to expand its presence in Europe. ### Hugging Face — Fine-tuning Llama 2 70B using PyTorch FSDP - Company: Hugging Face (huggingface.co) - Announced: 2023-09-13 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://huggingface.co/blog/ram-efficient-pytorch-fsdp - Record: https://forck.live/items/1998-fine-tuning-llama-2-70b-using-pytorch-fsdp - Subject: Platform Best practices for fine-tuning Llama 2 70B using PyTorch FSDP, addressing challenges such as CPU RAM usage during model loading, checkpoint saving, and training speed optimization. ### Hugging Face — Introducing Würstchen: Fast Diffusion for Image Generation - Company: Hugging Face (huggingface.co) - Announced: 2023-09-13 - Category: new-model - Coverage: not counted - Announcement: yes - Group: models - Source: https://huggingface.co/blog/wuerstchen - Record: https://forck.live/items/1997-introducing-w-rstchen-fast-diffusion-for-image-generation - Subject: Platform Hugging Face introduces Würstchen, a diffusion model that achieves 42x spatial compression and generates images faster and with less memory than Stable Diffusion XL, requiring significantly fewer training GPU hours. ### Hugging Face — Overview of natively supported quantization schemes in 🤗 Transformers - Company: Hugging Face (huggingface.co) - Announced: 2023-09-12 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://huggingface.co/blog/overview-quantization-transformers - Record: https://forck.live/items/1999-overview-of-natively-supported-quantization-schemes-in-transformers - Subject: Platform Hugging Face published a blog post comparing the bitsandbytes and auto-gptq quantization schemes supported natively in Transformers, discussing their benefits and drawbacks. ### Hugging Face — SafeCoder vs. Closed-source Code Assistants - Company: Hugging Face (huggingface.co) - Announced: 2023-09-11 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://huggingface.co/blog/safecoder-vs-closed-source-code-assistants - Record: https://forck.live/items/2000-safecoder-vs-closed-source-code-assistants - Subject: Platform Hugging Face compares its SafeCoder code assistant to closed-source alternatives, emphasizing benefits in transparency, customization, IT flexibility, and privacy.