This was… a lot easier than I expected…? It was obviously a good idea to back up all my files, make sure I had access to all my most important accounts on my phone, and make sure that I had the latest version of Mint Xfce on the bootable USB, but I was so worried about doing something slightly wrong and breaking my computer that I just feel kinda silly now that I know everything just kinda works.

Well, everything except two things: firstly, it’s gonna be a pain to set up my custom Russian keyboard; secondly, RVC seems like a pain in the ass to install, but we’ll see if it’s the type of pain in the ass I can deal with in the span of six days, and it’s no huge deal if it isn’t.

Edit: I had some trouble getting it to connect to my TV via HDMI. The solution turned out to be to switch back to the open source display driver instead of Nvidia’s proprietary one LMFAO

Edit 2: Tumblerd kept me from safely ejecting my external hard drive; I ended up just terminating Tumblerd through the task manager, and that fixed the problem. Tumblerd is evidently just a thingy that generates thumbnails for files and it got stubck because it was a big hard drive with hundreds of video files.

Edit 3: I’ve got trackpad gestures working more or less like they did on Windows. Hooray.

Edit 4: I’ve got the Japanese IME (Mozc) set up and added em dash and interrobang among a few other special characters to the custom dictionary.

  • PorkrollPosadist [he/him, they/them]@hexbear.netEnglish
    16·
    25 days ago

    spongebob-party

    Edit:

    RVC seems like a pain in the ass to install, but we’ll see if it’s the type of pain in the ass I can deal with in the span of six days, and it’s no huge deal if it isn’t.

    Give it a shot. Make another post if you run into trouble. I’m too lazy to machine-translate the instructions, but it looks like it’s a ‘fairly basic’ python install process.

    'fairly basic' possibly TMI

    Most distributions go through great lengths to de-duplicate dependencies and un-bundle software. This way when you install two programs which require the same library, that library only gets installed once. This frequently requires small modifications or patches by the distribution maintainers to make tens of thousands of packages coexist happily in the same repository, but it means if a security vulnerability is found in a very commonly used library (like xz-utils for instance), the patch only needs to be applied to that one package instead of to thousands of individual programs which use it.

    Many programming languages come with their own package managers - cpan for Perl, pip for Python, cargo for Rust, npm for Node.js, etc. While these make getting started on a software project more simple, these are the bane of most distribution package maintainers because they undermine the goals of a distribution package manager (by installing different versions of the same software in multiple places and not tracking them globally for updates).

    Long story short, instead of installing Python packages globally through pip (the instructions you’ll see on many Python-based project READMEs), you’ll first want to create a Python “virtual environment”. After activating the virtual environment, you can proceed using pip as normal. This installs the software and all of its dependencies within a sub-directory, without requiring any privilege escalation or interference with the distribution package manager. This also means that if you screw up, you can simply delete the virtual environment and try again, instead of sifting trough /usr/ and trying to figure out which files were created by your distro and which files were created by pip and praying none of the former got overwritten by the latter.

    Of course, this is only necessary if you distro doesn’t package the program in the first place.

    • Erika3sis [she/her, xe/xem]@hexbear.netOPEnglish
      2·
      25 days ago

      So, I need to get PyTorch, and there are five install commands on PyTorch’s website: one for computers with CUDA 12.6, one for 12.8, one for 12.9, one for computers with ROCm 6.4, and one for computers with neither ROCm nor CUDA. Checking my version of CUDA (nvcc -V) shows 12.0. Is there a way for me to update CUDA; should I use the command for the non-CUDA, non-ROCm version of PyTorch; or will it go fine for me to install a CUDA version of PyTorch even if my version of CUDA is apparently a bit too old?

      But even when I use any of the install commands from PyTorch’s website (variants of pip3 install torch torchvision), each of them gives me an “externally-managed-environment” error — which recommends just as you say creating a virtual environment. But it also says “make sure you have python3-full installed [before creating the virtual environment]”, and I’m not sure how to do this, or if I don’t have python3-full how to get it.