Technische Forensiek - Daan & Team: Nefit Bosch
Nefit Bosch Compress 5800i: Storing F32 Persgastemperatuursensor
#NefitBosch #Compress5800i #F32
🔍 Volledig Rapport: https://www.wpstoring.org/nefit-bosch/compress-5800i/nefit-bosch-compress-5800i-storing-f32-persgastemperatuursensor
System Diagnostic by Marcus: EG4
EG4 18KPV: Fault Code F32 - M3 Microprocessor Rx Failure
🔍 Full Report: https://www.storagefaults.com/eg4/18kpv/eg4-18kpv-fault-code-f32-microprocessor-rx-failure
Possibly-Smallest ESP32 Board uses Smallest-Footprint Parts.
Whenever there’s a superlative involved, you know that degree of optimization has to leave something else on the table. In the case of [PegorK]’s f32, the smallest ESP32 dev board I’ve seen, the cost of miniaturization is GPIO.
https://github.com/PegorK/f32?tab=readme-ov-file
#f32 #miniaturization #esp32 #dev #board #diy #engineering #media #tech #art #news
F32 – An Extremely Small ESP32 Board
#HackerNews #F32 #ESP32 #Small #Board #IoT #Technology #Electronics #Innovation
OLED Rückleuchten für den BMW| EINBAU | BMW F36
https://themotorbikechannel.com/oled-ruckleuchten-fur-den-bmw-einbau-bmw-f36/?feed_id=14256&_unique_id=671045b82395e
Source: OLED Rückleuchten für den BMW| E...
#3er #418d #420d #420i #425d #425i #430d #430i #435i #440i #4er #anleitung #auto #autoteile #bmw #BMWoled #dm #Einbau #einbauen #erfahrung #f30 #f31 #f32 #f33 #f34 #f36 #oled #reperatur #review #rückleuchte #vland #wechseln
Source: OLED Rückleuchten für den BMW| EINBAU | BMW F36 by Mazo Vlogs. Please don’t forget to give the Video a “Like” on Youtube and subscribe to the channel! Wenn du Lust auf mein Equipment, Merch oder Sozial Media Seiten hast: 🔥 https://linktr.ee/mazovlogs 🔥
Efficient $1$-bit tensor approximations
Alex W. Neal Riasanovsky, Sarah El Kazdadi
https://arxiv.org/abs/2410.01799 https://arxiv.org/pdf/2410.01799 https://arxiv.org/html/2410.01799
arXiv:2410.01799v1 Announce Type: new
Abstract: We present a spatially efficient decomposition of matrices and arbitrary-order tensors as linear combinations of tensor products of $\{-1, 1\}$-valued vectors. For any matrix $A \in \mathbb{R}^{m \times n}$, $$A - R_w = S_w C_w T_w^\top = \sum_{j=1}^w c_j \cdot \mathbf{s}_j \mathbf{t}_j^\top$$ is a {\it $w$-width signed cut decomposition of $A$}. Here $C_w = "diag"(\mathbf{c}_w)$ for some $\mathbf{c}_w \in \mathbb{R}^w,$ and $S_w, T_w$, and the vectors $\mathbf{s}_j, \mathbf{t}_j$ are $\{-1, 1\}$-valued. To store $(S_w, T_w, C_w)$, we may pack $w \cdot (m + n)$ bits, and require only $w$ floating point numbers. As a function of $w$, $\|R_w\|_F$ exhibits exponential decay when applied to #f32 matrices with i.i.d. $\mathcal N (0, 1)$ entries. Choosing $w$ so that $(S_w, T_w, C_w)$ has the same memory footprint as a \textit{f16} or \textit{bf16} matrix, the relative error is comparable. Our algorithm yields efficient signed cut decompositions in $20$ lines of pseudocode. It reflects a simple modification from a celebrated 1999 paper [1] of Frieze and Kannan. As a first application, we approximate the weight matrices in the open \textit{Mistral-7B-v0.1} Large Language Model to a $50\%$ spatial compression. Remarkably, all $226$ remainder matrices have a relative error $<6\%$ and the expanded model closely matches \textit{Mistral-7B-v0.1} on the {\it huggingface} leaderboard [2]. Benchmark performance degrades slowly as we reduce the spatial compression from $50\%$ to $25\%$. We optimize our open source \textit{rust} implementation [3] with \textit{simd} instructions on \textit{avx2} and \textit{avx512} architectures. We also extend our algorithm from matrices to tensors of arbitrary order and use it to compress a picture of the first author's cat Angus.
We present a spatially efficient decomposition of matrices and arbitrary-order tensors as linear combinations of tensor products of $\{-1, 1\}$-valued vectors. For any matrix $A \in \mathbb{R}^{m \times n}$, $$A - R_w = S_w C_w T_w^\top = \sum_{j=1}^w c_j \cdot \mathbf{s}_j \mathbf{t}_j^\top$$ is a {\it $w$-width signed cut decomposition of $A$}. Here $C_w = "diag"(\mathbf{c}_w)$ for some $\mathbf{c}_w \in \mathbb{R}^w,$ and $S_w, T_w$, and the vectors $\mathbf{s}_j, \mathbf{t}_j$ are $\{-1, 1\}$-valued. To store $(S_w, T_w, C_w)$, we may pack $w \cdot (m + n)$ bits, and require only $w$ floating point numbers. As a function of $w$, $\|R_w\|_F$ exhibits exponential decay when applied to #f32 matrices with i.i.d. $\mathcal N (0, 1)$ entries. Choosing $w$ so that $(S_w, T_w, C_w)$ has the same memory footprint as a \textit{f16} or \textit{bf16} matrix, the relative error is comparable. Our algorithm yields efficient signed cut decompositions in $20$ lines of pseudocode. It reflects a simple modification from a celebrated 1999 paper [1] of Frieze and Kannan. As a first application, we approximate the weight matrices in the open \textit{Mistral-7B-v0.1} Large Language Model to a $50\%$ spatial compression. Remarkably, all $226$ remainder matrices have a relative error $<6\%$ and the expanded model closely matches \textit{Mistral-7B-v0.1} on the {\it huggingface} leaderboard [2]. Benchmark performance degrades slowly as we reduce the spatial compression from $50\%$ to $25\%$. We optimize our open source \textit{rust} implementation [3] with \textit{simd} instructions on \textit{avx2} and \textit{avx512} architectures. We also extend our algorithm from matrices to tensors of arbitrary order and use it to compress a picture of the first author's cat Angus.
My #F32 resub application to study #TCR repertoire changes after #AAV #genetherapy was...
Not discussed 🫠 #immunology
Me closing eRA commons:
Daan | Analyse-Lead & Team 🇳
Marcus | Storage Expert 🇺🇸
Olly 👾
N-gated Hacker News
Hacker News
Marcin
The Motorbike Channel
arXiv math.CO bot
McDonald_69
Mark Brimble