Joshua Patterson

@datametrician

Engineering Sr Director leading at . Ex (#44) . Love making data science faster.

Atlanta, GA
linkedin.com/in/pattersonjo…
Joined September 2008
101 Photos and videos Photos and videos

Tweets

You blocked @datametrician

Are you sure you want to view these Tweets? Viewing Tweets won't unblock @datametrician

  1. Pinned Tweet
    Aug 14

    On 👇🏾 ran . Using 16x nodes (~6% of the cluster) we did a 10TB benchmark in ~15mins. 20x faster than 16x CPU nodes & 7x cheaper!!! Can't wait to 10x this... 100TB on half of Selene 🔥🔥🔥 Fastest cluster built in 3 weeks! And it does AI!

    2:18
    Meet Selene, the seventh most powerful supercomputer in the world. Built with DGX SuperPOD, this machine is being used to drive forward.
    2 replies 26 retweets 54 likes
    Undo
  2. Retweeted
    Sep 7

    Did you know that you can deploy Python code straight into production without rewriting it? explains how to do this and enable real-time performance, using a combination of NVIDIA and Iguazio technologies.

    3 retweets 7 likes
    Undo
  3. Retweeted
    6 hours ago

    Using RAPIDS to accelerate single-cell analysis on , analyzing one million cells is made possible in just 11 minutes.

    8 retweets 5 likes
    Undo
  4. Retweeted
    16 hours ago

    グラフデータの探索処理をcugraphで高速にできるか検証しました。 networkxと比べて300〜1800倍程度、高速化しました。

    6 retweets 6 likes
    Undo
  5. Retweeted
    11 minutes ago
    Replying to

    This sort of breaks the stranglehold Bigquery has on SQL querying over large datasets.

    3 retweets 4 likes
    Undo
  6. Retweeted
    9 hours ago

    For those who are interested, the NMF method using GPU is implemented here:

    1 reply 1 retweet 4 likes
    Show this thread
    Show this thread
    Undo
  7. Retweeted
    9 hours ago

    This work is awesome and if you use NMF you should def look at this.

    For those who are interested, the NMF method using GPU is implemented here:
    Show this thread
    1 retweet 4 likes
    Undo
  8. Retweeted
    57 minutes ago

    If you are using sklearn modules such as KDTree & have a GPU at your disposal, please take a look at sklearn compatible CuML modules. For a forthcoming paper, I got a 122x speedup with *1* line code change and 3 minutes of documentation shopping on TitanRTX

    1 reply 10 retweets 23 likes
    Undo
  9. Retweeted
    4 hours ago

    While moving cuStreamz into production at , we noticed streamz needed a few features. So we contributed checkpointing. . Learn about checkpointing and how RAPIDS leverages the larger ecosystem.

    2 replies 12 retweets 22 likes
    Undo
  10. Retweeted
    4 hours ago

    Very excited to have long-time friend joining me for DC_THURS this week. Bring your questions for Wes on Thurs am!

    1 reply 2 retweets 12 likes
    Undo
  11. Retweeted
    10 hours ago
    Replying to

    I'm really excited to play with this :)

    2 replies 3 retweets 8 likes
    Undo
  12. Retweeted
    Sep 7

    “BlazingSQL is a SQL engine on DataFrames. And it allows you to run a SQL query on GPU-accelerated DataFrames.” , Accelerating Data Science with BlazingSQL and

    3 replies 28 retweets 109 likes
    Undo
  13. Retweeted
    9 hours ago

    Very proud to work with to accelerate cancer research using GPUs. In this work we accelerated the NMF method using GPUs (via pytorch) for mutational signature analysis

    1 reply 9 retweets 27 likes
    Show this thread
    Show this thread
    Undo
  14. Retweeted
    Sep 7

    SMH…

    Same author. 🤔
    349 replies 4,930 retweets 32,597 likes
    Undo
  15. Retweeted
    Sep 5

    We are happy to have the support of as Gold sponsors, helping us bring to you—our sponsors make possible this celebration of !

    1 reply 5 retweets 12 likes
    Show this thread
    Show this thread
    Undo
  16. Sep 5

    thank you

    2:19
    Kirk Herbstreit for the win. Please listen to all of it.
    1 retweet 6 likes
    Undo
  17. Retweeted
    Sep 5

    Kirk Herbstreit for the win. Please listen to all of it.

    1,831 replies 22,686 retweets 79,285 likes
    Undo
  18. Sep 5

    Seismic facies analysis accelerated >250x with k-mean and on end to end pipelines in ! Woot! 🔥🔥🔥

    19 retweets 45 likes
    Undo
  19. Retweeted
    Aug 28

    If you want to accelerate your machine learning training with GPUs seamlessly, check out cuML: . It has pretty much similar use and machine learning algorithms as sklearn so

    1 reply 15 retweets 33 likes
    Undo
  20. Retweeted
    Sep 1

    XgboostなどのDeep Learning以外の機械学習がGPUで高速化できるcumlを試してみました。 試したのは推論のみですが約7倍程度、高速化できました。 GPUのスレッドブロックごとにデータの行の推論をしているので高速に動作していると思われています。

    5 retweets 13 likes
    Undo
  21. Retweeted
    Sep 4

    dask-image 0.4.0 is now available! Highlights include a new local threshold function, and GPU support combining with for the ndfilters and imread sub-packages. See the latest at

    1 reply 19 retweets 50 likes
    Undo

@datametrician hasn't Tweeted yet.

Loading seems to be taking a while.

Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.

    You may also like

    ·