gunalx

7 ms·

91.
by gunalx 6mo ago
The title should be corrected to not editorialize.
92.
by gunalx 6mo ago
Metawith the llama series as well,they just didn't manage to keep upping the game after and with llama4.
93.
by gunalx 6mo ago
Running q3 xss with full and quantizised context as options on a 16gb gpu and still has pretty decent quality and fitting fine with up to 64k context.
94.
by gunalx 6mo ago
I habe found llama-fit sometimes just selects a way to conservative load with VRAM to spare.
95.
by gunalx 6mo ago
Look into opencv.
96.
by gunalx 6mo ago
I'm guessing incomplete migrations or similar problems where stuff is just partially written, or out of sync. The Plex db is kinda complex under the hood.
97.
by gunalx 6mo ago
Ham radio, can be either really technical designing transmit protocols, antennas, SDRs or really just talking to random people as far away as you can reach without being bound to the internet.
98.
by gunalx 6mo ago
Indeed it does. Once i see degraded state i revert to last task and run a compact, before starting up again.
99.
by gunalx 7mo ago
We didnt get deepseek v4, but gemma 4. Cant complain.
100.
by gunalx 7mo ago
Glm also has their subscription witch I would assume heavy users to use.
101.
by gunalx 7mo ago
Only onlyoffice being petty. A good reason to use LibreOffice or Collabora instead.
102.
by gunalx 7mo ago
Most use definition is just awailable weigths. This kids make sense because "compiling" (training) the model cost inhibitly much, and we can still benefit from the artifacts.
103.
by gunalx 7mo ago
If you have to ask you dont really need the answer. Seems to not be to difficult in finding or creating training code. So a pretty decent amount of high quality training data should be many hours. And a few hours in high end data enter GPU
104.
by gunalx 7mo ago
Without memory you have no data to compute on. Memory and compute scaling only makes sense in tandem.
105.
by gunalx 7mo ago
Well, they are to some degree.
106.
by gunalx 7mo ago
Yes, especially when considering a dataceter needed the energy of pretty many people to be built. A single human is indeed more efficent, and way more flexible and actually just general intelligence.
107.
by gunalx 7mo ago
True. Desktops are becoming a nishe. Casual pc use most likely has a laptop. More and more just a phone and maybe a ipad or other tablet.
108.
by gunalx 7mo ago
Probably just means SFT fine-tuning a base model, vs behavioural dpo and/or SFT fine-tuning a instruction model.
109.
by gunalx 7mo ago
I also got a slight feeling of ai assistance as well (especially on the drawings), but the story was well written and really sucked me in all in all.
110.
by gunalx 7mo ago
not as simple as just OCR and map though. Some letters want space above them some want to be placed lower. take g and f and c for examples g and f are about the same height but different ofsets, and c would look like a capital C if scaled
111.
by gunalx 7mo ago
3 not 3.5? I think I would even prefer the qwen3.5 0.8b over GPT 3.
112.
by gunalx 8mo ago
qwen 3.5 is really decent. oOtside for some weird failures on some scaffolding with seemingly different trained tools. Strong vision and reasoning performance, and the 35-a3b model run s pretty ok on a 16gb GPU with some CPU layers.
113.
by gunalx 8mo ago
for privacy preserving direct inference: Fireworks ai nebius otherwise openrouter for routing to lots of different providers.
114.
by gunalx 8mo ago
Sad to not see smaller distills of this model being released alongside the flaggship. That has historically been why i liked qwen releases. (Lots of diffrent sizes to pick from from day one)
115.
by gunalx 8mo ago
It does habe dotfiles. (.config/noctalia. it is possible to declare mostly declaratively if you want)
116.
by gunalx 8mo ago
You can also kinda read the 3 categories as office, azure, windows. But that is a gross oversimplification.
117.
by gunalx 8mo ago
Glm models with vision ends on a V.
118.
by gunalx 8mo ago
I think this is it. https://huggingface.co/mistralai/Voxtral-Mini-4B-Realtime-26...
119.
by gunalx 9mo ago
Agree, felt a little less packed than before. H building was brutal though.
120.
by gunalx 9mo ago
That is exactly the only feature I would like to disable. (have custom theming already)

More ›