-
-
We did a thing and made
#cuDF bigger, better, faster, stronger
. Super proud of the team for this release and it just seems to be getting better and better from here. @rapidsai 0.10 will be even more fun!
https://twitter.com/rapidsai/status/1166745659040419840 …
-
I've never previously considered using a
#GPU to unpack variable structure binary data. I'm seeing crazy speedups with@numba_jit. Now I'm getting a 258x speedup unpacking float64 (c/w np.frombuffer) ESRI shapefile points into a@rapidsai#cudf.@tomekdrabas@adamlikesaipic.twitter.com/u5b7Bb2HWI
-
For reference, in
#cuDF we had 64 threadsperblock for our `apply_rows` UDFs and in changing it to use the `.forall(...)(...)` syntax we saw a ~10x speedup from getting much better occupancy on the GPU in 0.10
-
-
-
Pearu created a nice small package to create copy-free views of arrays in Python on the GPU and on the host. Supports
#NumPy,@pandas_dev,@ApacheArrow,#xnd,@numba_jit,#CuPy, and#cudf (@rapidsai)! https://github.com/plures/arrayviews …pic.twitter.com/M6zwPRcbxv
-
I tried a larger shapefile California blocks from
@uscensusbureau TIGER/LINE. ~1 sec to unravel 40m points in 750K blocks & similar to read dbf from string into@rapidsai#cudf. Huge potential with#cupsatial here and I believe that further speedup is possible!@tomekdrabaspic.twitter.com/jQ9yr9A8CG
Prikaži ovu nit -
Second big news of the day,
@RAPIDSai is on@GCPcloud#DataProc! If you want to try multi-node multi-GPU@XGBoostProject on@dask_dev or#RAPIDS#cuDF you now can, easier than ever, on@NvidiaAI#T4 nodes. We started with@GoogleColab, now on Dataproc for scale. Try it today!pic.twitter.com/14xynqhQoD
Prikaži ovu nit -
WOOT!
@rapidsai data frame library#cuDF now#conda installable! conda install -c numba -c conda-forge -c rapidsai -c defaults cudf=0.2.0 Even more to come this week! So proud of everyone on the#RAPIDSAI team!https://twitter.com/mike_wendt/status/1056694021383512065 … -
Another bonus article, want to accelerate
@pandas_dev, add a#GPU and use@RAPIDSai#cuDF https://link.medium.com/NT0v08Zbm3 plus this scales with@dask_dev & integrates with@XGBoostProject,@PyTorch,@ChainerOfficial,@datashader, and more just like Pandas.Prikaži ovu nit -
Been using
#cudf nightly for the last few weeks to enable our Multi-Node demos. Onwards and upwards with 0.9! https://twitter.com/rapidsai/status/1166745659040419840 …
-
Interested in seeing how
@rapidsai can help accelerate your network security research and use cases? Check out our latest notebook from#kdd19. It uses#cugraph and#cudf to investigate a network graph for incidents. https://github.com/rapidsai/notebooks-contrib/tree/master/conference_notebooks/KDD_2019/notebooks/cyber … -
Woot!
@rapidsai ML library#cuML is now#conda installable! conda install -c numba -c conda-forge -c rapidsai -c defaults cuml=0.2.0 K-means and K-NN (#FAIR#FAISS) added w/#cuDF integration! I know I said this before, but more to come next week! https://twitter.com/mike_wendt/status/1058566337663041536 … -
With the addition of Java bindings in the upcoming
@rapidsai 0.8 release,#cuDF is becoming a colorful project!#CUDA,#cplusplus,#Python,#Javapic.twitter.com/JKGCblEz6U
-
I just spent an entire afternoon trying to install
@rapidsai's#cudf to run Python code on a GPU, circumventing the lack of Windows support by hosting a notebook from subsystem shell (Ubuntu), only to realize that my GPU is not supported
pic.twitter.com/CXJeQsCrFq
-
Built to scale,
@RAPIDSai 0.9 is here!#cuDF: more robust, up to 10x faster;#cuGraph: PageRank 300GB in 30 secs;#cuML: multi-node kmeans/random forest & the new Forest Inference Library that accelerates@XGBoostProject/#LightGBM inference over 30x! https://nvda.ws/30GPnq2 pic.twitter.com/9L6QbZVYRt
-
Fantastic work by
@MurrayData doing tree detection In LiDAR using@rapidsai libraries#cuDF and#cuML DBSCAN. Awesome work!pic.twitter.com/5JWTPwC4cu
-
Check @blazingdb out if you're looking for
#SQL on@rapidsai. They are heavy contributors to#cudf and an integral part of the#GPU analytics ecosystem. Super happy that I get to consistently work with them to help build the future!https://twitter.com/blazingsql/status/1113870774945714176 … -
.
@BlazingSQL: a#GPU (!) powered#SQL interface to#ETL massive datasets in#Python?! I am in@rodaramburu!
Try it on @GoogleColab via the link below
h/t @datametrician@RAPIDSAI @BlazingDB@apachearrow#cuDF#DataSciencehttps://colab.research.google.com/github/BlazingDB/bsql-demos/blob/master/colab_notebooks/blazingsql_demo.ipynb#scrollTo=aMwNKxePSwOp …
Čini se da učitavanje traje već neko vrijeme.
Twitter je možda preopterećen ili ima kratkotrajnih poteškoća u radu. Pokušajte ponovno ili potražite dodatne informacije u odjeljku Status Twittera.


