Alex @ stsquad @lemmy.ml

FLOSS virtualization hacker, occasional brewer

---

Read more

Posts

10

Comments

349

Joined

2 yr. ago

I'm an August baby so was the youngest of my cohort going into uni and while I've done ok I do wonder if I'd have done a bit better academically if I hadn't been so distracted by all the other things on offer to a independent 18 year old.

How do you get the messages to your laptop? My googling only lead to various hacks to enable hidden iCloud folders and copying from there. Anyway I suspect I'm going a bit of topic but the point was it wasn't as smooth a process as I was expecting from my experience upgrading phones on the same OS.

I dunno. I tried to transition my sister in law's WhatsApps from her iPhone to Android and it's as far as I can tell impossible because the backup on the respective cloud services are hidden and can't be loaded manually.

Do you usually have some other front-end over the model? I can run llama.cpp directly in interactive mode but the results are a little underwhelming. However there seem to be various front ends that get better results? Is this down to better prompting and parameter control? I've seen temperature mentioned in relation to ChatGPT but I have no idea what rope and yarn factors are for?

Is there a standard for the suffixes? For example the OpenLlama models here: https://huggingface.co/SlyEcho/open_llama_7b_v2_gguf/tree/main have qN and and then a mix of K, M, 0 and 1 suffixes. The q I assume is the quantisation level but measured how? Does q2 mean t 2bits per weight? That seems very small - and what is it fixed float, integers?

Where is the sweet spot for running CPU bound models? I've just started playing with llama.cpp but the big models do make the cores work pretty hard. Should I look at using quantisation or more fine tuned models for the tasks I care about (developer assistance mainly).

I'm very lucky that I get to work in an upstream focused open source job. But I also maintain a few small packages personally and those only get attention when there are contributions to review or I have a personal itch to scratch. I'll leave enhancement requests in the trackers and just mark them as such and occasionally have a go at them if I feel the urge. No one not paying should expect anymore from maintainers.

Is this what people who haven't been introduced to #magit use?

Or where commanders stand in tanks.

A couple I've enjoyed recently:

Kleo - an ex Stasi assassin adjusts to the fall of the Berlin wall and takes revenge for her betrayal by the state (German)

Extraordinary Attorney Woo - a legal procedural following the autistic Woo as she navigates the law and life. You wouldn't think it was my thing but I found it very charming (Korean)

The Risa community seems by far one of the more participatory ones and their meme game is strong. But I am slowly seeing engagement go up on other subs. I'm patient.

Git is a distributed block chain, from a certain point of view.

You can do one thing and buy it for Android and then run it under either mcpelauncher or waydroid. Sure paying for the thing multiple times sucks but welcome to multi-platform propriety software. At least I only payed for it once for the whole family.

A parody of a parody no less. When will I see the likes of Airplane or Hotshots again?

I found Sausage Party hilarious because it's concept was fun and the ending was WTF but in a good way. Certainly not worst ever material from my point of view.

There are a fair number of YouTube creators who post ad-free versions of their content on Nebula. I think I got Nebula as part of a curiosity stream sub which last I checked was pretty cheap compared to the other streaming services.

Some sort of orbital death beam? I seem to recall a 2000ad story around a space energy beaming facility that goes horribly wrong.

I saw video of leaflets being dropped on the BBC news last night.

You can integrate Mastodon comments into your website with a little bit of JavaScript.

Yes - New Zealand introduced this policy first.