Show HN: AI Shopping Assistant https://ift.tt/WRs7kgm

Show HN: AI Shopping Assistant Hi! My friend and I built an AI Shopping Assistant powered by ChatGPT, called ShopMigo. The primary purpose of this project was to learn more about LLMs, embeddings, and AI in general. This started more as a pet project, but turned into something that we actually find really useful. I’m the kind of person that will take weeks to do research before buying an electronic or an expensive product. ShopMigo helps to guide you through a purchase and aggregates reviews so you don’t have to spend weeks/months doing your due diligence. Try it out! Let us know what you think! https://ift.tt/H9aMukW December 3, 2023 at 12:23AM

Show HN: 80% faster, 50% less memory, 0% loss of accuracy Llama finetuning https://ift.tt/xTQbYzh

Show HN: 80% faster, 50% less memory, 0% loss of accuracy Llama finetuning Hi HN! I'm just sharing a project I've been working on during the LLM Efficiency Challenge - you can now finetune Llama with QLoRA 5x faster than Huggingface's original implementation on your own local GPU. Some highlights: 1. Manual autograd engine - hand derived backprop steps. 2. QLoRA / LoRA 80% faster, 50% less memory. 3. All kernels written in OpenAI's Triton language. 4. 0% loss in accuracy - no approximation methods - all exact. 5. No change of hardware necessary. Supports NVIDIA GPUs since 2018+. CUDA 7.5+. 6. Flash Attention support via Xformers. 7. Supports 4bit and 16bit LoRA finetuning. 8. Train Slim Orca fully locally in 260 hours from 1301 hours (5x faster). 9. Open source version trains 5x faster or you can check out Unsloth Pro and Max codepaths for 30x faster training! https://ift.tt/WthiKuR... has more info about Unsloth! Hopefully you can try it out! Wrote a blog post at https://ift.tt/k2CrGqt if you want to learn more about our manual hand derived backprop or Triton kernels and stuff! Thanks once again! https://ift.tt/MyoiNQY December 1, 2023 at 08:12PM

Show HN: Block bots and spam on Instagram with SocialGuard https://ift.tt/S2p4uCd

Show HN: Block bots and spam on Instagram with SocialGuard Hey HN, I've been working on a project to combat bots and spam on Instagram. SocialGuard provides an overview of all the comments flowing through your IG account, with options to automatically delete or reply to comments based on keywords + phrases. If you use IG for work purposes, I hope it can help keep your posts clean of spam! https://ift.tt/DZxoglH December 1, 2023 at 11:11PM

Show HN: Bi-directional sync between Postgres and SQLite https://ift.tt/UCgemAH

Show HN: Bi-directional sync between Postgres and SQLite Hi HN, Today we’re launching PowerSync, a Postgres<>SQLite bi-directional sync engine that enables an offline-first app architecture. It currently supports Flutter, React Native and web (JavaScript) using Wasm SQLite in the browser, with more client SDKs on the way. Conrad and I (Ralf) have been working on our sync engine since 2009, originally as part of a full-stack app platform. That version of the system is still used in production worldwide and we’ve learnt a lot from its use cases and scaling. About a year ago we started on spinning off PowerSync as a standalone product that is designed to be stack-agnostic. If you’d like to see a simple demo, check out the pebbles widget on the landing page here: https://ift.tt/7kELyRZ We wrote about our architecture and design philosophy here: https://ift.tt/tFsoAcE... This covers amongst other things how we designed the system for scalable dynamic partial replication, why we use a server authority architecture based on an event log instead of CRDTs for merging changes, and the approach to consistency. Our docs can be found here: https://ift.tt/GpTI4fx We would love to hear your feedback! - Ralf, Conrad, Kobie, Phillip and team https://powersync.com November 30, 2023 at 07:43PM

Show HN: Status on DP – Add Status on Profile Picture Like LinkedIn https://ift.tt/dTeL37R

Show HN: Status on DP – Add Status on Profile Picture Like LinkedIn https://ift.tt/TVuUC1a November 30, 2023 at 12:34PM

Show HN: Collider – the platform for local LLM debug and inference at warp speed https://ift.tt/oj3xWlK

Show HN: Collider – the platform for local LLM debug and inference at warp speed ChatGPT turns one today :) What a day to launch the project I'm tinkering with for more than half a year. Welcome new LLM platform suited both for individual research and scaling AI services in production. GitHub: https://ift.tt/zWqTZdK Some superpowers: - Built with performance and scaling in mind thanks Golang and C++ - No more problems with Python dependencies and broken compatibility - Most of modern CPUs are supported: any Intel/AMD x64 platofrms, server and Mac ARM64 - GPUs supported as well: Nvidia CUDA, Apple Metal, OpenCL cards - Split really big models between a number of GPU (warp LLaMA 70B with 2x RTX 3090) - Not bad performance on shy CPU machines, fast as hell inference on monsters with beefy GPUs - Both regular FP16/FP32 models and their quantised versions are supported - 4-bit really rocks! - Popular LLM architectures already there: LLaMA, Starcoder, Baichuan, Mistral, etc... - Special bonus: proprietary Janus Sampling for code generation and non English languages https://ift.tt/zWqTZdK December 1, 2023 at 02:02AM

Show HN: Turn Emoji into Artwork via AI https://ift.tt/QL6bels

Show HN: Turn Emoji into Artwork via AI https://ift.tt/OveDrdx November 30, 2023 at 05:29PM