The best Hacker News stories from Show from the past day

from

All Show

from the past*

day week month

with these search terms:

Create follow Go back

Latest posts:

Show HN: MonoDevelop

Show HN: finetune LLMs via the Finetuning Hub

Hi HN community, I have been working on benchmarking publicly available LLMs these past couple of weeks. More precisely, I am interested on the finetuning piece since a lot of businesses are starting to entertain the idea of self-hosting LLMs trained on their proprietary data rather than relying on third party APIs.<p>To this point, I am tracking the following 4 pillars of evaluation that businesses are typically look into: - Performance - Time to train an LLM - Cost to train an LLM - Inference (throughput / latency / cost per token)<p>For each LLM, my aim is to benchmark them for popular tasks, i.e., classification and summarization. Moreover, I would like to compare them against each other.<p>So far, I have benchmarked Flan-T5-Large, Falcon-7B and RedPajama and have found them to be very efficient in low-data situations, i.e., when there are very few annotated samples. Llama2-7B/13B and Writer’s Palmyra are in the pipeline.<p>But there’s so many LLMs out there! In case this work interests you, would be great to join forces.<p>GitHub repo attached — feedback is always welcome :)<p>Happy hacking!

Show HN: finetune LLMs via the Finetuning Hub

Hi HN community, I have been working on benchmarking publicly available LLMs these past couple of weeks. More precisely, I am interested on the finetuning piece since a lot of businesses are starting to entertain the idea of self-hosting LLMs trained on their proprietary data rather than relying on third party APIs.<p>To this point, I am tracking the following 4 pillars of evaluation that businesses are typically look into: - Performance - Time to train an LLM - Cost to train an LLM - Inference (throughput / latency / cost per token)<p>For each LLM, my aim is to benchmark them for popular tasks, i.e., classification and summarization. Moreover, I would like to compare them against each other.<p>So far, I have benchmarked Flan-T5-Large, Falcon-7B and RedPajama and have found them to be very efficient in low-data situations, i.e., when there are very few annotated samples. Llama2-7B/13B and Writer’s Palmyra are in the pipeline.<p>But there’s so many LLMs out there! In case this work interests you, would be great to join forces.<p>GitHub repo attached — feedback is always welcome :)<p>Happy hacking!

Show HN: A .bit gateway for resolving ipfs:// to https://

bit.site dynamically captures all requests to *.bit.site, automatically resolving the IPFS/IPNS/Skynet content hashes for the corresponding .bit account records. It returns the relevant static content via HTTPS, enabling users to access decentralized networks with zero configuration on modern browsers.<p>At the same time, we provide global CDN acceleration, striving to enhance the availability of decentralized content.

Show HN: A .bit gateway for resolving ipfs:// to https://

bit.site dynamically captures all requests to *.bit.site, automatically resolving the IPFS/IPNS/Skynet content hashes for the corresponding .bit account records. It returns the relevant static content via HTTPS, enabling users to access decentralized networks with zero configuration on modern browsers.<p>At the same time, we provide global CDN acceleration, striving to enhance the availability of decentralized content.

Show HN: Automating Job Search with AI

This is a personal experiment that uses LLMs to rank unstructured job posting data based on user-defined criteria. Traditional job search platforms rely on rigid filtering systems, but many users lack such concrete criteria.<p>One of the superpowers of LLMs is understanding unstructured data, like the job postings in the monthly "Ask HN: who's hiring" threads. So I built a little tool that lets you define your preferences in a more natural way and then rates each job postings based on the relevance.<p>You can define what you're looking for in simple terms and get a custom list ranked by relevance. It's not flawless (especially with cheaper models like gpt-3.5), but it's a lot better than searching through hundreds of listings manually.

Show HN: Extract an RSS feed from almost anything

Howdy! RSSfeedASAP scratches my own itch. I run a regional podcasting directory which gets dozens of messy submission for podcasts. Often they don't even include an xml file and me being a good samaritan I sometimes do the manual work and find it myself. I got tired of that manual work and decided to build a microapp.<p>RSSfeedASAP is this app and I decided to release it in case someone else finds any use in it.

Show HN: Extract an RSS feed from almost anything

Howdy! RSSfeedASAP scratches my own itch. I run a regional podcasting directory which gets dozens of messy submission for podcasts. Often they don't even include an xml file and me being a good samaritan I sometimes do the manual work and find it myself. I got tired of that manual work and decided to build a microapp.<p>RSSfeedASAP is this app and I decided to release it in case someone else finds any use in it.

Show HN: Extract an RSS feed from almost anything

Howdy! RSSfeedASAP scratches my own itch. I run a regional podcasting directory which gets dozens of messy submission for podcasts. Often they don't even include an xml file and me being a good samaritan I sometimes do the manual work and find it myself. I got tired of that manual work and decided to build a microapp.<p>RSSfeedASAP is this app and I decided to release it in case someone else finds any use in it.

Show HN: Modular Diffusion – A modular Python library for diffusion models

Hello everyone! I've been working on this project for a few months as part of my thesis in Machine Learning. It's meant to be a library that provides an easy-to-use but flexible API to design and train Diffusion Models. I decided to make it because I wanted to quickly prototype a Diffusion Model but there were no good tools to do it with. I think it really can help people prototype their own Diffusion Models a lot faster and only in a few lines of code.<p>The base idea is to have a Model class that takes different modules corresponding to the different aspects of the Diffusion Model process (noise schedule, noise type, denoising network, loss function, guidance, etc.) and allow the user to mix and match different modules to achieve different results. The library ships with a bunch of prebuilt modules and the plan is to add many more. I also made it super easy to implement your own modules, you just need to extend from one of the base classes available.<p>Contrary to HuggingFace Diffusers, this library is focused on designing and training your own Diffusion Models rather than finetuning pretrained ones (although this is possible).<p>I would really appreciate your feedback.

Show HN: Modular Diffusion – A modular Python library for diffusion models

Hello everyone! I've been working on this project for a few months as part of my thesis in Machine Learning. It's meant to be a library that provides an easy-to-use but flexible API to design and train Diffusion Models. I decided to make it because I wanted to quickly prototype a Diffusion Model but there were no good tools to do it with. I think it really can help people prototype their own Diffusion Models a lot faster and only in a few lines of code.<p>The base idea is to have a Model class that takes different modules corresponding to the different aspects of the Diffusion Model process (noise schedule, noise type, denoising network, loss function, guidance, etc.) and allow the user to mix and match different modules to achieve different results. The library ships with a bunch of prebuilt modules and the plan is to add many more. I also made it super easy to implement your own modules, you just need to extend from one of the base classes available.<p>Contrary to HuggingFace Diffusers, this library is focused on designing and training your own Diffusion Models rather than finetuning pretrained ones (although this is possible).<p>I would really appreciate your feedback.

Show HN: Menu Bar Calendar on macOS

Show HN: Menu Bar Calendar on macOS

Show HN: XRss – RSS Reader and web stack demo

XRss is a simple RSS reader web app built to showcase xtemplate, a new web development tool based on Go's html/template and Caddy server.<p>The entire site UI for XRss comes from <i>a single HTML template file</i>. This index.html includes everything from SQL queries and route definitions and handlers to htmx state transition attributes and tailwindcss classes, and developing it requires <i>zero build steps</i> (amortized).<p>Check out the source which manages to be at once banal and gnarly: <a href="https://github.com/infogulch/xrss/blob/master/templates/index.html">https://github.com/infogulch/xrss/blob/master/templates/inde...</a><p>xtemplate preloads the whole template structure into memory and builds the router at startup, so responses to matching requests are rendered after a single lookup. Combined with direct queries to sqlite makes for a very snappy experience typically responding in less than 5ms. (Fingers crossed.)<p>There are various places where XRss could be improved (PRs welcome!), but it already delivers on its purpose of demonstrating the plausibility of xtemplate. See the xtemplate readme for an overview of what you can do with it. I think of it as 'PHP but the syntax looks like Go templates'.<p><a href="https://github.com/infogulch/caddy-xtemplate">https://github.com/infogulch/caddy-xtemplate</a><p>Let me know what you think! Does remaking PHP from scratch out of Go templates make me a lunatic? (yes) Is it a good idea anyway? (yes) What kind of web application do you think would be a good fit for a platform like this?

Show HN: XRss – RSS Reader and web stack demo

XRss is a simple RSS reader web app built to showcase xtemplate, a new web development tool based on Go's html/template and Caddy server.<p>The entire site UI for XRss comes from <i>a single HTML template file</i>. This index.html includes everything from SQL queries and route definitions and handlers to htmx state transition attributes and tailwindcss classes, and developing it requires <i>zero build steps</i> (amortized).<p>Check out the source which manages to be at once banal and gnarly: <a href="https://github.com/infogulch/xrss/blob/master/templates/index.html">https://github.com/infogulch/xrss/blob/master/templates/inde...</a><p>xtemplate preloads the whole template structure into memory and builds the router at startup, so responses to matching requests are rendered after a single lookup. Combined with direct queries to sqlite makes for a very snappy experience typically responding in less than 5ms. (Fingers crossed.)<p>There are various places where XRss could be improved (PRs welcome!), but it already delivers on its purpose of demonstrating the plausibility of xtemplate. See the xtemplate readme for an overview of what you can do with it. I think of it as 'PHP but the syntax looks like Go templates'.<p><a href="https://github.com/infogulch/caddy-xtemplate">https://github.com/infogulch/caddy-xtemplate</a><p>Let me know what you think! Does remaking PHP from scratch out of Go templates make me a lunatic? (yes) Is it a good idea anyway? (yes) What kind of web application do you think would be a good fit for a platform like this?

Show HN: XRss – RSS Reader and web stack demo

XRss is a simple RSS reader web app built to showcase xtemplate, a new web development tool based on Go's html/template and Caddy server.<p>The entire site UI for XRss comes from <i>a single HTML template file</i>. This index.html includes everything from SQL queries and route definitions and handlers to htmx state transition attributes and tailwindcss classes, and developing it requires <i>zero build steps</i> (amortized).<p>Check out the source which manages to be at once banal and gnarly: <a href="https://github.com/infogulch/xrss/blob/master/templates/index.html">https://github.com/infogulch/xrss/blob/master/templates/inde...</a><p>xtemplate preloads the whole template structure into memory and builds the router at startup, so responses to matching requests are rendered after a single lookup. Combined with direct queries to sqlite makes for a very snappy experience typically responding in less than 5ms. (Fingers crossed.)<p>There are various places where XRss could be improved (PRs welcome!), but it already delivers on its purpose of demonstrating the plausibility of xtemplate. See the xtemplate readme for an overview of what you can do with it. I think of it as 'PHP but the syntax looks like Go templates'.<p><a href="https://github.com/infogulch/caddy-xtemplate">https://github.com/infogulch/caddy-xtemplate</a><p>Let me know what you think! Does remaking PHP from scratch out of Go templates make me a lunatic? (yes) Is it a good idea anyway? (yes) What kind of web application do you think would be a good fit for a platform like this?

Show HN: XRss – RSS Reader and web stack demo

XRss is a simple RSS reader web app built to showcase xtemplate, a new web development tool based on Go's html/template and Caddy server.<p>The entire site UI for XRss comes from <i>a single HTML template file</i>. This index.html includes everything from SQL queries and route definitions and handlers to htmx state transition attributes and tailwindcss classes, and developing it requires <i>zero build steps</i> (amortized).<p>Check out the source which manages to be at once banal and gnarly: <a href="https://github.com/infogulch/xrss/blob/master/templates/index.html">https://github.com/infogulch/xrss/blob/master/templates/inde...</a><p>xtemplate preloads the whole template structure into memory and builds the router at startup, so responses to matching requests are rendered after a single lookup. Combined with direct queries to sqlite makes for a very snappy experience typically responding in less than 5ms. (Fingers crossed.)<p>There are various places where XRss could be improved (PRs welcome!), but it already delivers on its purpose of demonstrating the plausibility of xtemplate. See the xtemplate readme for an overview of what you can do with it. I think of it as 'PHP but the syntax looks like Go templates'.<p><a href="https://github.com/infogulch/caddy-xtemplate">https://github.com/infogulch/caddy-xtemplate</a><p>Let me know what you think! Does remaking PHP from scratch out of Go templates make me a lunatic? (yes) Is it a good idea anyway? (yes) What kind of web application do you think would be a good fit for a platform like this?

Show HN: Shaq, a CLI for Shazam

Show HN: Shaq, a CLI for Shazam

Show HN: Shaq, a CLI for Shazam