Show HN: Ftail – logger for Rust with multiple drivers (console/daily file/etc.) https://ift.tt/pBHugVO

Show HN: Ftail – logger for Rust with multiple drivers (console/daily file/etc.) https://ift.tt/nae6TPy September 14, 2024 at 12:27AM

Show HN: Fabrix – Instantly Generate UI Frontend from GraphQL https://ift.tt/GAMNq8T

Show HN: Fabrix – Instantly Generate UI Frontend from GraphQL https://ift.tt/2VgeJLl September 13, 2024 at 03:40AM

Reimagined Potrero Yard Tour: Walk the Project Site and Visualize Its Future

Reimagined Potrero Yard Tour: Walk the Project Site and Visualize Its Future
By John Angelico

Join us this weekend at Potrero Yard to learn how we’re reimagining the site. You might have heard that we are rebuilding Potrero Yard. Join us for a self-guided tour to learn about the last hundred years of this bus yard and what’s in store next. On Saturday, Sept. 14, come experience a community walking tour of the Potrero Yard Modernization Project site. This self-guided tour starts at Franklin Square and goes along 17th and Bryant streets.   Date: Saturday, Sept. 14, 2024  Time: 1-3 p.m.  Where: Franklin Square, 2500 17th Street, San Francisco  (Enter at the corner of 17th and Bryant...

Published September 12, 2024 at 05:30AM
https://ift.tt/zgcXsBy

Show HN: Hacker News for Creators/Makers https://ift.tt/9v4dxUE

Show HN: Hacker News for Creators/Makers https://ift.tt/u21vs67 September 13, 2024 at 02:51AM

Show HN: A YouTube videos course generator https://ift.tt/r086qmg

Show HN: A YouTube videos course generator REPOST! (to get more feedback from HN) An exciting aspect of what I'm working on is that users can create their own playlists. For example, you can provide a description of what you're learning, such as a lesson title, and the pipeline will create a youtube playlist. Since the content is highly customized, it will be relevant and high-quality, matching current lessons or lectures in school or university. I'm trying to find a product structure that users will find valuable and effective. The web app is completely free to use right now as I figure out the best model. https://ift.tt/Jk5hPSK September 12, 2024 at 10:35PM

Show HN: Mandoline – Custom LLM Evaluations for Real-World Use Cases https://ift.tt/4niTk5R

Show HN: Mandoline – Custom LLM Evaluations for Real-World Use Cases Hi HN! We're a small team of AI engineers who've spent the last few years building AI applications. Through this, we've experienced firsthand many of the challenges that come with evaluating and improving AI systems in real-world contexts. Standard LLM evaluations (and evaluation methods) often use simplified scenarios that don't reflect the complexity LLMs encounter in actual use. This leads to a disconnect between reported performance and real-world usefulness. We built Mandoline to bridge this gap, helping developers evaluate and improve LLM applications in ways that matter to end-users. Our approach allows you to design custom evaluation criteria that align with your specific product requirements. For a quick overview of how it works, check out our Python and / or Node SDK READMEs: - Python: https://ift.tt/x79CG4u... - Node / Typescript: https://ift.tt/M5vdNmW... Hopefully this design is flexible yet scalable, and helps you do things like: track LLM progress over time, make informed AI system design decisions, choose the best model for your use case, prompt engineer more systematically, and so on. Under the hood, Mandoline is a hybrid system using a combination of our own models and top general-purpose LLM APIs. We used Mandoline to evaluate and improve itself, which helped us make better decisions about system design. In the future, we’ll be adding visualization tools to more easily analyze trends, and expanding our in-house models capabilities to reduce reliance on (and hopefully outperform) external models. Check out our website ( https://mandoline.ai/ ) and documentation ( https://ift.tt/FcY4uKR ) to learn more. We’d love to hear about your experiences with evaluating AI systems for production use. What have you found most challenging in evaluating AI systems? What behaviors are hard to quantify? How could Mandoline fit into your workflow? You can reach us here in the comments or send us an email (hi@mandoline.ai). We appreciate you taking the time to learn a bit about Mandoline. https://mandoline.ai September 12, 2024 at 12:33AM

Show HN: Tune LLaMa3.1 on Google Cloud TPUs https://ift.tt/wpZQuHt

Show HN: Tune LLaMa3.1 on Google Cloud TPUs Hey HN, we wanted to share our repo where we fine-tuned Llama 3.1 on Google TPUs. We’re building AI infra to fine-tune and serve LLMs on non-NVIDIA GPUs (TPUs, Trainium, AMD GPUs). The problem: Right now, 90% of LLM workloads run on NVIDIA GPUs, but there are equally powerful and more cost-effective alternatives out there. For example, training and serving Llama 3.1 on Google TPUs is about 30% cheaper than NVIDIA GPUs. But developer tooling for non-NVIDIA chipsets is lacking. We felt this pain ourselves. We initially tried using PyTorch XLA to train Llama 3.1 on TPUs, but it was rough: xla integration with pytorch is clunky, missing libraries (bitsandbytes didn't work), and cryptic HuggingFace errors. We then took a different route and translated Llama 3.1 from PyTorch to JAX. Now, it’s running smoothly on TPUs! We still have challenges ahead, there is no good LoRA library in JAX, but this feels like the right path forward. Here's a demo ( https://ift.tt/FmrGgEu ) of our managed solution. Would love your thoughts on our repo and vision as we keep chugging along! https://ift.tt/AMLQDwU September 11, 2024 at 08:44PM