Kovács Dávid Péter

Machine Learning researcher

[Kovács Dávid Péter @davkovacs10]

💥I am happy to share our work on next gen transferable ML force fields for bio-organic / molecular chemistry.

MACE-OFF23 is a series of pretrained MACE models accurately describing molecules in vacuum as well as condensed phase with a wide range of tests
https://t.co/zTszpSXSOT

[Kovács Dávid Péter @davkovacs10]

@robert_csordas Hi! Sorry I missed the poster, are you still around tomorrow?

[Jim Fan @DrJimFan]

Hitchhiker's guide to rebranding:
- Machine learning -> statistical mechanics
- Loss function -> energy functional
- Optimize the model -> minimize free energy
- Trained model -> reached equilibrium distribution
- KL divergence -> free energy difference
- Gaussian noise -> random thermal fluctuations
- Random step -> Brownian motion
- SGD -> directional Brownian motion
- GPU -> simulated particle accelerator
- Diffusion models -> Langevin dynamics
- Reinforcement learning -> control theory
- Robotics -> physical computation
- Audio learning -> 1D signal processing
- Image learning -> 2D signal processing
- Video learning -> 3D signal processing
- Multimodal models -> multidimensional signal processing
- Sora -> learned physics engine

You're welcome

[Kovács Dávid Péter @davkovacs10]

@jigyasa_nigam @lab_COSMO @MicheleCeriotti Congratulations! 😊😊

[Kovács Dávid Péter @davkovacs10]

@jburnmurdoch Why do you think people would have voted the same way in a PR? If you know the tory vote is lost you might as well vote reform to make a point.
Why do you think labour would have focused the campaign on winning marginals instead of as many votes as possible in safe seats?

[Kovács Dávid Péter @davkovacs10]

@fiszczyp Taylor Swift?

[Chaitanya K. Joshi @chaitjo]

There's been many new Geometric GNNs in the past couple weeks combining 3D equivariance and topological ideas (simplices, cells).

I think folks should really try their best to compare to MACE (https://t.co/hy0Wo46wOb), if possible, b/c its a way to get similar capabilities.

[Kovács Dávid Péter @davkovacs10]

@CatAstro_Piyush @lars__schaaf Hi! We have a notebook linked on the mace github repo.

[Kovács Dávid Péter @davkovacs10]

Check out this great summary of our foundation model for chemistry / materials science by @lars__schaaf

This thread hints at the emergent behaviour of our model with applicability well beyond the training domain of crystals. 👇👇👇

[Lars Schaaf @lars__schaaf]

The first time I used this model I was pretty breath taken. The coolest things: with just a few lines, **anyone** can now run decent atomic scale simulations.

Try it out yourself! Details in 🧵
paper: https://t.co/c3BzaWRWvP

[Kovács Dávid Péter @davkovacs10]

@josejimlun sorry to hear. I also lost a bike ca once a year in Cambridge…

[Fabian Zills @PythonFZ]

Check out the interactive demo on Zero Shot Molecular Generation via Similarity Kernels https://t.co/u5c5O5gXOA for our preprint https://t.co/873cKlejHP and share your best molecules with us! Great collaboration with @EliRokas, @IlyesBatatia and @davkovacs10
#simgen #zndraw

[Andrej Karpathy @karpathy]

Have you ever wanted to train LLMs in pure C without 245MB of PyTorch and 107MB of cPython? No? Well now you can! With llm.c:
https://t.co/PoGTZIwASL

To start, implements GPT-2 training on CPU/fp32 in only ~1,000 lines of clean code. It compiles and runs instantly, and exactly matches the PyTorch reference implementation.

I chose GPT-2 to start because it is the grand-daddy of LLMs, the first time the LLM stack was put together in a recognizably modern form, and with model weights available.

[Kovács Dávid Péter @davkovacs10]

@jrib_ Do we know that your test set is not in their training set?

[Aditi Krishnapriyan @ask1729]

1/ Introducing a new method, the Gaunt Tensor Product, to improve the efficiency of E(3)-equivariant neural networks and speed up all sorts of equivariant operations.

Accepted at #ICLR2024 as a spotlight paper: https://t.co/9RYDgI3y7A

With Shengjie Luo and Tianlang Chen

[Kovács Dávid Péter @davkovacs10]

And to try the models quickly just install mace from https://t.co/XEYMuoyT8n

pip install mace-torch

And import them as an ASE calculator

calc = mace_off(model="medium", device='cuda')

[Kovács Dávid Péter @davkovacs10]

We present 3 models of increasing accuracy parameterised for 10 (H C N O P S F Cl Br I) chemical elements trained to reproduce accurate (hybrid DFT with large basis) QM energies and forces. The models are available at https://t.co/nMOGN8mCEd

[Kovács Dávid Péter @davkovacs10]

As well as tagging all other co-authors who were integral to this project! @IlyesBatatia Nick Browning @venkatkapil24 @w_c_witt @ColeGroupNCL @joshhorton93 @MagdauIoan and Gabor of course!