Tootfinder

Opt-in global Mastodon full text search. Join the index!

No exact results. Similar results found.
@netzschleuder@social.skewed.de
2025-12-26 07:00:03

contiguous_usa: Contiguous states (USA)
A network of contiguous states in the USA, in which each state is a node and two nodes are connected if they share a land-based geographic border. The dataset includes the lower 48 states, and the District of Columbia.
This network has 49 nodes and 107 edges.
Tags: Transportation, Roads, Unweighted

contiguous_usa: Contiguous states (USA). 49 nodes, 107 edges. https://networks.skewed.de/net/contiguous_usa
@underdarkGIS@fosstodon.org
2025-12-26 18:50:08

Finally found some time to look into the QGISMCP plugin.
The original connects #QGIS :qgis: to #Claude through the Model Context Protocol #MCP. I rewired it to work with a local European alternative: the new m…

@arXiv_csLG_bot@mastoxiv.page
2026-02-25 10:45:11

Untied Ulysses: Memory-Efficient Context Parallelism via Headwise Chunking
Ravi Ghadia, Maksim Abraham, Sergei Vorobyov, Max Ryabinin
arxiv.org/abs/2602.21196 arxiv.org/pdf/2602.21196 arxiv.org/html/2602.21196
arXiv:2602.21196v1 Announce Type: new
Abstract: Efficiently processing long sequences with Transformer models usually requires splitting the computations across accelerators via context parallelism. The dominant approaches in this family of methods, such as Ring Attention or DeepSpeed Ulysses, enable scaling over the context dimension but do not focus on memory efficiency, which limits the sequence lengths they can support. More advanced techniques, such as Fully Pipelined Distributed Transformer or activation offloading, can further extend the possible context length at the cost of training throughput. In this paper, we present UPipe, a simple yet effective context parallelism technique that performs fine-grained chunking at the attention head level. This technique significantly reduces the activation memory usage of self-attention, breaking the activation memory barrier and unlocking much longer context lengths. Our approach reduces intermediate tensor memory usage in the attention layer by as much as 87.5$\%$ for 32B Transformers, while matching previous context parallelism techniques in terms of training speed. UPipe can support the context length of 5M tokens when training Llama3-8B on a single 8$\times$H100 node, improving upon prior methods by over 25$\%$.
toXiv_bot_toot

@netzschleuder@social.skewed.de
2025-11-26 14:00:04

sp_high_school_new: High school dynamic contacts (2011-2012)
These datasets contain the temporal network of contacts between students in a high school in Marseilles, France. The first dataset gives the contacts of the students of three classes during 4 days in Dec. 2011, and the second corresponds to the contacts of the students of 5 classes during 7 days (from a Monday to the Tuesday of the following week) in Nov. 2012.
This network has 126 nodes and 28561 edges.
Tags: Soc…

sp_high_school_new: High school dynamic contacts (2011-2012). 126 nodes, 28561 edges. https://networks.skewed.de/net/sp_high_school_new#2011
@Techmeme@techhub.social
2026-02-24 15:30:54

Kraken launches what it claims are the first regulated perpetual futures contracts based on tokenized stocks, available to non-US users and trading 24/7 (Krisztian Sandor/CoinDesk)
coindesk.com/business/2026/02/

@inthehands@hachyderm.io
2026-02-26 03:07:45

…sometimes there’s •not• a clear standard way. Sometimes you need flexibility. For example, a lot of what makes UI programming hard is layout: you have to make your own very specific application look good on a variety of devices and screens, which means coming up with an •algorithm• for adjusting your design for all those different contexts.
That’s an intrinsically hard problem that requires design chops and nuance and contextual knowledge. Attempts to abstract the decisions out of that problem have been stubbornly unsuccessful. (How many layout engines are out there now?)

@Techmeme@techhub.social
2026-02-24 15:06:57

Sources: Meta aims to enter the stablecoin space in H2, with plans to integrate a vendor to administer stablecoin-backed payments and implement a new wallet (Ian Allison/CoinDesk)
coindesk.com/business/2026/02/

@netzschleuder@social.skewed.de
2026-02-25 00:00:05

sp_office: Workplace contacts (2013)
A temporal network of contacts between individuals, measured in an office building in France, from June 24 to July 3, 2013.
This network has 92 nodes and 9827 edges.
Tags: Social, Offline, Weighted, Temporal, Metadata
networks.skewed.de/net/sp_offi

sp_office: Workplace contacts (2013). 92 nodes, 9827 edges. https://networks.skewed.de/net/sp_office
@Techmeme@techhub.social
2026-01-24 19:10:52

Nifty Gateway, the NFT trading platform which had facilitated $300M in sales by mid-2021, enters withdrawal-only mode and plans to shut down on Feb. 23, 2026 (CoinDesk)
coindesk.com/business/2026/01/

@netzschleuder@social.skewed.de
2026-02-26 19:00:04

sp_kenyan_households: Kenyan households contacts (2012)
A network of proximity contacts measured between members of 5 households of rural Kenya, between April 24 and May 12, 2012.
This network has 47 nodes and 32643 edges.
Tags: Social, Offline, Unweighted, Temporal, Metadata
networks.skewed.…

sp_kenyan_households: Kenyan households contacts (2012). 47 nodes, 32643 edges. https://networks.skewed.de/net/sp_kenyan_households