dharma1

76 ms·

1. 931.▲
   by dharma1 9y ago

   No counterparty risk on decentralised exchanges
2. 932.▲
   by dharma1 9y ago

   I think bitcoin could go a lot higher since the supply is limited and as of yet there hasn't been an easy way for institutional investors to buy into it. But at the same time I don't think the comparison to Facebook or any other l
3. 933.▲
   by dharma1 9y ago

   I dunno if this is good. Maybe for miners who want to hedge. But seeing how much bitcoin price reacts to news and how relatively easy it is to spread fake news, I think this just creates more incentives for market manipulation. It's pr
4. 934.▲
   by dharma1 9y ago

   Are there any frameworks yet supporting the advertised 120 TFLOPs mixed precision training?
5. 935.▲
   by dharma1 9y ago

   You can use Tensor Cores on AWS p3 instances (v100) now. Just doesn't seem to be a whole lot of ML framework support for mixed precision training with CUDA 9 yet https://devblogs.nvidia.com/parallelforall/programmi
6. 936.▲
   by dharma1 9y ago

   couldn't you trade on decentralised exchanges where counterparty risk isn't a thing? at least if you're only trading between crypto pairs
7. 937.▲
   by dharma1 9y ago

   Yeah, will be interesting to see if they can bring the price of GPUs with tensor cores down, from what I've read it's too expensive to make them right now for consumer market. BTW - do you happen to know what DL frameworks current
8. 938.▲
   by dharma1 9y ago

   Yep, Nvidia is quoting 125 TFLOPs mixed precision on V100, boosted by Tensor Cores. Vega 64 can in theory do 25 TFLOPs half precision. But as you say there's a large price difference too. For a market segment that needs 1-8 GPU rigs fo
9. 939.▲
   by dharma1 9y ago

   Exactly this. A handful of good AMD engineers should be able to squeeze everything out of their own hardware by tuning fast convolution, gemm and other ML related kernels optimised for Vega. That happens once, not for every framework. AMD&#
10. 940.▲
    by dharma1 9y ago

    Vega is meant to have someting called HBCC - that effectively allows you to use system RAM or even SSD as GPU memory for VERY large datasets. I have no idea if it would be fast enough for deep learning workloads over PCIe https:/&#x2F
11. 941.▲
    by dharma1 9y ago

    Another thing to note is that you can in theory train with half precision without sacrificing much accuracy. There isn't that much support for it yet, for AMD anyway that I've seen, but Vega 64 has 2x speed half precision (25 TFLO
12. 942.▲
    by dharma1 9y ago

    Essentia does audio fingerprinting with Chromaprint, which is essentially that. There are also other open source projects using Chromaprint focused specifically on that - https://github.com/worldveil/dejavu
13. 943.▲
    by dharma1 9y ago

    I hope Volta based gaming cards will happen sometime in 2018 and they don't gimp it too much. V100 kills it with mixed precision compute for deep learning because of the Tensor Cores. If you can get mixed precision training working, yo
14. 944.▲
    by dharma1 9y ago

    1080Ti doesn't have double speed FP16, Vega 64 should be more than twice as fast for half precision training/inference. The problem is that AMD is has been lacking in framework support and fast optimised kernels. This seemed promi
15. 945.▲
    by dharma1 9y ago

    the project seems very cool. related: https://medium.com/@stevewaterhouse/how-token-sales-can-be-a...
16. 946.▲
    by dharma1 9y ago

    Some stress testing with older SSD's here - https://vi-control.net/community/threads/performance-to-be-e... I use Spitfire stuff, mainly Albion, but the voice requirements these guys have is on another level
17. 947.▲
    by dharma1 9y ago

    I've been thinking about moving my Kontakt libraries to a fast SSD. Would you see a difference noticeable between this and Samsung 960 for that purpose?
18. 948.▲
    by dharma1 9y ago

    Yes exactly. The instances with 8 or 16 GPUs. Does the training time reduce linearly, is the GPU utilisation 100%, is it plug and play with TF
19. 949.▲
    by dharma1 9y ago

    Thanks! Curious how this would scale on the 8x or 16x instances
20. 950.▲
    by dharma1 9y ago

    If you're going to be running it for 24x7 for 3 years, I think it'd be worth doing the apples-to-apples comparison of buying your own V100s vs renting them from AWS. The DGX Station with 4 V100s is $70k
21. 951.▲
    by dharma1 9y ago

    The reason Nvidia quote 120 TLFOPS mixed precision on V100 is because of the new tensor core. https://devblogs.nvidia.com/parallelforall/inside-volta/
22. 952.▲
    by dharma1 9y ago

    Yep agreed. Didn't want to post the spot pricing since it changes all the time :)
23. 953.▲
    by dharma1 9y ago

    Price: p3.2xlarge - $3/hr, p3.8xlarge - $12/hr, p3.16xlarge - $25/hr These look very good for half precision training
24. 954.▲
    by dharma1 9y ago

    That photek 12" is tight. Would be nice to have curated collections in this app by people whose taste you trust, and to be able to quickly skim through them, like back in the day when you were checking out a stack of vinyl the shop own
25. 955.▲
    by dharma1 9y ago

    Good stuff. Was looking into cheap array mics with linux drivers a few times in the past but not much is available. Speech separation - Mitsubishi Research has done some pretty impressive stuff on that - http://www.merl.com/
26. 956.▲
    by dharma1 9y ago

    Very true. Which is why all the home assistants (Google Home, Amazon Echo etc) use array mics and beamforming - they get a much cleaner speech signal from far field audio, and better WER as a result
27. 957.▲
    by dharma1 9y ago

    or have more effective ways to collect tons of open source speech data. The Mozilla Common Voice project is really cool, but they should make it way easier for people to contribute. Like, adding a mic button for voice search next to their m
28. 958.▲
    by dharma1 9y ago

    You're right in that TPU's allow Google to train very large datasets faster and using less power. But I think a reasonable ASR model should be trainable with GPUs alone. The issue previously has been a lack of large enough high qu
29. 959.▲
    by dharma1 9y ago

    Kaldi is the best, there was just Tensorflow integration added which will hopefully speed up development (though I haven't seen any pretrained models for that yet). Here's a blog post - http://www.googblogs.com/kal
30. 960.▲
    by dharma1 9y ago

    Same here, it's becoming useless for that. "What are the best tools for X" questions on Quora are filled with marketing drivel from SEO specialists/employees of companies plugging their product. Probably works for a whil