Rebeauty Atlas는 무엇인가요?

Rebeauty Atlas는 AI·디자인·마케팅·경제·의료 등 다양한 주제의 Medium 아티클을 수집해 태그·트랙·위키·인사이트로 탐색할 수 있게 정리한 리서치 아카이브입니다.

글은 어떻게 모이나요?

Rebeauty Atlas는 Medium에 공개된 아티클을 자동 수집해 AI·디자인·마케팅·경제·의료 등 주제별 태그·트랙·위키로 정리하는 리서치 아카이브입니다.

사이트는 어떤 섹션으로 구성되어 있나요?

Articles · Wiki · Insights · Tags 섹션으로 구성되어 있으며, PHP MVC · MySQL · shadcn-style UI · Wiki via keyword rules 로 만들어졌습니다.

Articles · Rebeauty Atlas

AI HUM TCH

AN Anubhav · Towards AI

What Is Inference Engineering? The Layer Doing 80% of Your LLM Bill.

FP8 KV cache, prompt caching, quantization, speculative decoding, MoE routing. The five 2026 levers between your model and your invoice…

Jul 16, 2026 · 10 min read · 154 claps paywalled #machine-learning #artificial-intelligence #programming

What Is Inference Engineering? The Layer Doing 80% of Your LLM Bill.

AI HUM

VI Vishal Rajput · AIGuys

NVIDIA Proved 4-Bit Training Works at Real Scale (Not Just Inference)

A 12-billion-parameter Mamba-Transformer, trained on 10 trillion tokens, entirely in 4-bit floating point, matched an FP8 baseline almost…

Jun 19, 2026 · 12 min read · 2 claps paywalled #machine-learning #technology #deep-learning

NVIDIA Proved 4-Bit Training Works at Real Scale (Not Just Inference)

AI

JA Jatinkumar Parmar

FP8 GEMM optimization on AMD CDNA4-based GPUs

Jun 19, 2026 · 64 min read · 0 claps paywalled #fp8-gemm #amd #gpu

SPT MDA AI

MK M K Pavan Kumar

Compressing Gemma 4 12B with LLM Compressor and Serving It on vLLM

This is the full journey of taking Gemma 4 12B from a 23.9GB BF16 model to a 15GB FP8 one that serves on a single GPU — start to finish…

Jun 13, 2026 · 14 min read · 3 claps #llm-compression #vllm #large-language-models

Compressing Gemma 4 12B with LLM Compressor and Serving It on vLLM

AI SPT

OM OmarEbnElKhattab Hosney

Running Microsoft’s Lens on a Single RTX 4090: Quantizing the DiT to Hit 1440 Without Offload

How a one-line FP8 quantization of the denoiser turned “1440 needs CPU offload” into “1440 runs fully on the GPU” on a 24 GB card.

Jun 08, 2026 · 8 min read · 3 claps paywalled #llm #text-to-image-generation #microsoft-lens

Running Microsoft’s Lens on a Single RTX 4090: Quantizing the DiT to Hit 1440 Without Offload

AI SPT

AD Adnan Strujic

Hermes Agent on DGX Spark running Qwen3.6–35B-A3B-FP8 model on vllm server and Telegram messages

Step by step introduction and configuration to Hermes Agent craze

May 14, 2026 · 5 min read · 28 claps paywalled #hermes-agent #dgx-spark #qwen

Hermes Agent on DGX Spark running Qwen3.6–35B-A3B-FP8 model on vllm server and Telegram messages

AI TCH

AN Ant Ling

FP8 Quantization for Large Model Training: Performance Optimization and Analysis

Background & TL;DR

Oct 12, 2025 · 7 min read · 3 claps #foundation-models #llm #open-source

FP8 Quantization for Large Model Training: Performance Optimization and Analysis

AI

NI Nikulsinh Rajput

FP8 + Structured Sparsity: Spend Less, Serve More

Combine 8-bit floating point with 2:4 sparsity to cut GPU memory and flops — while keeping accuracy and p99s in check.

Oct 01, 2025 · 5 min read · 15 claps paywalled #gpu #fp8 #structured

FP8 + Structured Sparsity: Spend Less, Serve More

HUM AI SPT

ST StackGpu

FP8, BF16, and INT8: How Low-Precision Formats Are Revolutionizing Deep Learning Throughput

Explore how FP8, BF16, and INT8 are transforming deep learning performance by optimizing throughput, reducing energy consumption, and…

Jun 20, 2025 · 7 min read · 6 claps #fp8 #bf16 #int8

FP8, BF16, and INT8: How Low-Precision Formats Are Revolutionizing Deep Learning Throughput

HUM AI

AD Adnan Masood, PhD.

Bits and Business: FP8 quantization deepdive into DeepSeek’s High-Precision, Low-Bit Revolution

Tim Dettmers of Making deep learning accessible, said something really profound and beautiful.

Feb 20, 2025 · 24 min read · 5 claps paywalled #fp8 #quantization #deepseek

Bits and Business: FP8 quantization deepdive into DeepSeek’s High-Precision, Low-Bit Revolution

AI

SH Shankar Shivamurthy

How DeepSeek R1, Using FP8 Instead of FP32, Beat OpenAI, Meta, Gemini, and Claude?

Why FP8 matters ?

Jan 28, 2025 · 3 min read · 20 claps #deepseek #artificial-intelligence #generative-ai-tools

How DeepSeek R1, Using FP8 Instead of FP32, Beat OpenAI, Meta, Gemini, and Claude?

HUM AI

FR FriendliAI Tech & Research · FriendliAI

Improve Latency and Throughput with Weight-Activation Quantization in FP8

Quantization is a popular technique used to reduce the size of a machine learning model by lowering the numerical precision of some of its…

Apr 15, 2024 · 5 min read · 105 claps #quantization #fp8

Improve Latency and Throughput with Weight-Activation Quantization in FP8

JA Jack omnixri

Nvidia GTC 2024 提出的 FP8/FP4 如何加速AI訓練及推論

作者：許哲豪(Jack Hsu), 2024/3/24

Mar 23, 2024 · 5 min read · 0 claps #gtc-2024 #b200 #fp8

AI HUM

1K 1kg

FP8: Deep Learning’s New Power Tool

Let’s take a journey into the realm of artificial intelligence, specifically deep learning. Here, we have immense structures, called neural…

Jul 28, 2023 · 3 min read · 3 claps #fp8 #deep-learning #ai