jzombie

14 ms·

1. 121.▲
   by jzombie 3y ago

   Thanks for the share! I'll post this in my README as an alternative approach.
2. 122.▲
   Show HN: sshfs on Mac without macFuse

   2 points by jzombie 3y ago|2 comments
3. 123.▲
   Show HN: Pipper Makes Pip Easier(github.com)

   1 points by jzombie 3y ago|0 comments
4. 124.▲
   by jzombie 3y ago

   "Thin air" is a weird metaphor for something that happens in space, where there is no air.
5. 125.▲
   by jzombie 3y ago

   Having read Napoleon Hill's book, "Think and Grow Rich", several times, and considering the book is now in the public domain, I thought that I would create a custom GPT using the book's epub as the context, and share the
6. 126.▲
   Think and Grow Rich GPT(chat.openai.com)

   2 points by jzombie 3y ago|1 comments
7. 127.▲
   by jzombie 3y ago

   Any business proposition that begins with "Hey, I hear you're a programmer!" doesn't sound like a good business proposition at all.
8. 128.▲
   by jzombie 3y ago

   I don't know if I totally agree with this. Maybe it's not nearly as bad as how the article frames it but just doing a Google image search of "landfills of old PCs" returns some pretty convincing images.
9. 129.▲
   by jzombie 3y ago

   I think nano LLMs are the next trendy thing, but I agree with your sentiment.
10. 130.▲
    by jzombie 3y ago

    Indeed. A landfill is the worst possible solution for this.
11. 131.▲
    by jzombie 3y ago

    Could potentially.
12. 132.▲
    by jzombie 3y ago

    Sorry to hear that. Apple Music is the only music subscription I have ever signed up for, and I like it a lot. Wondering how other music services compare to it.
13. 133.▲
    Opinion: Web popups do not work

    3 points by jzombie 3y ago|0 comments
14. 134.▲
    by jzombie 3y ago

    Thanks for sharing.
15. 135.▲
    by jzombie 3y ago

    I don't personally know. I've been curious if DALL-E has been truly been mixed in w/ ChatGPT as a single model using a Mixture of Experts (MoE)- type learning gate to train them all together.
16. 136.▲
    by jzombie 3y ago

    Curious as to why some sort of model specifically trained for prompt engineering couldn't be placed before a LLM in a pipeline, or used in a transfer learning scenario. I'm also curious as to why existing LLMs can't be fine t
17. 137.▲
    by jzombie 3y ago

    If you could squeeze in LLM integration for that price, somehow undercutting GPT+, but with RAG on your own data, it would be a steal.
18. 138.▲
    by jzombie 3y ago

    > the label wanted more in licensing fees than the whole thing cost to produce. Hopefully this message was sent instead.
19. 139.▲
    by jzombie 3y ago

    Nice! Thanks for the tip. Anyway, I appreciate this conversation and am going to make some coffee.
20. 140.▲
    by jzombie 3y ago

    > I think my next attempt will be a handful of hand crafted rules on daily data. I'm definitely never going to try day trading scale strategies or try to pick good points intraday to buy. It should be resilient to the price differen
21. 141.▲
    by jzombie 3y ago

    I started w/ an LSTM and wound up w/ a mixture of transformers, CNNs and skip connections.
22. 142.▲
    by jzombie 3y ago

    Yes, actually I was planning on training a separate model to determine buy/sell signals, using this prediction model as possibly part of a transfer learning process, and do paper trading w/ it to start w/.
23. 143.▲
    by jzombie 3y ago

    Thanks for the feedback. Will try that. Been mostly using Keras Tuner and Tensorboard to slowly get to the point I'm at and have had some wild results in the process.
24. 144.▲
    by jzombie 3y ago

    I thought MSE was tied to the predictions it makes and not the structure itself (i.e. if you get the same value out for the same input, using different structures, you should at least get the same MSE).
25. 145.▲
    by jzombie 3y ago

    I can assure the data is not shuffled and is ordered in chronological order (time series model). However at present I've only got training/validation splits and not an additional testing.
26. 146.▲
    Ask HN: 0.000068 MSE for TensorFlow Model?

    1 points by jzombie 3y ago|14 comments
27. 147.▲
    by jzombie 3y ago

    No problem. My apologies for my tone in this thread, as well. Good luck with your learning adventure!
28. 148.▲
    by jzombie 3y ago

    And ChatGPT itself is NOT built using JS or React if that's your question... It's a model built using PyTorch, I believe, using some very expensive hardware. Just the user interface is [theoretically] coded in React.
29. 149.▲
    by jzombie 3y ago

    Why not just take it for its word? I don't see, for instance, why they might use Svelte instead, and some evil genius decided to just hardcode that into the window object.
30. 150.▲
    by jzombie 3y ago

    A lot of people actually do this: https://news.ycombinator.com/from?site=discussions.apple.com