How I Ran a 100K Context LLM on an 8GB GPU (TurboQuant vs RotorQuant)
22:03
How I Ran a 100K Context LLM on an 8GB GPU (TurboQuant vs Roto…
183 views4 weeks ago
YouTubeLight The Deep
TurboQuant The algorithm that crashed RAM prices 30% Overnight
13:48
TurboQuant The algorithm that crashed RAM prices 30% Overnight
1.6K views1 month ago
YouTubeAI Depth School
TurboQuant The End of the Memory Wall 🏗️ Why Your AI Just Got 8x Faster ⚡
1:03
TurboQuant The End of the Memory Wall 🏗️ Why Your AI Just Got 8x F…
2.3K views1 month ago
YouTubePinaka Labs
Google 研究團隊發表 TurboQuant，這是一種 AI 壓縮技術，能用極致方式大幅減少大型語言模型的記憶體用量，同時完全不損失準確度。AI 模型處理文字、圖片時，會使用大量「高維度向量」，這些向量雖然強大，但非常吃記憶體，尤其在「關鍵-值快取」（KV Cache）中會造成嚴重瓶頸。傳統壓縮方法雖然能縮小尺寸，卻會產生額外記憶體負擔。TurboQuant 巧妙解決這個問題，它結合兩大創新技術：1. PolarQuant：把向量從一般座標轉成「極座標」（像用距離 + 角度來描述位置），讓壓縮更有效率，完全消除傳統方法的額外記憶體開銷。2. QJL（Quantized Johnson-Lindenstrauss）：只用 1 個位元就能修正剩餘的微小誤差，確保計算結果精準無誤，而且完全沒有額外負擔。實際測試顯示，TurboQuant 可以把 KV 快取壓縮到僅 3 位元，記憶體用量至少減少 6 倍，在 LongBench、Needle In A Haystack 等長上下文任務中，模型表現幾乎零損失。在 H100 GPU 上，注意力計算速度還能提升最高 8 倍。用在向量搜尋時，準確率
0:03
Google 研究團隊發表 TurboQuant，這是一種 AI 壓縮 …
19.3M views1 month ago
x.com駿HaYaO
Someone from uWaterloo Computer Engineering implemented Google Research’s TurboQuant as a full CUDA-native compression engine on Blackwell B200 5x KV cache compression on Qwen 2.5-1.5B, near-lossless attention scores, live generation from compressed memory, and 5 custom cuTile kernels with fused attention, online softmax, and on-chip decompression.Respect @anirudhbv_ce 🙌
4:46
Someone from uWaterloo Computer Engineering implemented Google …
789.5K views1 month ago
x.comNull
【中配】测试谷歌的 TurboQuant 方法：我得到了 5 倍的压缩效果，准确率达到了 99.5%！ - Onchain AI Garage
16:39
【中配】测试谷歌的 TurboQuant 方法：我得到了 5 倍的压缩效果，准确率 …
2.7K views1 month ago
bilibili头号个体
The End of AI Memory Bottlenecks: Breaking Down Google s TurboQuant 🚀🧠 | Revanth Reddy T
The End of AI Memory Bottlenecks: Breaking Down Google s TurboQu…
1.4K views1 month ago
linkedin.com
11:41
Quantmod R package
43.7K viewsJul 15, 2015
YouTubeMelvin L
15:09
Quant Finance with R Part 1: Intro and Data
40.7K viewsDec 16, 2018
YouTubecodebliss
12:31
Quantile Regression in R
10.2K viewsMay 17, 2021
YouTubeJustin Eloriaga
12:06
QuantLib Python Cookbook
100 viewsJun 16, 2016
leanpub.com
11:50
Introduction to Quantlib part 1 Build up an Option
4.8K viewsNov 16, 2015
YouTubeCarol Zheng
30:05
Qlib开源AI量化框架
6.2K viewsAug 5, 2023
YouTube刘英杰 Yingjie
7:14
[Detailed Explanation] Google TurboQuant: Achieving Ultimate Z…
222 views1 month ago
YouTubeAI Learning Notes
17:02
QuantLib 入门 Day01 | 零息债券与固定利息债券定价实战（Python 实 …
72 views8 months ago
YouTube我在学量化
7:40
QuantMod in R Part 1
4.2K viewsJul 3, 2018
YouTubeEd Boone
4:58
R - quantmod - download data
2K viewsJul 31, 2019
YouTubeChiu Yu Ko
6:32
R Programming: r volatility with quandl
9K viewsJul 16, 2015
YouTubeBionic Turtle
2:17
13. Computing Quantiles in R
10.9K viewsJun 16, 2018
YouTubeData Science Materials
18:21
Computational Finance in R: quantmod-package
6.7K viewsJun 15, 2020
YouTubeJan-Hendrik Meier
5:41
TQuant Lab 系統入門教學 - TQuant Lab 安裝
376 viewsOct 13, 2023
YouTubeTEJ台灣經濟新報
16:53
Getting Data For Multiple Assets (With Calculations): R For Finance
8K viewsSep 28, 2017
YouTubecodebliss
10:36
QuantLib notebooks: interest-rate sensitivities
5.1K viewsMar 15, 2016
YouTubeLuigi Ballabio
14:44
QuantLib C++ on Ubuntu 22.04 (AWS EC2 Setup)
1 views6 months ago
YouTubeAnalytica Tv
18:03
Introduction to QuantLib. Part 8d: InterestRate, YieldCurve and Stoc…
1.5K viewsFeb 14, 2021
YouTubeeefelix
42:36
XRP & QNT Bombshell: How Quant's Programmability Will Ben…
4.4K views1 month ago
YouTubeApex Crypto Insights (NFA)
27:27
【課程試看】Python-QuantLib套件金融計算應用：Part I 期貨與現貨套 …
905 viewsMay 29, 2023
YouTubeMasterTalks線上課程
14:08
QuantX Reverb by Relab Developement (No Talking)
6.7K viewsNov 18, 2023
YouTubeElektronick Musick
7:23
TurboQuant: Redefining AI Efficiency with Extreme Compress…
8.7K views1 month ago
YouTubeResearch Paper Review
12:42
Easily Import Financial Data to R with Quantmod
6.9K viewsDec 20, 2021
YouTubeData Garden
See more videos
Static thumbnail place holder