Most ML development is done in Python to my knowledge. The reason is mostly the readily available side packages like torch, scikit-learn, etc. And Python, although not constructed as such, does indeed support a functional style. A quick search revealed a HowTo:
- 1 Post
- 21 Comments
I dunno. P2P traffic always seems to overburden light users and it would indeed require the apps to always run in the background to relay the traffic. Although the idea seems compelling I wouldn’t install the software on a machine of mine.
Appreciate the KISS perspective.
For me, the project management features of a forge are extremely helpful. Setting milestones, assigning issues to them, defining timelines and regularly reiterating the planning has proven to accelerate our work as a team significantly. This experience refers to huge code bases (climate models) and medium to large team sizes, though. And probably also my bad memory 😵💫
I suppose it’s always good, though, to evaluate how much management a code will actually need in the end, and what tools correspond to that need.
gsvOPto Selfhosted@lemmy.world•GDPR compliant / European SMTP relay for self hosted services?English1·2 days agoThat’s a great recommendation, thanks! The pay per use model seems very fair and I like their approach to sustainability matters very much.
gsvOPto Selfhosted@lemmy.world•GDPR compliant / European SMTP relay for self hosted services?English4·3 days agoThanks! Will have a look
While I totally agree that this should be the case, I’m not sure it really works. Voluntary participation is among the first things to be cut when it comes to monetary gain maximization, and is often not even considered. And in some instances, like the publicly funded research institute I work at, there’s no funds dedicated to voluntary contribution to open source projects.
The article is unfortunately a stub. The original insight stems from the peer reviewed report of Tannu and Nair (2022) and is based on a meta analysis of several life cycle analyses. It’s actually cited and linked in the Seagate brochure:
https://hotcarbon.org/assets/2022/pdf/hotcarbon22-tannu.pdf
Long story short: The reason a combination of the higher production related CO2e emissions and the higher power consumption given the current power mix.
Using it for years and am fully satisfied. Basically, any imap client capable of encryption will do.
Schleswig-Holstein therefore follows the general strategy to move towards an open source driven administration. In fact, several federn institution already migrated to the openDesk administration bundle (https://gitlab.opencode.de/bmi/opendesk/). Great!
gsvto Privacy•UK creating ‘murder prediction’ tool to identify people most likely to killEnglish1·15 days agoExactly my thought. Or the beginning of psycho-history?
Checked again. Beeper will be running “local bridges”. Once these are implemented into a modular system, there is not really a need for a local chat protocol in my understanding. The matrix connection can simply become another bridge and a Beeper server for Beeper chats would basically be yet another matrix server.
https://blog.beeper.com/2024/06/04/2024-beeper-roadmap/?ref=textscom
The FAQ says that “all integrations were implemented in-house using the Texts Platform SDK”. Whether that sdk is a derivative of the Matrix protocol? No idea. Texts.com does not offer connections to matrix, which kind of suggests it’s not 🤷🏼♂️ We will have to see whether the announced unified app will be running a solution based on Matrix or not.
Not exactly what was asked but a thought as I’ve been considering the same. After merging with Texts.com, Beeper seems to be redesigning the bridge architecture. I read that the implementation will move towards running the bridges on the client device so that the decryption is happening in the RAM of the end user’s machine rather than the server. In that case, the mentioned security problem will be at least partially resolved. Self-hosting the bridges is already possible now. One will still have to trust Beeper, though. As I am using their software already, it looks like there’s no reason, yet, to mistrust them.
There’s many pathways to get there. In my case it’s a classic science career. Whether that is better for the WLB is likely open for debate. 🙈 There’s a growing understanding in science that there’s a need for tenured scientific programmers supporting the compute infrastructure and development. These roles are quite rare though and thus there is tons of competition. Long story short: It’s a challenge. The good thing is that there’s many jobs out there with the meaning we seek.
I’m involved scientific programming tasks (climate models) and we are largely using a mix of prototyping, feature-driven, and big bang models. It’s a result of the requirements for our work. Important is, that our “user base” is extremely small and in the beginning of a project mostly ourselves. The required features are given by the scientific questions and timelines dictated by project timelines from soft funding. Iterations are thus mostly more like “that didn’t quite work, I need another method” kind of arguments. Hence, the implementation of modular and fast evolving design is important and often we try to build our software such that fast development can be done on individual models.
Sometimes theres an idea on how to solve a problem more efficiently or with better performance. And then it’s all about coming up with any proof of concept.Finally, there is not really anything like a management in our line of work. The professor is usually satisfied if the results work out and papers go out quickly. The rest ist largely up to the group of devs.
The initiative to build FOSS alternatives for the administration in Europe goes back years. In a nutshell: Corporate software is getting very expensive for the administration and poses security risks. As an alternative, FOSS Software is put together to replace the administrative systems. The reason there is so many news now is that the first Software bundles are being released and are coming online in databases.
To give that more context: The BundesMessenger is developed for secure communication in the German army (named Bundeswehr in German). So it’s most likely not only about cash but also about security. It’s e2e encrypted, can be federated to several locations, etc. Seems a smart move to me. Edit: Actually I’m mistaken. It’s for the whole public administration. Not only the army. Not sure how I got that impression. Argument holds, though. 🙈
So and so. The projects I like a lot and heavily use I give 5€ a month, but there are very few. That would also be the usual member fee for instance for the Wikimedia foundation. Beyond that, it’s usually between a coffee and 20€, depending on how much potential I see, how much it has helped me, and how much the project needs the money. Maybe needless to say: Bug reports help a lot so I place them where I find one. Finally, when it comes to code contributions I find it hard. Usually that requires a lot of preparation and time to get into the depth of the project beyond the API/fronted. And that is, more often than not, a road block for me.
I think though, that the amount given is not a good measure on its own. It depends a lot on what one has and is able to give. There are so many people out there who earn way more than they need (including me, although I am nowhere rich). They could easily give a lot more and cover for a general user base. And the latter, making the project more popular and maybe contributing some bug reports is just as essential.
gsvto Programming•Anthropic can now track the bizarre inner workings of a large language modelEnglish11·28 days agoFor some reason I don’t find it very bizarre. I’d even speculate that a random human mind isn’t any less weird. Surly, the pathways of my thoughts are often very bizarre. 😅
I’m not sure I understand this, tbh. Does that mean the P2P network works on a chat group basis? Or does the user explicitly choose who to build a P2P network with? And then, there are lots of follow up questions already looking around the corner.
Their website seems to explain very little and the app itself is closed source. Although there are open source dependencies, it’s for instance unclear whether they are complete. So I guess it’ll all come down to trust into the software and the company. Btw. their privacy statement allows the usage of aggregate data for marketing purposes and the sharing of data with third parties, such as search engines. And latest at that point I’d rather self-host a matrix instance.