141–150 of 175 posts
Earlier quoted context omitted.
In Europe, TeamSpeak is still very popular.
I used to play EVE Online a fair bit, and always thought it interesting how some of the groups used Discord but only for text communications. Voice was still done over Teamspeak or Mumble.
Earlier quoted context omitted.
That didn't stop IRC being popular in the 1990s. There has long been a place in the ecosystem for ephemeral chat. Often alongside non-ephemeral things like written documentation.
> That didn't stop IRC being popular in the 1990s. IRC chats, especially in opensource projects channels, could and would be archived, published over the web and indexed by search engines.
#haskell on Libra is publicly logged, but I couldn’t get Google to return a quoted phrase from a message a few weeks ago.
Many people on IRC don’t enjoy being in logged channels. I’ve also heard that there are GDPR implications to publicly logging people’s messages without their consent.
Discussion of the difficulty and downsides of IRC logging, from a coulple years ago:
=> https://news.ycombinator.com/item?id=22892015
=> https://web.archive.org/web/20200417001532/https://echelog.c...
The HN blowback to developers choosing to use Discord is just wildly out of proportion.
Earlier quoted context omitted.
In my experience it doesn't work like that. It's like if you take an idiot and spend a bunch of time training them, they'll still perform much worse than a moderately intelligent person with much less training. And smaller models can be pretty idiotic.
Then why do Chess AI perform much better than LLMs trying to play chess.
An LLM could have picked up some chess patterns through osmosis, but it can not reason explicitly in the domain.
Suppose I’ve written code that calls the OpenAI API. Is there some library that helps me easily switch to a local/other LLM. I.e a library that (ideally) provides the same OpenAI interface for several models, or if not then at least the same interface.
Hi all, I'm the main maintainer from the OpenLLM team here. I'm actively developing the fine-tuning feature and will release a PR soon enough. Stay tuned. In the meanwhile, the best way to track the development workflow is at our discord, so feel free to join!!
Earlier quoted context omitted.
Thanks for the great project! Any chance, your team might consider more open platform than Discord for posting updates? I personally find Discord hard to use, and there’s no way to have sensible subscription (like RSS). Discord is usually muted.
Discord is a black hole where information goes to die. Its search and scrollback is awful. It's awful at being an archive, as finding anything that was asked more than a day or two ago is impractical. To use Discord in good faith and with open eyes, you have to prioritize communication in the present, and give up hope of archiving anything that was said for people who might need the information in the future.
Earlier quoted context omitted.
For falcon 40b you probably need an A100 40gb or so. Every model is drastically different. If you want to run something on consumer hardware, your best bet is using anything ported to the ggml framework, especially if you're on Apple silicon.
Do you know any "standard" way or measures to determine the approximate hardware requirements of a model?
If you want to train a model, that’s a different story.
What kind of hardware do I need to run something small scale (1 user concurrently) and get reasonable result? Are we talking about Raspberry Pi, Core i5, Geforce 4090?
I'm upgrading my M1 laptop on Thursday to a newer model, M2 MAX with 96GB of memory. I'm totally going to try Falcon-40B on it, though I do not expect it to run that well. But I do expect it (the M2, I mean) will be snappier on the smaller models than my original M1 is.
Earlier quoted context omitted.
I’m not sure ML researchers would agree that number of (compressed) bytes are more meaningful than number of parameters. Parameters have mathematical meaning – bytes doesn’t.
My point is that they don't have a mathematical meaning. They have a training-time impact, but that's more relevant when creating than using. You'd need to know how many parameters were independently covarying for any given class of predictions. It certainly isnt all of them. You could cite the "average dropout percent to random-level accuracy on a given class of problems" (my guess is that this would show 5-20% of p…
Yes.
A more helpful bit of information could be what the model was pre-trained on. Assuming they’re trying to refine it for a more specific task.
Size is helpful for “what can I run on my machine” (or how much would it cost to run on a server.) Not all models are created equal, given a byte size, for a given task.