Hey. Yeah you. No don’t look over your shoulder. I’m not talking to the guy behind you. Look, we’ve been meaning to tell you that you’re doing a pretty good job out there. Proud of you. Keep up the good work.
Yeah I feel Linux has a lot of dead ends. Its easy to follow the wrong path. My saving grace has always been that once you get things working, you know how you did it and it likely won’t change much.
So really its a big search, but once you hit a steady state it really feels like home.
Never change lemmy
Its great! It will not hesitate to kill you until you learn it.
Binding of Isaac has a bit of invincibility after every hit, and that works for the game really well. You need to be able to recover and not die immediately.
Noita has no invincibility frames like that. You need to be able to die immediately. An instant death is essential for the game tone and balance.
Surge XT can do 98% of everything I need. It has a ton of waveforms, filter types, and modulation options. It’s a great synth.
In the rare case that SurgeXT can’t do what I want, I turn to zynaddsubfx. It is complicated, but really one of the wildest synths out there.
Last resort is this open source fork of VCV rack I found. So at the end of the day if you need to you can just build the synth you need.
Noita!: I agree with many in dubbing it my favorite single player game.
It is a pixel art alechemy and magic roguelike where every pixel is physically simulated and destroyable. Find spells and wands and combine them in programmer like ways to multiply their power. Many tricks in the game are so strong they feel like cheating, but the game has been built with challenges that surpass every trick.
It is the epitome of endlessly repayable. Especially switching between the few mods that drastically change the map.
I can easily have a god run last me 5-10 hours. But I still have a lot of fun with the more common 30min-2hr runs.
Here is my software set up for open source music making (I often use midi and audio input):
Core software:
For SFZ instruments:
Raw samples:
For effects:
Ardour is supposed to have a midi related update for its next big release, so stay tuned.
From my perspective, the simplest set up would be Ardour and SFZ instruments. Mainly because I’m quite used to those two.
For set up you’d just open Ardour, make a new midi track, place liquidsfz as the first plugin on the track, then open liquidsfz and browse to the SFZ file you want to use.
Then you just draw in the midi notes you want using either the edit or draw tool.
Thats a good one
Yes with llamacpp its easy to put just the experts on the CPU. Since only some of the experts are used every time, the GB moved to RAM slows things down way less than moving parts of the model that are used every time. And now parts that are used every time get to stay on the GPU. I was able to get llama4 scout running at around 15 T/s on 96GB RAM and 24GB VRAM with a large context. The whole GGUF was about 80GB.
Also they actually are a Chinese company. I am pretty sure it is the company that makes RedNote (Chinese tiktok) and thats why they had access to so much non-synthetic data. I tried the demo on huggingface and never got any Chinese characters.
I also really enjoyed it’s prose. I think this will be a winner for creative writing.
Well they get to use my friends social inertia against me. Hard to move off of snap even if I wanted to and so they can just change the TOS and get my data.
But they don’t have to do it that way.
It could be a pretty bomb model though.
Yo thanks for the resources! I never even considered doing it myself. Kinda scared it would reflect me too well yk lol.
I may give this a shot.
Best I can offer is Hannah Montana linux
If you were down to use hugging face DeepHeremes is a reasoning model built on top of Mistral Small 24b. It’d fit decently well in 20GB.
Maybe the ollama run hf.co/{username}/{repository} command would make it easy enough for you.
Reasoning models usually are better for math.
I love being on Lemmy. This is exactly why I’m on Lemmy.
Nous has been awesome. That Ascension Maze model card was interesting. Definitely sounded like the model wrote it itself lol.
I am the biggest fan of folks diving into thought processes and how to amplify cognition, and I really appreciate Nous doing so much of it out in the open.
Huh. I didn’t realize that thanks. Lame that they would hold back the one that is the biggest size most consumers would ever run.
There actually is a 32b dense
Here is link for ollama for Gemma 3 QAT https://ollama.com/eramax/gemma-3-27b-it-qat:q4_0
There are ggufs around if you want to try it on another back end.
I think the next bit of performance may be leaning hard into QAT. We know there is a lot of wasted precision in models, so the more we understand that during training the better quality small quants can get.
I also think diffusion LLMs ability to change previous tokens is amazing. As well as the ability to iteratively use an auto regressive LLM to increase output quality.
I think a mix of QAT and iterative interference will bring the biggest upgrades to local use. It’ll give you a smaller higher quality model thay you can decide to run for even longer for higher quality outputs.
I’m pretty convinced it is just a bad set of models.
I’m waiting for qwq2 or something. Llama hasn’t been my choice for a while.
I am young and have a computer science degree, and I still struggle at times. I get it.
For games, I’d try to install steam and run them through steam if thats how you’d normally do it on windows. Then for me the main setting to play with (on a game by game basis) is setting the game to use proton (in the compatibility settings of the game) and whether or not to use steam input for controller support.
If you are trying to install a non steam game, maybe look into lutris. Though I’m on the techy side, and I hear a lot of people like heroic game launcher on the less techy side.
Good luck. I think it’s fair to run out of energy while trying get the right combo, but if ya stick to it I’m confident you’ll find the set up that works for you.