Discover and read the best of Twitter Threads about #jax

Most recents (13)

There's a new programming language in town - it's Mojo! I'm more than a little excited about it. It's Python, but with none of Python's problems.

You can write code as fast as C, and deploy small standalone applications like C.

My post is below, and a šŸ§µ
fast.ai/posts/2023-05-ā€¦
Python is the language that I have used for nearly all my work over the last few years. It is a beautiful language. It has an elegant core on which everything else is built.

But it comes with a downside: performance. It's thousands of times slower than C.
Python programmers learn to avoid using Python for the implementation of performance-critical sections, instead using Python wrappers over C, FORTRAN, Rust, etc.

But this ā€œtwo-languageā€ approach has serious downsides. It's complex, hard to debug or profile, & hard to deploy.
Read 11 tweets
Fresh!!
Fully differentiable HĆ¼ckel model using #JAX.
i) We optimize the atom types for different molecular frameworks to find the molecule with the targeted property (HOMO-LUMO) šŸ§µ (1/3)
@kjelljorner @robpollice @A_Aspuru_Guzik @chemuoft
#compchem
ii) Parameter optimization a la machine learning for diff. molecular property, for example, polarizability
30% improvement with only 100 DFT training data.
šŸ§µ (2/3) Image
#arxiv arxiv.org/abs/2211.16763 #github
(inverse molecular design)
github.com/RodrigoAVargasā€¦
(parameter optimization)
github.com/RodrigoAVargasā€¦
šŸ§µ (3/3)
Read 3 tweets
2/2 4 pm EST: Estos son los mensajes clave para la tormenta subtropical #Nicole. Consulte hurricanes.gov para obtener mƔs detalles.
1/2 #Florida #Georgia
Follow instructions from local officials as Nicole nears the coast
Siga las instrucciones de los funcionarios locales a medida que Nicole se acerca a la costa.
Read 15 tweets
"Eqxvision" is a new #JAX library for computer vision!āš”ļø

Reddit: reddit.com/r/MachineLearnā€¦
GitHub: github.com/paganpasta/eqxā€¦

Built by @|paganpasta, it's now at feature-parity with torchvision for classification models. (With segmentation, object detection etc. on the way!)

1/2
It's been super cool watching this project grow, and a big thank-you to @|paganpasta for all their upstream contributions to Equinox. (Read: fixing my bugs!šŸ›šŸŖ²)

2/2
I guess I'm taking a leaf out of @DynamicWebPaige's book and tweeting about #JAX ecosystem stuff... :D
Read 3 tweets
āš”ļø My PhD thesis is on arXiv! āš”ļø

To quote my examiners it is "the textbook of neural differential equations" - across ordinary/controlled/stochastic diffeqs.

w/ unpublished material:
- generalised adjoint methods
- symbolic regression
- + more!

arxiv.org/abs/2202.02435

všŸ§µ 1/n
If you follow me then there's a decent chance that you already know what an NDE is. (If you don't, go read the introductory Chapter 1 to my thesis haha -- it's only 6 pages long.) Put a neural network inside a differential equation, and suddenly cool stuff starts happening.

2/n
Neural differential equations are a beautiful way of building models, offering:
- high-capacity function approximation;
- strong priors on model space;
- the ability to handle irregular data;
- memory efficiency;
- a foundation of well-understand theory.

3/n
Read 21 tweets
Watch me code a Neural Network from Scratch in pure JAX! šŸ„³ in this 3rd video of the Machine Learning with JAX series.

YouTube:
GitHub: github.com/gordicaleksa/gā€¦

@DeepMind @GoogleAI @jakevdp @froystig @SingularMattrix @cdleary

#jax
In this video, I build an MLP (multi-layer perception) and train it as a classifier on MNIST (although it's trivial to use a more complex dataset) - all this in pure JAX (no Flax/Haiku/Optax).

2/
I then add cool visualizations such as:
* Visualizing MLP's learned weights
* Visualizing embeddings of a batch of images in t-SNE
* Finally, we analyze the dead neurons

3/
Read 5 tweets
2nd video of the JAX tutorials series is out! šŸ„³

YT:
Code: github.com/gordicaleksa/gā€¦

@DeepMind @GoogleAI @jakevdp @froystig @SingularMattrix @cdleary @huggingface

#jax
In this one, we learn all the necessary components to train complex ML models in parallel on multiple machines.

2/
As always - open up the notebook in Colab to make this an active learning experience (code + video):

github.com/gordicaleksa/gā€¦

3/
Read 3 tweets
End-to-end learning of multiple sequence alignments with differentiable Smith-Waterman
biorxiv.org/content/10.110ā€¦
A fun collaboration with Samantha Petti, Nicholas Bhattacharya, @proteinrosh, @JustasDauparas, @countablyfinite, @keitokiddo, @srush_nlp & @pkoo562 (1/8)
Many methods like GREMLIN, MSA_transformer, RosetTTAFold and AlphaFold rely on input MSA generated by non-differentiable methods. (2/8)
We ask the question, what if we make the red arrow differentiable and optimize end-to-end. (3/8)
Read 12 tweets
Toy MuJoCo + Box2d envs in OpenAI Gym are moving to #brax! 100x GPU/TPU speedup + purely pythonic + jax/pytorch-enabled ready to be unleashed! An exciting news for #brax #braxlines #jax teams. Also check out #composer, where I am adding more demos github.com/openai/gym/issā€¦
#braxlines:

#composer: github.com/google/brax/trā€¦

arxiv: arxiv.org/abs/2110.04686

#brax still cannot (and probably won't ever) match the full specs with mujoco/pybullet. But esp with open-sourcing plans of mujoco, excited to see where could be synergies.
Good to see a lot of large-scale, algorithmic deep RL researchers are aligned: "I personally believe that hardware accelerator support is more important, hence choosing Brax."
Read 4 tweets
A thread on our latest optimizers work! We tune Nesterov/Adam to match performance of LARS/LAMB on their more commonly used workloads. We (@jmgilmer, Chris Shallue, @_arohan_, @GeorgeEDahl) do this to provide more competitive baselines for large-batch training speed measurements Image
We are **not** trying to prove that any optimizer is better than any other (more on that later). However, we believe that well tuned baselines are very important, especially in optimization where there are so many confounding factors.
In general, we found that LR schedules were the most impactful part of the pipeline to tune (that is, once we got all the model and dataset details to be the same!).
Read 10 tweets
The Lie group SU(N) is very important in fundamental particle physics as a Gauge symmetry group. Building SU(N) Gauge-equivariant models is key to make ML useful in this field. It was not easy, but WE DID IT!
SU(N) gauge equivariant flows for LatticeQCD:
arxiv.org/abs/2008.05456
This work is the result of the most amazing collaboration between @MIT, @nyuniversity and @DeepMind and I couldn't be more thankful and proud to have the opportunity to work with this team.
The team: Denis Boyda, Gurtej Kanwar, @sracaniere,@DaniloJRezende,@msalbergo,@kylecranm, Daniel Hackett, Phiala Shanahan
Read 24 tweets
šŸ“¢šŸ“¢Dopamine now also runs on #JAX !!šŸ“¢šŸ“¢

Happy to announce that our #RL library, dopamine, now also has JAX implementations of all our agents, including a new agent: QR-DQN!
github.com/google/dopamine
1/X
The JAX philosophy goes very well with that of Dopamine: flexibility for research without sacrificing simplicity.
I've been using it for a while now and I've found its modularity quite appealing in terms of simplifying some of the more difficult aspects of the agents.
2/X
Consider the projection operator used for the C51 and Rainbow agents, which is a rather complex TF op. This went from 23 lines of fairly complex TF code to 9 lines of JAX code that is more straightforward.
3/X
Read 9 tweets
Good morning! Weā€™ve got the incredible @CynthiaNixon firing up our Central Florida grassroots volunteers today! #FloridaForBernie #Bernie2020
Getting ready for the onslaught of volunteers coming who are ready to launch a canvass out of Miami Shores! #FloridaForBernie #Bernie2020
Miami for Bernie volunteer Nikola cast her ballot today for Bernie and brought her daughters to watch their mom vote for not just their future, but also the future of those who are less fortunate. #NotMeUs #IVoted #Bernie2020 #FloridaForBernie
Read 62 tweets

Related hashtags

Did Thread Reader help you today?

Support us! We are indie developers!


This site is made by just two indie developers on a laptop doing marketing, support and development! Read more about the story.

Become a Premium Member ($3.00/month or $30.00/year) and get exclusive features!

Become Premium

Too expensive? Make a small donation by buying us coffee ($5) or help with server cost ($10)

Donate via Paypal Become our Patreon

Thank you for your support!