AI Paper Atlas

Research map / Large language models / Language, safety and interpretability

Generating QM1B with PySCF$_{\text{IPU}}$

Alexander Mathiasen, Hatem Helal, Kerstin Klaeser, Paul Balanca, Josef Dean, Carlo Luschi, Dominique Beaini, Andrew W Fitzgibbon, Dominic Masters

NeurIPS 2023 · Large language models › Language, safety and interpretability · approach: Data and evaluation methodology · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
pretrain on QM1B before fine-tuning on QM9SchNet 9M · QM9 (88k subset)validation MAE (meV): 54.13 -> 30.2Improved
run DFT on IPUs instead of CPU PySCFPySCF · DFT data generation, one POD16DFTs per second: 82 -> 228Improved
use float32 instead of float64 in libcint (ablation)PySCF_IPU · libcint on IPUspeed: around 6x fasterImproved
use float32 DFT instead of float64 (ablation)PySCF_IPU · HL gap vs PySCF float64MAE (meV): 0.2Improved
use STO-3G labels instead of 6-31G* labels for QM9 (ablation)SchNet · QM9 training loss curveslossNo clear effect
use RDKit ETKDG conformer positions instead of equilibrium positions for QM9 (ablation)SchNet · QM9 trainingMAEHurt
increase QM1B training subset sizeSchNet 9M · QM1B validationvalidation MAEImproved
increase QM1B training size beyond about 100M samplesSchNet 9M · QM1B validationvalidation MAENo clear effect
exploit the 8x symmetry of the electron repulsion integralsPySCF_IPU · DFT computeFLOPs and memory: 8x reductionImproved
recompute electron repulsion integrals every iteration instead of precomputingPySCF_IPU · cycle countcycles: 105M -> 395.7MHurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy