bestcoder69

8 ms·

31.
by bestcoder69 3y ago
Whether someone is “unable to meaningfully consent” is unknowable, so I don’t put any stock it in at all in these kind of decisions. I’d say that deciding that question isn’t that important (as described, fine, don’t ban WorldStarHipHop). R
32.
by bestcoder69 3y ago
Looking for a llama2 fine-tune guide specific to Apple silicon, if anyone has one. I wanna see how big of a model I can tune on my 64gb mac studio
33.
by bestcoder69 3y ago
transaction=voluntary=good is an extreme libertarian position, eg bumfights. I’m not saying you support bumfights, but your argument does. Or child labor etc. If there’s more nuance you gotta say it.
34.
by bestcoder69 3y ago
Same here (mps errors). I tried after the initial musicgen release. I’ll have to check again, but I remember AFAICT my hardware wasn’t getting saturated, so maybe there’s headroom for mac cpu performance. And of course in the meantime I’ll
35.
by bestcoder69 3y ago
Anyone know if there are ways, as-is, to speed this up on Apple Silicon? This setup takes 5 minutes: - Mac Studio M1 Max 64GB memory - running musicgen_app.py - model: facebook/musicgen-medium - duration: 10s
36.
by bestcoder69 3y ago
https://twitter.com/iris_IGB/status/1685322871306928128
37.
by bestcoder69 3y ago
Maybe thinking of this? https://twitter.com/quantian1/status/1686174330793660416
38.
by bestcoder69 3y ago
Could be a skill issue on the part of the replicators? The russian anime girl scientist had to make a bunch of modifications (and was cursing the orig authors) before she could make her rocks float.
39.
by bestcoder69 3y ago
This could’ve been an interesting article. E.g.: - how to use flask with recent pytest versions (testing isn’t totally trivial in flask IME - maybe it’s better in 2023?) - recent language features that make flask nicer - best ways to deploy
40.
by bestcoder69 3y ago
Is this a bet against LLMs?
41.
by bestcoder69 3y ago
Re langchain: > we were able to accomplish most of our needs with a relatively small volume of Python code that we wrote ourselves Every single time someone posts a trip report building something on LLMs I love to ctrl+f to the part wher
42.
by bestcoder69 3y ago
Free ChatGPT runs 3.5. You have to upgrade to plus to use GPT-4. The APIs seem close to ChatGPT, but it’s a little opaque what they’re actually doing. If you inspect network requests the models are named something like “chat-render-3.5” ins
43.
by bestcoder69 3y ago
Hm, well, ChatGPT is a boy.
44.
by bestcoder69 3y ago
Why this instead of GPT Functions?
45.
by bestcoder69 3y ago
Hindsight is 20/20 but I feel like that mistake is on the Founding Fathers. If you want us to do something, say it in the place where you put all the rules instead of hoping we'll read some other thing, y'know. Maybe a datapo
46.
by bestcoder69 3y ago
React works. NextJS works. OpenAI's ChatCompletion API with `functions` works. Go on langchain's discord and ask their documentation bot anything.
47.
by bestcoder69 3y ago
The core data structure, the Chain, is basically just a function. Combining chains is function composition, like literally it's just f(g(x)), but incompatible with _your_ f's and g's without an adapter. Read this page and men
48.
by bestcoder69 3y ago
https://github.com/search?q=repo%3Asimonw%2Fllm%20OPENAI_API... Let us know what you find.
49.
by bestcoder69 3y ago
Pro-tip (though could be outdated): if your reply gets a yellow-tick reply then your post gets ranked higher than the blue ones. To make this happen, @-mention an automated customer service bot.
50.
by bestcoder69 3y ago
Yes, this seems like ChatGPT’s style of writing. Another post has a conclusion that matches ChatGPT’s conclusion style: > Remember, namespaces are a powerful feature that requires careful configuration and management. With proper knowled
51.
by bestcoder69 3y ago
Did they keep all their data in RAM on a single server too?
52.
by bestcoder69 3y ago
You know you can just look up income stats for SF right?
53.
by bestcoder69 3y ago
I've only seen people mention that it runs really slow, even on like A100s.
54.
by bestcoder69 3y ago
Definitely don't quantize to 4-bit at this param count, if that's what you're doing. The redpajama.cpp repo is a bit crazy for making that the default. I'd say at this size don't count on accuracy. Think: fancy auto
55.
RedPajama 7B (an Apache 2.0-licensed LLaMa) is now available(together.xyz)
85 points by bestcoder69 3y ago|25 comments
56.
by bestcoder69 3y ago
Today I used it to rewrite a function in llama.cpp that compiled in gcc9 but not gcc7 (an avx2 intrinsic wasn’t available… idk specifics, I barely know C/C++). First attempt was wrong, so I had it write test code to debug. Using the te
57.
by bestcoder69 3y ago
Can also get rid of the `run` and move parens to simplify even further: function doWork() { const x = () => { if (foo()) return f(); if (bar()) return g(); return h(); }; return x() * 10; }
58.
by bestcoder69 3y ago
ChatGPT plugins will be usable through the OpenAI API, and both of those services are available as ChatGPT plugins.
59.
by bestcoder69 3y ago
Same. I feel like once you strip away the abstractions that you probably aren't going to use long-term, you're left with something that competes with python f-strings. And yep I've tried it at multiple points in its evolution
60.
by bestcoder69 3y ago
Has langchain had positive ROI for anyone beyond the initial prototype of their app? My experience has me skeptical - I end up feeling like I'm painted into a corner and need to start over. Maybe if I just used PromptTemplate and LLMCh

More ›