py-spy 0.4.0 has just been released - and adds support for Python 3.12/3.13: github.com/benfred/py-spy/re…
Ben Frederickson retweeted
Want to speed up your Pandas code by 10-1000x?
With no code change?
The folks from @nvidia have created cuDF pandas accelerator mode. By using this line in Jupyter, you automatically leverage your GPU to run Pandas code:
%load_ext cudf.pandas
From command-line:
python -m cudf.pandas yourscript.py
(It also accelerates 3rd party libraries that leverage Pandas.)
Check out this demo notebook: nvda.ws/3rNqETS
I've had a chance to use the pre-release version and am very impressed!
Ben Frederickson retweeted
Check out this (free) webinar. Especially relevant for users of Pandas. 🤯
Have you checked out our lineup for the AI and Data Science Virtual Summit? On 11/8, we'll be joined by @fastdotai founding researcher @jeremyphoward, author of "Effective Pandas" @__mharrison__, and many more. Register for free today: nvda.ws/497s7Fk
Ben Frederickson retweeted
I wrote a blog post about making a #Python library 100x faster with less than 100 lines of Rust
@rustlang
ohadravid.github.io/posts/20…
Ben Frederickson retweeted
In #recsys you might want to predict different types of user events (click, like, share, purchase).
In this post we introduce multi-task learning (MTL) for ranking models and how to easily build and train such models using #Merlin and #tensorflow.
medium.com/nvidia-merlin/bui…
Ben Frederickson retweeted
Love to see the new RAPIDS release 23.02 out now! Awesome improvements for anyone who wants their data science code to run faster in Python (medium.com/rapids-ai/rapids-…). We also overhauled the rapids.ai/ site. Hopefully it's much easier to find things now!
Nice blog post showing how to use py-spy to diagnose a hanging python process : tomaugspurger.github.io/post…
Ben Frederickson retweeted
An in-depth comparison of using
• Petastorm
• TensorFlow
• Merlin Dataloader
for loading data on DataBricks!
Merlin is 10x - 1000x faster 🙂
Thank you so much @Andompesta90 for this fascinating study!
Read more here: andompesta.github.io/2022/11…
Ben Frederickson retweeted
For all my 🇺🇸 friends getting over their 🦃hangovers here's a black friday blog for you. Ever wanted to scale your recommender system to the Terabyte level. Check out this amazing work by Hao Wu and Deya Fu on the Merlin Distributed Embeddings library: developer.nvidia.com/blog/fa…
Ben Frederickson retweeted
これクッソ早かった。pytorchのdataloaderが200M件のデータのロードに3分かかったのに対し、merlin-dataloaderだとわずか500msだった。400倍の高速化。
github.com/NVIDIA-Merlin/NVT…
Ben Frederickson retweeted
Merlin Dataloader is 119x faster than my own PyTorch Dataset + Dataloader combo!
This is revolutionary for tabular data 🥳
Let's take a closer look at what is going on.
Great blog post by my colleague @radekosmulski :
I am launching a new blog -- TabularMusings 🥳
Here is the first blog post: tabularmusings.com/posts/fea…
And here is the technology I am using and the reasons for starting the blog:
Ben Frederickson retweeted
Very interesting work & Q&A on parameter tuning. And a special shout to @benfrederickson for being a great session chair! #RecSys2022
And the last (but not least) talk of Session 8C (Models and Learning II) at #RecSys2022
"Revisiting the Performance of iALS on Item Recommendation Benchmarks" by Steffen Rendle (Google Research)
Great thread on using py-spy to improve GPU utilization:
Ben Frederickson retweeted
Before I start a new project or build a large feature I like to spend some time outlining the announcement blog post/talk abstract.
This helps me figure out exactly which bits I'm excited to tell folks about so I can focus my efforts there!
jacobtomlinson.dev/posts/202…
Just released a new version of py-spy - which adds support for python 3.11 : github.com/benfred/py-spy/re…
Ben Frederickson retweeted
Wrote a short piece dissecting the allure of complexity and defending the beauty of simple methods and systems. Also, how to reframe our thinking of both.
eugeneyan.com/writing/simpli…
Ben Frederickson retweeted
H&Mコンペの解法共有スライドで知ったこのライブラリを業務で使ってみたんだけど、何も考えずともGPUで学習出来たりして便利だった。
最初なのでとりあえず控えめに4000万行ぐらいデータ突っ込んでみたんだけど10秒ぐらいでfit終わってびっくりした。
github.com/benfred/implicit
