# forck.live — 7–13 August 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-W32, 7–13 August 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-W32 - The current issue, and the catalogue of the other weeks: https://forck.live/briefing.md - Catalogue: https://forck.live/llms.txt ## Announcements ### Replit — Improving the Inline Ghostwriter Experience - Company: Replit (replit.com) - Announced: 2023-08-11 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://replit.com/blog/ghostwriter-inline - Record: https://forck.live/items/18072-improving-the-inline-ghostwriter-experience - Subject: Replit Agent Replit improved Ghostwriter's inline actions (Explain Code, Generate Code, Edit Code) by replacing popovers with persistent inline widgets, adding streaming for faster first-response times (sub-one second), syncing chat messages with the Ghostwriter pane, and introducing diff views for reviewing AI suggestions. Edit Code was renamed from Transform Code. ### Hugging Face — Hugging Face Hub on the AWS Marketplace: Pay with your AWS Account - Company: Hugging Face (huggingface.co) - Announced: 2023-08-10 - Category: availability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://huggingface.co/blog/aws-marketplace - Record: https://forck.live/items/2010-hugging-face-hub-on-the-aws-marketplace-pay-with-your-aws-account - Subject: Platform Hugging Face Hub is now available on AWS Marketplace, allowing AWS customers to subscribe and pay for Hugging Face services (Inference Endpoints, Spaces Hardware Upgrades, AutoTrain) directly through their AWS account, with integrated billing. ### Hugging Face — Deploying Hugging Face Models with BentoML: DeepFloyd IF in Action - Company: Hugging Face (huggingface.co) - Announced: 2023-08-09 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://huggingface.co/blog/deploy-deepfloydif-using-bentoml - Record: https://forck.live/items/2012-deploying-hugging-face-models-with-bentoml-deepfloyd-if-in-action - Subject: Platform To integrate DeepFloyd IF with BentoML by following the workflow of defining a model, saving it, creating a BentoML Service, building a Bento, and deploying it. ### Hugging Face — Optimizing Bark using 🤗 Transformers - Company: Hugging Face (huggingface.co) - Announced: 2023-08-09 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://huggingface.co/blog/optimizing-bark - Record: https://forck.live/items/2011-optimizing-bark-using-transformers - Subject: Platform Optimizing the Bark text-to-speech model using Hugging Face Transformers, Optimum, and Accelerate libraries, including benchmarking of latency and memory usage. ### Replit — Expand possibilities on Replit with Expandable Storage - Company: Replit (replit.com) - Announced: 2023-08-08 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://replit.com/blog/expandable-storage - Record: https://forck.live/items/18041-expand-possibilities-on-replit-with-expandable-storage - Subject: Replit Agent Replit introduced Expandable Storage, removing the previous 1 GiB per-Repl limit and allowing Repls to reach up to 256 GiB, with account-wide storage limits increased for all plans: Free to 10 GiB, Hacker to 20 GiB, and Pro to 50 GiB. Builders can also purchase à la carte storage to scale up to 1 TiB across their account. The new infrastructure uses incremental snapshots via a system called Margarine to improve performance for larger Repls. ### Replicate — Fine-tune SDXL with your own images - Company: Replicate (replicate.com) - Announced: 2023-08-08 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://replicate.com/blog/fine-tune-sdxl - Record: https://forck.live/items/2560-fine-tune-sdxl-with-your-own-images - Subject: Platform Replicate announces fine-tuning support for SDXL 1.0, enabling users to train the model on their own images using methods like Dreambooth, LoRA, and Textual Inversion. ### Hugging Face — Releasing Swift Transformers: Run On-Device LLMs in Apple Devices - Company: Hugging Face (huggingface.co) - Announced: 2023-08-08 - Category: developer-tool-release - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://huggingface.co/blog/swift-coreml-llm - Record: https://forck.live/items/2014-releasing-swift-transformers-run-on-device-llms-in-apple-devices - Subject: Platform Hugging Face released Swift Transformers, an in-development Swift package for text generation, along with swift-chat, updated exporters, transformers-to-coreml, and converted models (Llama 2 7B, Falcon 7B) to help developers run LLMs on Apple devices using Core ML. ### Hugging Face — Fine-tune Llama 2 with DPO - Company: Hugging Face (huggingface.co) - Announced: 2023-08-08 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://huggingface.co/blog/dpo-trl - Record: https://forck.live/items/2013-fine-tune-llama-2-with-dpo - Subject: Platform The Direct Preference Optimization (DPO) method available in the TRL library and demonstrates how to fine-tune the Llama v2 7B model on the stack-exchange preference dataset. ### Replit — Why We Changed Our Resource Limits and Plans - Company: Replit (replit.com) - Announced: 2023-08-07 - Category: not stated - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://replit.com/blog/new-limits-and-plans - Record: https://forck.live/items/18192-why-we-changed-our-resource-limits-and-plans - Subject: Replit Agent Replit gives users a computer in the cloud so they can code from any device. Our plan from the start has been to be the best place for anyone in the world to bring their idea to life. ### LG AI Research — LG AI Talk Concert 2023, Large-Scale AI EXAONE 2.0 Unveiled for the First Time - Company: LG AI Research (lgresearch.ai) - Announced: 2023-08-07 - Category: new-model - Coverage: not counted - Announcement: yes - Group: models - Source: https://www.lgresearch.ai/news/view?seq=336 - Record: https://forck.live/items/11046-lg-ai-talk-concert-2023-large-scale-ai-exaone-2-0-unveiled-for-the-first-time - Subject: EXAONE LG AI Research unveiled EXAONE 2.0, a large-scale multimodal AI, at its LG AI Talk Concert 2023. The model reduces inference time by 25% and memory usage by 70% compared to its predecessor, achieving the same performance with a 78% cost reduction. EXAONE 2.0 will be offered through three platforms—Universe, Discovery, and Atelier—and as customized models tailored to client needs. ### LG AI Research — LG AI Talk Concert 2023, Large-Scale AI EXAONE 2.0 Unveiled for the First Time - Company: LG AI Research (lgresearch.ai) - Announced: 2023-08-07 - Category: new-model - Coverage: not counted - Announcement: yes - Group: models - Source: https://www.lgresearch.ai/blog/view?seq=336 - Record: https://forck.live/items/4894-lg-ai-talk-concert-2023-large-scale-ai-exaone-2-0-unveiled-for-the-first-time - Subject: EXAONE LG AI Research unveiled EXAONE 2.0, an advanced large-scale multimodal AI model, at the LG AI Talk Concert 2023. It features improved data quality, cost efficiency, and customized model options for clients.