I'll be giving a talk on our paper `Reverse-mode automatic differentiation and optimization of GPU kernels via Enzyme` at SC today at 3:30 CT! #SC21 sc21.supercomputing.org/presentation/?…

Quick clarification, the workshop starts at 9am EST, 8am central time, (the 10pm CST below refers to Chinese Standard Time). Check out diffprogramming.mit.edu/schedule for an up-to-date schedule and link to talks/posters!

My code's error when rushing towards a paper deadline: ``` ERROR: LoadError: UndefVarError: fun not defined ```

In this paper, researchers have developed an end-to-end GPU-to-CPU transpilation system based on Polygeist/MLIR, showing CUDA kernels from PyTorch can efficiently run and scale on the CPU-only Supercomputer Fugaku without user intervention. #HPC arxiv.org/pdf/2207.00257…

We actually won the best student paper award this time around!! 🤩🤩🥳🥳 #SC22 Massachusetts Institute of Technology (MIT) Argonne National Lab TU München

🎉Congratulations to #DOECSGF alumnus William Moses (2018-2022) for winning Best Student Paper for "Scalable Automatic Differentiation of Multiple Parallel Paradigms through Compiler Augmentation" at #SC22

Shook hands with the President of France! Was getting lunch in the city during #NeurIPS

Official a Dr after an amazing speech (featuring a flying cap) from Mark Rober

Super happy to be able to finally announce the first iteration or our new code data "ComPile", a 182B tokens/2.4TB textual IR dataset of compiled production sources. In collaboration with aidengrossman Konstantinos Parasyris Tal Ben-Nun Billy Moses José M Monsalve Diaz Mircea Trofin & Johannes Doerfert

For everyone developing, using or interested in Enzyme: Just 6 days left to submit your talk/tutorial/hackathon proposal to the annual conference for the Enzyme automatic differentiation framework! When: March 25-27, 2024 Where: MIT, Boston, MA, USA

Academic Twitter: I have a new office and I want to fill my bookshelf with great CS books (and beyond), what should I get?

Congratulations to Dr. William (Billy) Moses (Billy Moses), the 2024 ACM SIGHPC Doctoral Dissertation Award winner Moses' Tapir compiler optimizes parallel code by embedding parallelism directly into the compiler's internal representation sighpc.org/opportunities/… #HPC #SC24

Want state-of-the-art data curation, data poisoning & more? Just do gradient descent! w/ Andrew Ilyas Ben Chen Axel Feldmann Billy Moses Aleksander Madry: we show how to optimize final model loss wrt any continuous variable. Key idea: Metagradients (grads through model training)

Metagradient descent is #1 on the DataComp small *and* medium leaderboards! Really excited about the potential to design training routines by "just doing gradient descent." w/ Logan Engstrom Ben Chen Billy Moses Axel Feldmann Aleksander Madry More exciting results coming soon :)

RAPTOR is accessible now, introducing a novel numerical profiler for your #HPC/science apps. You can get it on arxiv before the SC'25 release here arxiv.org/pdf/2507.04647. Faveo + Ivan did a fantastic job and are still improving the tool. Thanks to Anshu, Akash, Wahib, Billy Moses

Thrilled to announce that our quantum computing research (Co-authored with William Moses (Billy Moses) and Neil Thompson (Neil Thompson) has finally come out in Proceedings of the IEEE (Proceedings of the IEEE) (IF 25.9), a leading journal in computer science! 🎉 Our paper, "The Quantum