- Joined
- Aug 23, 2019
They are just so easy to update and access from different devices because browser developers handle compatibility for them.
Programming thread
- Thread starter Splendid
- Start date
- Joined
- Oct 19, 2023
If only we lived in a perfect world where shit like GTK worked on phones too.
It's technically possible...
- Joined
- Feb 24, 2023
I've been trying constantly to learn vanilla javascript MVC but I just cant do it. I feel like im getting burned out. How do you guys avoid getting burn out when your learning something hard? Also what would you do if you were in my situation?
- Joined
- Jul 29, 2021
take breaks and break up things in manageable chunks.
- Joined
- Jul 6, 2020
Bounce between multiple projects to keep fresh
- Joined
- Aug 2, 2021
Worked with HR people.
Everyone was doing weird stuff. Eg people would not paste without formatting; instead they would paste to notepad then copy-paste to the actual document.
The reason? "It just doesn't work sometimes"? They never bothered to diagnose these things. It just doesn't work I guess, the computer must be stupid...
granted, google docs, libra and word all have different way to paste without formatting, and word online is completely different from word on your pc, but still
prodigylogic
kiwifarms.net
- Joined
- Jul 4, 2022
I work mainly with a WinForms/WPF thick client, so it's necessary, and since most of the thick client's libraries are targeted to the .NET standard, they are reusable for Xamarin apps. It makes it an easy choice for our thick client on our mobile apps.
The one thing I like about XAML is that the markup and styling is done in the same file; I've never liked switching between three different file types to develop one page. I've also noted that if I'm strict with the MVVM architecture pattern, development with XAML frameworks produces clean first runs more often than with other frameworks. Designing with this pattern has a mean learning curve; though, and I've seen people quit their jobs before learning it. For MAUI specifically, I think it beats out react-native. You get a lot of reusability if you work things right, and if the demand is that you have support for web, desktop, and mobile platforms.
That said, it's not worth learning for web development. Some people still use WebForms or Silverlight, but you don't want to stick your bets on dying horses. Blazor looks very promising for web development, and it uses a mixture of C# and HTML that I find very agile.
- Joined
- Mar 10, 2019
If you copy some data from Excel, and you want to paste that data into Excel, as values, and some of the cells that you copied had formulas that used
="" to be blank, those blank cells are not actually empty cells as far as charts are concerned (for the purpose of treating empty cells as "gaps" or "connect data points with line"). The chart pretends that they contain zeros.But, if you paste to Notepad then copy-paste to the actual document, it works perfectly.
And yeah, I could use
=NA(), and then turn on "Show #N/A as an empty cell", but then the data would need to have a bunch of #N/A errors in it, instead of empty cells, and that would be fucking retarded.
- Joined
- Aug 2, 2021
- Joined
- Aug 23, 2019
I'm guilty of that, I like using notepad++ to sanitise and standardise things.
sadrefrigerator
kiwifarms.net
- Joined
- Jul 5, 2022
What is "vanilla JavaScript MVC"
You would use vanilla JavaScript for your whole stack. You shouldn't use vanilla js for much of anything.
- Joined
- Mar 28, 2023
Doing ML shit is making me really despise Python, I'm not going to lie.
- Joined
- Feb 8, 2021
The package situation may not be as bad as in Javascript, but by God they're doing their best to change that.
- Joined
- Mar 28, 2023
The libraries are just incredibly overbearing. They do not expose enough of the codebase effectively, so it's just a black box. Additionally, getting this shit to actually utilize my GPU has been a task and a half
- Joined
- Jun 30, 2023
I remember pasting copied portions Wikipedia articles into the URL bar of the Internet Explorer to clear formatting back when I was a kid doing school projects in Word. They had me fixing the few old computers we had when I was in elementary school. I can't think of a single instance where I would genuinely be helped by the copied formatting from the browser.
I have said it before, perhaps even in this thread, but I am convinced that ML researchers write the most painful to read Python in existence. I think it may be due to a lot of them being statsfags who write barely working shit in R. My former colleagues when I was still doing research wrote code that was so bad, I had to insist they stop asking me for advice on how to fix it.
Aside from that, Pytorch uses a lot of function passing which makes it difficult to debug stuff like HF libraries without a high res monitor. This isn't necessarily bad practice, but it's painful to deal with when you just want the shit you write to work like it should.
Edit: I forgot about the reply directly above this. Since I can't quote, please bear with me as I work to avoid double posting:
I am torn as to whether the culture of one-liner solutions to inference/training/general model utilization is a good thing. While it helps beginners get into a very powerful field of computer science, I worry that it will further perpetuate the increasingly evident societal issue of average people not knowing how to do anything. Doing things the long way allows for more customization of the process you are using, but most beginners don't care about that. The only way to find the more detailed information is from the people who wrote the libraries themselves, while those who are learning from third-party sources will become well acquainted with the idea that programming requires little to no thought on the end user's part.
Last edited:
- Joined
- Jun 7, 2020
I have horror stories of working with OpenAI gym models that I have to convert to Gymnasium or Stable Baselines code into Stable Baselines 3 code that it makes me want to never touch the damn thing ever again.
- Joined
- Jan 15, 2023
I don’t get the love for VS Code. Sure most of it is coming from front end kids, but it sits at the exact worst point between the “customize everything yourself” of a terminal editor like vim or emacs, and the specialization + power provided by PyCharm or some other full featured IDE. Also I’ve seen it crash trying to search repos that grep handles just fine.
The average ML practitioner will give you deployment instructions that require you to run eight shell scripts in a row and think it’s easy and self-evident. They haven’t yet realized that you can combine them into a single shell script and run that, with proper logging, error checking and everything else you’d want.
The average ML practitioner will give you deployment instructions that require you to run eight shell scripts in a row and think it’s easy and self-evident. They haven’t yet realized that you can combine them into a single shell script and run that, with proper logging, error checking and everything else you’d want.
- Joined
- Jul 6, 2021
I will never use a IDE written in Electron ever. VS Code is complete garbage and somehow has worse performance than Atom (RIP) despite being supported by Microsoft. Sublime Text will still remain objectively the best lightweight text editor and it's all you need for most projects.
The problem with Python for machine learning is not Python itself but the dependency hell it creates. I remember one time having 10 different Tensorflow packages installed across virtualenvs because each ML project had different tensorflow package requirements.
- Joined
- Jan 17, 2021
Heh that brings back memories from uni days. Half the struggle of learning basic neural net models was getting the damn code to use the GPU and not the CPU. Did admittedly get fun watching the Apple fags attempt to do that on uni computers other students had already screwed with.
Can relate to this as well. Part of it I think is people using it without a firm grasp of programming fundamentals. ML is the flavour of the day so a lot of tangential fields are throwing it into thesis proposals without appreciating the underlying work required; they might have a strong grasp of the theory going into ML, but they lack the skills to put it into practice effectively. The result is kiddies whose whole programming experience might be simple R or Matlab scripts trying to build a legitimate program.
- Joined
- Mar 28, 2023
I know this is kind of a lame thing to get hung up on, but this AutoML tech really eliminates any "sexiness" from making models. you spend more time loading a dataframe than actually tweaking anything.