📊 Machine Learning - tyler · Scour

You can put a Softmax in front of CrossEntropyLoss. PyTorch won’t stop you. Here are 16 other architecture bugs it won’t catch. 🐛Fuzzing

gaox.substack.com·3d·Substack

Show HN: Dataset for AI training and fine tuning 🧠AI

neurvance.com·2d·Hacker News

zero-intelligence/zero-intel: Every codebase has a confession. Most people never ask it the right question. 🔧Code Generation

github.com·13h·Hacker News

InferenceBench: A Benchmark for Open-Ended Inference Optimization by AI Agents 📊LLM Evals

inferencebench.ai·5h·Hacker News

How Snapchat Serves a Billion Predictions Per Second 💬Prompt Engineering

blog.bytebytego.com·1d

Andrej Karpathy Joined Anthropic. What It Says About Where AI Is Heading. 🧠AI

firethering.com·20h·Hacker News

Issue 651 🗂️Vector Databases

datascienceweekly.substack.com·6d·Substack

Running PyTorch Models on Apple Silicon GPUs with the ExecuTorch MLX Delegate 🦙Local LLM

pytorch.org·2d·Hacker News

Slax Reader CLI: a read-later library your AI agents can use 💬Prompt Engineering

slax.com·22h·Hacker News

How to Clean Time Series Data in Python 🐍Python

freecodecamp.org·2d·Hacker News

Mathlib Initiative: Roadmap ⌚Quantified Self

mathlib-initiative.org·5d·Hacker News

End of the semester 🔧Code Generation

mtendekuyokwa.github.io·1d·Lobsters, Hacker News

The Prehistory of A.I. Slop 🖥️Computing History

longreads.com·17h

noahgolmant/pytorch-hessian-eigenthings: Efficient PyTorch Hessian eigendecomposition tools! 🗂️Vector Databases

github.com·6d·Hacker News

LoRA and Weight Decay (2023) 📊LLM Evals

irhum.github.io·2d·Hacker News

AI-aided code migration: Google got 6x faster migration from TensorFlow to Jax 🤖Creative Automation

cloud.google.com·6d·Hacker News

Google Tensor SDK Beta with LiteRT 🧠AI

developers.googleblog.com·1d

AI Function calling in the browser, 258 KB WASM 🧠AI

needle-rs.pages.dev·2d·Hacker News

Benchmarking Subquadratic's latest model and SSA Kernel ⚙️Performance Profiling

appen.com·6d·Hacker News

Runway News | Why Distributed Training Is Hard: DTensor, Correctness and the Costs of Abstraction ⚡SIMD

runwayml.com·2d·Hacker News

Log in to enable infinite scrolling