Tuesday, April 30, 2024
Show HN: I replicated Anthropic's monosemanticity research using just my MacBook https://ift.tt/5t4Li0h
Show HN: I replicated Anthropic's monosemanticity research using just my MacBook Hi everyone, I've been working on an open-source implementation of Anthropic's research on monosemanticity ("Towards Monosemanticity"). The problem Anthropic is trying to solve is that language models are hard to interpret because individual neurons can be responsible for multiple different things. The research finds that training a small autoencoder on neuron activations can result in "features" which are much easier to interpret. When I was reading the original research, I got really excited when I realized that the models they used were really small, and I could probably train them from scratch with just my M3 MBP. My models are somewhat undertrained compared to what Anthropic produced, but I think my results are still very compelling. Let me know what you think! https://ift.tt/Qwz58uV April 30, 2024 at 09:26PM
Monday, April 29, 2024
Show HN: Attorch – PyTorch's nn module written in Python using OpenAI's Triton https://ift.tt/f5MyGmT
Show HN: Attorch – PyTorch's nn module written in Python using OpenAI's Triton attorch is a subset of PyTorch's nn module, written purely in Python using OpenAI's Triton. Its goal is to be an easily hackable, self-contained, and readable collection of neural network modules whilst maintaining or improving upon the efficiency of PyTorch. In other words, it intends to be a forkable project endowed with a simple, intuitive design that can serve as an accessible starting point for those who are seeking to develop custom deep learning operations but are not satisfied with the speed of a pure PyTorch implementation and do not have the technical expertise or resources to write CUDA kernels. There already exist a number of wonderful PyTorch-like frameworks powered by Triton, but most concentrate solely on Transformers and NLP applications, whereas attorch aims to be more inclusive by also presenting a variety of layers pertaining to areas besides NLP such as computer vision. Moreover, attorch is not an inference-only package and fully supports both forward and backward passes, meaning it can be used during training as well as inference, though its performance for the latter is generally not on par with dedicated inference engines. Questions and feedback are welcome in the comments sections. https://ift.tt/rQtjiXH April 29, 2024 at 11:37PM
Sunday, April 28, 2024
Show HN: Dotenv, if it is a Unix utility https://ift.tt/d9gQCKS
Show HN: Dotenv, if it is a Unix utility I like the idea of using dotenv files, but I dislike having to use different language-specific libraries to read them. To solve this, I created a small utility that lets you prefix any command with "dotenv" to load the ".env" file. This is how I imagine dotenv would work if it had started as a UNIX utility rather than a Node.js library. https://ift.tt/Y2Ogk83 April 29, 2024 at 12:25AM
Saturday, April 27, 2024
Show HN: Htpy – generate HTML from Python without templates https://ift.tt/lVHBFwm
Show HN: Htpy – generate HTML from Python without templates I built a library that to generate HTML from Python. We have been using this library with Django the last couple of months instead of classic templates and find it to be productive. It is easy to debug, works great with static type checkers and it is easy to build reusable components/partials. Give it a try! https://htpy.dev April 27, 2024 at 11:34PM
Friday, April 26, 2024
Show HN: I made a site for practicing front end debugging with real-world bugs https://ift.tt/BArOz7q
Show HN: I made a site for practicing front end debugging with real-world bugs I am excited to introduce my solo project, a platform built with the frontend development community in mind. It's an interactive environment aimed at refining debugging skills through exposure to real-world bugs. This project stems from the need for a hands-on, practical method of learning to debug. The site features a variety of intentional bugs for users to solve, mirroring the types of challenges faced in professional settings. Over the coming weeks, I will be adding many more exercises and new features to enhance your learning experience further. Check out CodeMender at https://ift.tt/Ied6ZGV https://ift.tt/Ied6ZGV April 26, 2024 at 09:59PM
Thursday, April 25, 2024
Wednesday, April 24, 2024
Tuesday, April 23, 2024
Show HN: Storing Private Keys in the Browser Securely https://ift.tt/QjbAgGs
Show HN: Storing Private Keys in the Browser Securely So the main purpose here is to show _a_ way that session-token theft can be mitigated. Clearly, this isn't NSA proof or something you'd use to secure a BL5 containment facility, but to prevent session-jacking; if feels like it could help a lot, and would be pretty quick and easy to roll out if an IDP wanted to implement it. https://ift.tt/e1d6Q0T April 23, 2024 at 11:02PM
Monday, April 22, 2024
Sunday, April 21, 2024
Show HN: AppView 1.0.0 is released Instrument, Observe, Secure your deployments https://ift.tt/aBMfn8g
Show HN: AppView 1.0.0 is released Instrument, Observe, Secure your deployments AppView is an open source, runtime-agnostic instrumentation utility for any Linux command or application. It helps users to explore, understand, and gain visibility with no code modification. With one instrumentation approach for all runtimes, AppView offers ubiquitous, unified instrumentation of any unmodified Linux executable. With AppView 1.0.0 comes the new threat detection logic that allows users to capture security-related events. Is your application accessing secure files? Is it making connections it shouldn't be? Is it exfiltrating data over DNS or is GOT poisoning in effect? Other features of the 1.0.0 release include the ability to: - Generate metrics on process and application performance. - Generate events, reporting on network, file, logs, console messages and http/s activity. - Capture (decrypted) payload data without the need for keys. - Generate a stack trace, and a core dump when an application crashes. - Generate network flow information. - Create a report on unique file and network activity. - Install AppView in a Kubernetes cluster. - Secure file and network access in an application. - Instrument both static and dynamic executables. - Attach to processes while they are running or start when the process does. - Normalize and forward metrics and events, in real time, to remote systems. - Summarize metrics and detect protocols. We are looking for users and contributors alike. https://appview.org/ April 21, 2024 at 09:22PM
Saturday, April 20, 2024
Show HN: LLM Scraper – turn any webpage into structured data https://ift.tt/XUgOdrf
Show HN: LLM Scraper – turn any webpage into structured data https://ift.tt/Ld576Eu April 21, 2024 at 12:37AM
Show HN: Edit This Page – Frictionless Content Collaboration https://ift.tt/G4tBdS3
Show HN: Edit This Page – Frictionless Content Collaboration https://ift.tt/N2KLDqV April 20, 2024 at 06:57PM
Friday, April 19, 2024
Show HN: Composable (as in iGoogle, but modern) privacy-friendly new tab https://ift.tt/gmp2E53
Show HN: Composable (as in iGoogle, but modern) privacy-friendly new tab I spent quite a lot of time working on this one over the last 1.5 years. It started as a small project for my personal use because I wanted to keep all my self-hosted services visible so I wouldn't forget they existed lol. Using a web page wasn't ideal because of the white flicker every time I opened a new tab, so I decided to make this into a browser extension. From that time on, it became a lot bigger and got some traction (which I'm very happy about). It's made with React, but I tried to squeeze maximum performance (limited by my skills and desire to keep it somewhat readable, though) out of it. UI/UX was a big priority for me in this project, so I also tried to streamline it as much as possible and make Anori a joy to use. If you decide to try it, let me know how good I did! Oh, and it's open source [1] and the process of adding new widgets is documented [2], so you can make your own! [1]: https://ift.tt/KNEtbWn [2]: https://ift.tt/DBr5TQ8... https://anori.app/ April 19, 2024 at 12:34PM
Thursday, April 18, 2024
Show HN: Vapi – Convince our voice AI to give you the secret code https://ift.tt/dF6itID
Show HN: Vapi – Convince our voice AI to give you the secret code https://ift.tt/KhlwzSd April 18, 2024 at 08:50PM
Wednesday, April 17, 2024
Show HN: Easy Folders – A folders based organization tool for ChatGPT and Claude https://ift.tt/Q8y2FjP
Show HN: Easy Folders – A folders based organization tool for ChatGPT and Claude Download for free: https://ift.tt/J50uQwU A simple and powerful tool to organize chats in ChatGPT and Claude. ** NEW! Claude support added - Create folders and organize your chats for Claude ** ** NEW! Multi-account support added - Create folders and organize your chats for multiple ChatGPT & Claude accounts, and switch between them seamlessly. ** - Privacy focused, your data is all stored on your local machine - Create folders & sub-folders for your chats - Easy-to-use drag & drop - Colored folders - Automatic folders for Custom GPTs - Sync folders across multiple devices - Full chat history search - Bookmark important chats - Bulk delete chats - Backup & Import data - Dark & light mode support April 17, 2024 at 11:41PM
Tuesday, April 16, 2024
Show HN: Generated Documentation for Linux Kernel https://ift.tt/0u9FrsE
Show HN: Generated Documentation for Linux Kernel We are excited to release SAFA's generated documentation for the Linux Kernel. You will be able to view: -Project Overview (Overview, Features, Entities, Dataflow) -Code Files and their summaries (518 Code Files) -Functional Requirements (202 Functional Requirements) -Features (100 Features) -Mapped Relationships with explanations Learn more at https://safa.ai https://ift.tt/Wh5eAMY April 16, 2024 at 11:23PM
Monday, April 15, 2024
Sunday, April 14, 2024
Show HN: Gemfast – A rubygems server written in Go https://ift.tt/6FCao2y
Show HN: Gemfast – A rubygems server written in Go https://ift.tt/uIs6Fig April 14, 2024 at 10:55PM
Show HN: Minipic convert and compress images multiple formats locally in browser https://ift.tt/Fd0qQKy
Show HN: Minipic convert and compress images multiple formats locally in browser https://minipic.app April 14, 2024 at 10:53PM
Saturday, April 13, 2024
Show HN: ZSV (Zip Separated Values) columnar data format https://ift.tt/vIj0TKk
Show HN: ZSV (Zip Separated Values) columnar data format A columnar data format built using simple, mature technologies. https://ift.tt/bk3jdF1 April 13, 2024 at 11:52PM
Show HN: My $1k self-install, off-grid solar backup build for renters https://ift.tt/tnz7uj1
Show HN: My $1k self-install, off-grid solar backup build for renters https://sunboxlabs.com April 13, 2024 at 10:59PM
Friday, April 12, 2024
Show HN: Facelessvideos.ai – An app to create viral TikTok videos with AI https://ift.tt/RufVDz8
Show HN: Facelessvideos.ai – An app to create viral TikTok videos with AI https://ift.tt/7cF2ywi April 12, 2024 at 11:26PM
Thursday, April 11, 2024
Show HN: Aarde – Self Replicating Programs https://ift.tt/4iKk1Nh
Show HN: Aarde – Self Replicating Programs Last weekend I created a new visualization/demo. It's a sort of crossover between artificial life and defragmentation on MSDOS. The organisms "live" in the memory and continuously copy themselves to a free piece of memory. The programming language consists of only 5 instructions to keep it simple: S - Program start E - Program end F - Find location to copy to C - Copy to location J - Jump to start of program Every time a code is copied, a mutation can occur. We start with the following simple life form: SFCJE The rest is evolution. https://ift.tt/PGs2Rvj April 11, 2024 at 11:59PM
Wednesday, April 10, 2024
Tuesday, April 9, 2024
Show HN: Visualize eBay laptops in bulk with laptopscout https://ift.tt/fen6gP7
Show HN: Visualize eBay laptops in bulk with laptopscout A bit late in the new diskprices-like website trend, but here's my contribution to the ecosystem. The goal here is to get an instant comparison of the laptops on ebay, using mostly 3 metrics of now: CPU benchmark rank, GPU benchmark rank, and price, without having to scroll through the item description or google the PC model. At the moment there are some parsing bugs, and some components that aren't parsed yet. I also quickly run into the ebay rate limit, making it hard to maintain an hour-by-hour listing. Curious to know what you guys think! https://laptopscout.xyz/EBAY_US April 10, 2024 at 12:34AM
Monday, April 8, 2024
Show HN: The fastest way to run Mixtral 8x7B on Apple Silicon Macs https://ift.tt/su7H3Bi
Show HN: The fastest way to run Mixtral 8x7B on Apple Silicon Macs I’d originally launched my app: Private LLM[1][2] on HN around 10 months ago, with a single RedPajama Chat 3B model. The app has come a long way since then. About a month ago, I added support for 4-bit OmniQuant quantized Mixtral 8x7B Instruct model, and it seems to outperform Q4 models at inference speed and Q8 models at text generation quality, while consuming only about 24GB of RAM[3] at 8k context length. The trick is: a) to use a better quantization algorithm and b) to use unquantized embeddings and the MoE gates (the overhead is quite small). Other notable features include many more downloadable models, support for App Intents (Siri, Apple Shortcuts), on-device grammar correction, summarization etc with macOS services and an iOS version (universal app), also with many smaller downloadable models and support for App Intents. There's a small community of users building and sharing LLM based shortcuts on the App's discord. Last week, I also shipped support for the bilingual Yi-34B Chat model, which consumes ~18GB of RAM. iOS users and users with low memory Macs can download the related Yi-6B Chat model. Unlike most popular offline LLM apps out there, this app uses mlc-llm for inference and not llama.cpp. Also, all models in the app are quantized with OmniQuant[4] quantization and not RTN quantization. [1]: https://privatellm.app/ [2]: https://ift.tt/UzDCYIO [3]: https://www.youtube.com/watch?v=4AE8yXIWSAA [4]: https://ift.tt/3YvrngM April 8, 2024 at 08:07PM
Sunday, April 7, 2024
Show HN: HomeStage – Instant virtual furnishing with one click https://ift.tt/GufONUr
Show HN: HomeStage – Instant virtual furnishing with one click https://homestage.app April 8, 2024 at 01:10AM
Saturday, April 6, 2024
Show HN: Bluetuith – A console TUI-based Bluetooth manager https://ift.tt/ph8QcqZ
Show HN: Bluetuith – A console TUI-based Bluetooth manager Hello HN, bluetuith is a Bluetooth manager for the terminal, which aims to replace most Bluetooth managers, and can perform any Bluetooth based operations and interact with devices. Please have a look at the repository and the documentation for more information. I have been working on this for some time now, and I would like constructive suggestions to further improve the application, especially with regards to cross-platform functionality, so that the user experience can be enhanced. Any suggestions will be highly appreciated. Contributors are welcome as well. https://ift.tt/b4d0fD7 April 6, 2024 at 10:47AM
Show HN: Kiwi – End-to-End Kafka Subscriptions with WebAssembly https://ift.tt/WR9pOuN
Show HN: Kiwi – End-to-End Kafka Subscriptions with WebAssembly Hey HN! I'm excited to announce the release of Kiwi v0.1.1. I started working on Kiwi a few months ago with the primary motivation of providing an extensible WebSocket adapter that allows client applications to securely subscribe to real-time data sources such as Kafka. Often times, there is a desire to extend the real-time data that flows through sources like Kafka to end users in a fast and secure manner. Rather than standing up custom solutions, Kiwi lets operators focus on writing business logic, by loading custom WASM plugins that handle tasks like authorization and event filtering per client. There is a lot more information in the README. Any feedback would be much appreciated :) Thanks! https://ift.tt/nLt9ipE April 6, 2024 at 08:30PM
Friday, April 5, 2024
Thursday, April 4, 2024
Show HN: Managed GitHub Actions Runners for AWS https://ift.tt/M6ZEKCR
Show HN: Managed GitHub Actions Runners for AWS Hey HN! I'm Jacob, one of the founders of Depot ( https://depot.dev ), a build service for Docker images, and I'm excited to show what we’ve been working on for the past few months: run GitHub Actions jobs in AWS, orchestrated by Depot! Here's a video demo: https://www.youtube.com/watch?v=VX5Z-k1mGc8 , and here’s our blog post: https://ift.tt/xy7IQ5X . While GitHub Actions is one of the most prevalent CI providers, Actions is slow, for a few reasons: GitHub uses underpowered CPUs, network throughput for cache and the internet at large is capped at 1 Gbps, and total cache storage is limited to 10GB per repo. It is also rather expensive for runners with more than 2 CPUs, and larger runners frequently take a long time to start running jobs. Depot-managed runners solve this! Rather than your CI jobs running on GitHub's slow compute, Depot routes those same jobs to fast EC2 instances. And not only is this faster, it’s also 1/2 the cost of GitHub Actions! We do this by launching a dedicated instance for each job, registering that instance as a self-hosted Actions runner in your GitHub organization, then terminating the instance when the job is finished. Using AWS as the compute provider has a few advantages: - CPUs are typically 30%+ more performant than alternatives (the m7a instance type). - Each instance has high-throughput networking of up to 12.5 Gbps, hosted in us-east-1, so interacting with artifacts, cache, container registries, or the internet at large is quick. - Each instance has a public IPv4 address, so it does not share rate limits with anyone else. We integrated the runners with the distributed cache system (backed by S3 and Ceph) that we use for Docker build cache, so jobs automatically save / restore cache from this cache system, with speeds of up to 1 GB/s, and without the default 10 GB per repo limit. Building this was a fun challenge; some matrix workflows start 40+ jobs at once, then requiring 40 EC2 instances to launch at once. We’ve effectively gotten very good at starting EC2 instances with a "warm pool" system which allows us to prepare many EC2 instances to run a job, stop them, then resize and start them when an actual job request arrives, to keep job queue times around 5 seconds. We're using a homegrown orchestration system, as alternatives like autoscaling groups or Kubernetes weren't fast or secure enough. There are three alternatives to our managed runners currently: 1. GitHub offers larger runners: these have more CPUs, but still have slow network and cache. Depot runners are also 1/2 the cost per minute of GitHub's runners. 2. You can self-host the Actions runner on your own compute: this requires ongoing maintenance, and it can be difficult to ensure that the runner image or container matches GitHub's. 3. There are other companies offering hosted GitHub Actions runners, though they frequently use cheaper compute hosting providers that are bottlenecked on network throughput or geography. Any feedback is very welcome! You can sign up at https://ift.tt/uAZyFYx for a free trial if you'd like to try it out on your own workflows. We aren't able to offer a trial without a signup gate, both because using it requires installing a GitHub app, and we're offering build compute, so we need some way to keep out the cryptominers :) April 4, 2024 at 06:32PM
Show HN: Generate a YC application with GPT-4 https://ift.tt/9cK0frF
Show HN: Generate a YC application with GPT-4 https://ift.tt/QVFHY3S April 4, 2024 at 11:43PM
Wednesday, April 3, 2024
Tuesday, April 2, 2024
Subscribe to:
Posts (Atom)