FL33TW00D

20 ms·

1. 301.▲
   by FL33TW00D 6y ago

   If you have a large model in excess of 500mb, any advice on the next best way of deploying as cost effectivtely as possible?
2. 302.▲
   by FL33TW00D 6y ago

   Sounds very interesting! I think something like this could be applicable: https://yjernite.github.io/lfqa.html Although with the finely tagged data it may not be necessary.
3. 303.▲
   by FL33TW00D 6y ago

   I've just spend the last few months working on a medication potency scoring system using ASHP data. Would be interested to hear more about your approach.
4. 304.▲
   by FL33TW00D 6y ago

   You beat me to it. Any time someone mentions an electric plane the MJ/Kg jumps to mind. Until battery energy density increases by an order of magnitude electric planes aren’t going to happen, at least commercially.
5. 305.▲
   by FL33TW00D 6y ago

   As far as I know to get a V100 you need Colab Pro? Did this change recently?
6. 306.▲
   by FL33TW00D 6y ago

   Is the mom test actually worth the money? I’ve seen it touted in a few places and again here.
7. 307.▲
   by FL33TW00D 6y ago

   Yeah currently experimenting with T5 large in FP16. Will then use the ONNX conversion script to decrease inference time and deploy it!
8. 308.▲
   by FL33TW00D 6y ago

   Not yet but I am working on it!
9. 309.▲
   by FL33TW00D 6y ago

   From my experience T5 is currently the best publicly available model for NLG.
10. 310.▲
    by FL33TW00D 6y ago

    I think that they're correct saying that GPT-3 isn't revolutionary, since it just demonstrates the power of scaling. However I would argue that the underlying architecture, the Transformer (GP(T)), is/was/will be revolu
11. 311.▲
    by FL33TW00D 6y ago

    Interesting post, I’m working on a whole graph classification problem at the moment. If GNNs aren’t the future, then where else should I be looking?
12. 312.▲
    by FL33TW00D 6y ago

    Thanks this was a great presentation
13. 313.▲
    by FL33TW00D 6y ago

    Anyone have any resources to read on how these things work?
14. 314.▲
    by FL33TW00D 6y ago

    Brilliant write up!
15. 315.▲
    by FL33TW00D 6y ago

    They’re seeing some interest in neuroscience as the brain can be represented with a graph structure.
16. 316.▲
    by FL33TW00D 6y ago

    Thanks for the link! Primary bit of interest for me was the announcement of collaboration with HuggingFace. NLP coming to handheld without a strong internet connection will be very interesting.
17. 317.▲
    Qualcomm Snapdragon 888(qualcomm.com)

    2 points by FL33TW00D 6y ago|2 comments
18. 318.▲
    by FL33TW00D 6y ago

    Great work! I've used dcm2niix[1] many times in the past to get around using DICOM directly, perhaps this will make it easier in future. [1] https://github.com/rordenlab/dcm2niix
19. 319.▲
    by FL33TW00D 6y ago

    After reading through https://www.gwern.net/GPT-3 , I suspect that GPT-3 is capable of higher order reasoning, given the right motivation (prompt).
20. 320.▲
    by FL33TW00D 6y ago

    100% Agree. The research is there but the engineering to make it easy to use is lagging. I look forward to the day it's easy to create visualizations like so: https://exbert.net/
21. 321.▲
    by FL33TW00D 6y ago

    Seminal papers in this area AFAIK are StyleGAN1 and StyleGAN2: https://github.com/NVlabs/stylegan2 https://arxiv.org/abs/1912.04958
22. 322.▲
    by FL33TW00D 6y ago

    Related video from 2018: https://www.youtube.com/watch?v=FlgLxSLsYWQ
23. 323.▲
    by FL33TW00D 6y ago

    Very interesting paper I had not seen before. Seems like with more targeted magnetic stimulation the effects could be enhanced.
24. 324.▲
    by FL33TW00D 6y ago

    +1 for Speech and Language Processing. Not only helped me with NLP but really solidified backprop.
25. 325.▲
    by FL33TW00D 6y ago

    I've just recently been on a journey to understand transformers in and out, here are the resources I've found managed to drill it into my head: 1. Chapters 7, 9, 10 https://web.stanford.edu/~jurafsky/slp3&#x2F
26. 326.▲
    by FL33TW00D 6y ago

    Have you been able to get LDAC to provide high quality audio whilst still being able to use the microphone on the XM3's?
27. 327.▲
    by FL33TW00D 6y ago

    I work in this field and whilst it's true that there is no direct medications that can seriously slow or halt the progression, there are effective avenues for slowing it via diet, exercise, reduction of stress factors: https:/&#x
28. 328.▲
    by FL33TW00D 6y ago

    Brilliant.