Disonantezko

joined 2 years ago
[–] Disonantezko@lemmy.sdf.org 2 points 4 months ago (1 children)

My gear is an old:

I7-4790 16GB RAM

How many tokens by second?

[–] Disonantezko@lemmy.sdf.org 1 points 4 months ago

It's not out of my scope, I'm just learning what can I do locally with my current machine.


Today I read about RAG, maybe I'm gonna try an easy local setup to chat with a PDF.

 

I have an GTX 1660 Super (6 GB)

Right now I have ollama with:

  • deepseek-r1:8b
  • qwen2.5-coder:7b

Do you recommend any other local models to play with my GPU?

[–] Disonantezko@lemmy.sdf.org 2 points 4 months ago (1 children)

Even at 160kbps, maybe 1/1.000.000 people can recognice a FLAC vs MP3 trying 10 times (continuous) using expensive headphones and players, 320kbps is overkill, I prefer a FLAC and just encode to Opus.

Right now Opus is better and can be played in web browsers, smartphones, YouTube and Netflix are using that for awhile.

[–] Disonantezko@lemmy.sdf.org 2 points 8 months ago* (last edited 8 months ago) (1 children)
  1. LTSC + WSL (Better than VM)
  2. Dual Boot
  3. Linux only
[–] Disonantezko@lemmy.sdf.org 7 points 8 months ago

Or Win10 IoT LTSC till 2032

[–] Disonantezko@lemmy.sdf.org 2 points 8 months ago

To me Adobe has very bad UI, I did try to use it, and first time was awful. Freehand was a lot more intuitive, but when Macromedia was bought, was killed.

I get, that a lot of people did learn to use Adobe UI, and of course they want the same because they're used to, but doesn't make it better.

Affinity is more friendlier that PS to me.

I'm not saying that GIMP UI is perfect or good, but right now, to my casual use case, is not bad. Obviously can be better, and get some ideas from other UIs.

[–] Disonantezko@lemmy.sdf.org 5 points 8 months ago* (last edited 8 months ago)

I prefer:

  • ImagePipe: fast edit
  • Snapseed: complex edit (not FOSS)
  • Aves: gallery
  • Superimage: AI upscaler (RealESRGAN)
  • Waifu2x NCNN: AI upscaler (Waifu2x, RealCuGAN)
[–] Disonantezko@lemmy.sdf.org 5 points 8 months ago

With ChaiNNer you can remove background, upscale (local), it's a lot more flexible and compatible with models than Upscayl, also a little bit more complex (node based, not as complex as comfyUI). You can upscale an image with a face model and use other model for everything else in the same image.

[–] Disonantezko@lemmy.sdf.org 3 points 10 months ago

Use Termux, you need:

  • Android Tablet/Smartphone.
  • Bluetooth keyboard.
  • Cheap stand 4 device.

That's my setup to play tabletop rpg (DND5E) for a couple of years, all CLI using any text editor you like with markdown. I use: tmux, vis.

In your case: SC-IM, visidata, any text editor.

With Termux you can use packages from repo or from other distros with proot-distro, like: Alpine, Avoid, Debian, Arch, Fedora, Ubuntu.

[–] Disonantezko@lemmy.sdf.org 3 points 10 months ago* (last edited 10 months ago)
  • ChaiNNer: is very flexible and can do all of that. Node based.
  • Upscayl: do only upscales with less models, but has simpler UI, very fast in AMD GPU (because NCNN).

  • I haven't tried automatic1111 or ComfyUI yet.
  • All of them work offline withou limits and are more flexible than online alternatives.

GIMP has:

[–] Disonantezko@lemmy.sdf.org 1 points 1 year ago

Works with Chromecast 2 (no Android)?

[–] Disonantezko@lemmy.sdf.org 25 points 1 year ago

Do you know the exploit was detected in Debian Sid? (by a PostgreSQL developer), Arch got the update (with both compromised versions), but because don't directly link openssh to liblzma (as Debian), and thus this attack vector is not possible.

Also, other rolling distros also got the compromised versions, maybe: openSUSE Tumbleweed, Endeavour OS, Fedora Rawhide, Slackware -current, etc.

view more: next ›