Peter Entschev retweeted
Optuna 2.0, the next major evolution of Optuna is here!
2.0 introduces hyperparameter importance to see which hyperparameters matter most, Hyperband pruning, and many performance improvements!
See what’s new ↓
medium.com/optuna/optuna-…
2.0 introduces hyperparameter importance to see which hyperparameters matter most, Hyperband pruning, and many performance improvements!
See what’s new ↓
medium.com/optuna/optuna-…
3
104
259
Peter Entschev retweeted
Wow! 19.5x faster TPCx-BB performance – performed with NVIDIA DGX A100 using RAPIDS for data processing and machine learning, AND UCX open source libraries for ultra-fast communication.
Check out the NVIDIA blog of the announcement: blogs.nvidia.com/blog/2020/06/2…
For a deeper dive into how this was possible and what we're planning for future improvements, check out the RAPIDS blog: medium.com/rapids-ai/no-m…
/2
0
12
16
Peter Entschev retweeted
Tomorrow, I’m giving a webinar/Q&A on the current state of @RAPIDSai. nvidia.com/en-us/gtc/sess… register to see how we scale to 10TB problems with @dask_dev @blazingsql @openucx & @numba_jit. Plus, we have new integrations with @plotlygraphs @raydistributed & more at #GTC2020!
2
28
49
Peter Entschev retweeted
TPCx-BB at 10TB run with @rapidsai @openucx @dask_dev @numba_jit @ChainerOfficial #cupy @blazingsql is open source here: github.com/rapidsai/tpcx-…. Beyond proud of the team that delivered these results in a short period of time with the @PyData ecosystem.
2
35
72
Peter Entschev retweeted
~2 years ago UCX-Py concept was presented at UCX hackathon as a “crazy” idea - let’s enable high performance communication for Python? and (!)it has to support any combination of GPU/CPU/remote/local/server/client! It is really amazing what progress has been made by UCX-Py team.
Distributed #DataScience processing on @rapidsai keeps getting faster. See the latest benchmarks from our high-performance #Python communication library UCX-Py. nvda.ws/3eteqEu
1
11
22
Peter Entschev retweeted
Distributed #DataScience processing on @rapidsai keeps getting faster. See the latest benchmarks from our high-performance #Python communication library UCX-Py. nvda.ws/3eteqEu
1
22
62
Peter Entschev retweeted
Thinking about hosting your own #hackathon with RAPIDS? Setting up @nvidia #gpu #datascience environments for hackathons is quick to build and easy to administer so people can enjoy RAPIDS and @ProjectJupyter. medium.com/rapids-ai/sett…
2
16
35
Peter Entschev retweeted
I just published Python Pandas at Extreme Performance medium.com/p/python-panda… @kubernetesio @nuclioio @rapidsai #serverless
2
30
80
Peter Entschev retweeted
Accelerating #datascience with #GPU is only the start. For optimal performance, #RDMA networks, #NVMe storage, @kubernetesio, and @Docker containers are needed to fully power @RAPIDSAI and the rest of the #analytics ecosystem on @NVIDIA GPU. Read more here medium.com/@patterson.jos…
2
31
64
Peter Entschev retweeted
Dask 2.2 is out!
blog.dask.org/2019/08/02/das…
We've rewritten our Parquet management, spun out I/O handling to a separate package, use async/await internally, and add support for Hyperband. (among hundreds of other things)
blog.dask.org/2019/08/02/das…
We've rewritten our Parquet management, spun out I/O handling to a separate package, use async/await internally, and add support for Hyperband. (among hundreds of other things)
0
57
138
Peter Entschev retweeted
Dive into this new introductory notebook that shows the ease of going between @rapidsai #cuDF and @ChainerOfficial #CuPy all in #Python on @nvidia #GPUS. #RAPIDSin10 nvda.ws/317s1cR
0
7
17
Peter Entschev retweeted
While scaling data frames with @dask_dev & #cuDF, many want to do the same with arrays. Instead of using a new library, here’s how to use @ChainerOfficial #CuPy and examples of speed up on @NvidiaAI #GPUs nvda.ws/2JXHu8n
0
6
14
Peter Entschev retweeted
Head to our Medium page to read our latest blog post by @_DPeterK_
medium.com/informatics-la…
medium.com/informatics-la…
0
7
11
Peter Entschev retweeted
Meet our speaker @PeterEntschev, senior system software engineer at @NvidiaAI where he works on @rapidsai.
He will speak on distributing computing with GPUs and @dask_dev, and will participate in the "Dataframe summit @ EuroSciPy".
More info and tickets: euroscipy.org/2019/
He will speak on distributing computing with GPUs and @dask_dev, and will participate in the "Dataframe summit @ EuroSciPy".
More info and tickets: euroscipy.org/2019/
0
12
27
My EuroPython 2019 slides on Distributed Multi-GPU Computing with Dask, CuPy and RAPIDS are now available at ep2019.europython.eu/media/conferen…
1
23
47
Peter Entschev retweeted
Slides from my #SciPy2019 talk:
"Refactoring the SciPy Ecosystem for Heterogeneity"
matthewrocklin.com/slides/scipy-2…
"Refactoring the SciPy Ecosystem for Heterogeneity"
matthewrocklin.com/slides/scipy-2…
2
35
72
Peter Entschev retweeted
This makes it all worth it! I can’t be prouder of the @rapidsai developers who made this possible, and we’re just getting started!
At #VBTransform, @WalmartLabs Chief Data Officer, @BillGroves13, explained how his team is working with #NVIDIA to speed up data processing and machine learning using @rapidsai during a fireside chat with @datametrician. Learn more: nvda.ws/2S8o6cu
0
8
31
Peter Entschev retweeted
Join @dante_dgd, @keithjkraus, and Nick Becker from the @NVIDIAAI RAPIDS team and @scopatz from @quansight to learn how you can execute #datascience pipelines entirely on #GPUs with RAPIDS. nvda.ws/2JtiCoS #SciPy2019
0
5
14