eurekin

22 ms·

1. 301.▲
   by eurekin 3y ago

   I'm using: ollama run dolphin-mixtral:8x7b-v2.5-q3_K_S
2. 302.▲
   by eurekin 3y ago

   So the spinal cord is only a wire and, besides reflexes, there is no processing done anywhere else, but the brain? How about neurons in the gut, I've read somewhere that intestines are lined up with some neural "mesh". Is tha
3. 303.▲
   by eurekin 3y ago

   Thanks! That is every interesting. Here's a direct amazon link: https://www.amazon.com/dp/B0BYGB3WV4 And a running demo: https://forums.developer.nvidia.com/t/llama-2-llms-w-nvidia-...
4. 304.▲
   by eurekin 3y ago

   In the spirit of hacker news, one can also build a rig such as this one: https://nonint.com/2022/05/30/my-deep-learning-rig/ and land a job at OpenAI as a side effect. Of all places, I wasn't expec
5. 305.▲
   by eurekin 3y ago

   Since n-times a 3090 is still a much better offering
6. 306.▲
   by eurekin 3y ago

   As the ancestor commenter mentioned: > If you’re interested in ML training Training - at least the one I tried - requires to be run in fp16 mode. So a 7b net needs 14 GB for the model weights alone, plus some extra for the context and th
7. 307.▲
   by eurekin 3y ago

   Attention mechanism, the core of LLM, is universal enough to be brought back to standard vision models. Which is kind of ironic, since vision models were dominated by convolutions, and, the transformer is dubbed "convolution for text&q
8. 308.▲
   by eurekin 3y ago

   Of course
9. 309.▲
   by eurekin 3y ago

   Thanks! That rules them out completely EDIT: For training
10. 310.▲
    by eurekin 3y ago

    My guess: because it's useful with following analogy. We use text embeddings to represent concepts written words. They lack the nuance, when the same word has different meaning in different concepts. LLM's use text embeddings and
11. 311.▲
    by eurekin 3y ago

    I'm betting a higher frequency, lower latency, high throughput dedicated part. The slow stuff thinks for longer and sends a "goal" to this part, which translates that into a series of higher frequency (comparing to the slow p
12. 312.▲
    by eurekin 3y ago

    I was also stumped by this exact quote. The whole article was in the best spirit, until this. It's a model we have, it will get updated in order to be more useful. Every engineering field builds a model of reality. Who are the connecti
13. 313.▲
    by eurekin 3y ago

    Well, MLPs is the actual neural network (the approximator), whereas the attention (the rest) is more of a text to relevant embedding extractor
14. 314.▲
    by eurekin 3y ago

    I always assumed ANN simply is a universal and learnable function approximator. That is, there is no direct equivalent of classical conditioning. Only data in, expected output pairs
15. 315.▲
    by eurekin 3y ago

    > A lot of our motor learning is in fact wiring the cerebellum correctly. That got me interested: since the wiring is so long (from limbs to cerebellum), what kinds of motor learning? Do we know, if cerebellum needs more energy than the
16. 316.▲
    by eurekin 3y ago

    ChatGPT knows nothing of those libraries and their code. I can't paste anything into context, because of security reasons :)
17. 317.▲
    by eurekin 3y ago

    So, there goes my next evening! :D Thank you!
18. 318.▲
    by eurekin 3y ago

    I'm dreaming of the day there would be a chatgpt like thing for all those internal proprietary tools and their "documentation"
19. 319.▲
    by eurekin 3y ago

    Force hirings? How's that any good? I'd rather be given a opportunity to learn and grow competencies so later, I'd apply as a valuable contributor. Edit: I genuinely don't know what's the intended meaning and the wa
20. 320.▲
    by eurekin 3y ago

    If anybody asks me who do I want to be, when I grow up I'll point at this author
21. 321.▲
    by eurekin 3y ago

    Exactly! Let's say Heimdall for example. It's a simple website. Can I "cast" that to the chromecast?
22. 322.▲
    by eurekin 3y ago

    I have family in Japan, it's complicated. Extending hand is OK, forcing others to comply is not
23. 323.▲
    by eurekin 3y ago

    So if I'd like a custom dashboard I can? How do I start?
24. 324.▲
    by eurekin 3y ago

    I think that's the common sense thing really
25. 325.▲
    by eurekin 3y ago

    What are kv cache params?
26. 326.▲
    by eurekin 3y ago

    Can you elaborate? :)
27. 327.▲
    by eurekin 3y ago

    I'd just do the best I could in that situation
28. 328.▲
    by eurekin 3y ago

    I think, if you're competent that's all that matters
29. 329.▲
    by eurekin 3y ago

    Accurate, but not precise. I'd imagine they were focused on catching as many violations as possible, at the cost of catching false positives into net.
30. 330.▲
    by eurekin 3y ago

    How is learning performed than? By means of genetic evolution - on the level of a individual or the whole colony (queens mutations)?