RoundPipe achieves near-zero-bubble pipeline parallelism for LLM training on consumer GPUs by dynamically dispatching computation stages round-robin, yielding 1.48-2.16x speedups and enabling 235B model fine-tuning on 8x RTX 4090.
super hub Mixed citations
PyTorch: An Imperative Style, High-Performance Deep Learning Library
Mixed citation behavior. Most common role is background (53%).
167
Pith papers citing it
Background
53% of classified citations
abstract
Deep learning frameworks have often focused on either usability or speed, but not both. PyTorch is a machine learning library that shows that these two goals are in fact compatible: it provides an imperative and Pythonic programming style that supports code as a model, makes debugging easy and is consistent with other popular scientific computing libraries, while remaining efficient and supporting hardware accelerators such as GPUs. In this paper, we detail the principles that drove the implementation of PyTorch and how they are reflected in its architecture. We emphasize that every aspect of PyTorch is a regular Python program under the full control of its user. We also explain how the careful and pragmatic implementation of the key components of its runtime enables them to work together to achieve compelling performance. We demonstrate the efficiency of individual subsystems, as well as the overall speed of PyTorch on several common benchmarks.
hub tools
citation-role summary
background 18
method 11
dataset 1
citation-polarity summary
claims ledger
- abstract Deep learning frameworks have often focused on either usability or speed, but not both. PyTorch is a machine learning library that shows that these two goals are in fact compatible: it provides an imperative and Pythonic programming style that supports code as a model, makes debugging easy and is consistent with other popular scientific computing libraries, while remaining efficient and supporting hardware accelerators such as GPUs. In this paper, we detail the principles that drove the implementation of PyTorch and how they are reflected in its architecture. We emphasize that every aspect o
co-cited works
representative citing papers
Looped transformers with recall and outer normalization produce reachable, input-dependent fixed points with stable gradients, enabling generalization, while those without recall cannot; a new internal recall variant performs competitively or better.
First in-orbit demonstration of a DRL-trained AI satellite attitude controller that performs robust inertial pointing after sim-to-real transfer.
A two-pass optimization framework with polynomial-based simulation discovers heralded ballistic circuits for 3-5 qubit graph states achieving up to 7.5x higher success probabilities than fusion baselines, including first known circuits for some 5-qubit states.
Task vectors from weight differences allow arithmetic operations to edit pre-trained models, improving multiple tasks simultaneously and enabling analogical inference on unseen tasks.
Quantitative Bayesian inference using a deep-learning emulator detects 0.018-0.020 M_sun of helium in the Type Ic supernova 2014L.
ffortissimo is a JAX-based freeform forward-modeling pipeline that fits complex dust distributions and infers scattering properties in KLIP-reduced images of circumstellar disks such as HR 4796A.
A matrix-free, GPU-compatible PyTorch implementation of phase-field fracture with explicit dynamics, custom differentiable implicit damage solve, benchmarks on dynamic and quasi-static cases, and inverse recovery of fracture energy G_c via L-BFGS.
A Set-Transformer architecture with self-attention encodes Pauli-string correlations, optimizes via commutation objective, and finds symmetries with near-deterministic success on physical models like Ising and Toric code.
A new lattice method recasts SIGW integrals as FFT convolutions to compute fully non-Gaussian spectra in seconds with ~10% error on a radiation-dominated background.
ARBITER models reasoning trajectory basins in test-time sampling and uses model-internal signals to correct majority-vote failures, recovering part of the oracle gap on math benchmarks.
Hybrid TimesFM plus ridge regression on covariates forecasts 1-MeV electron flux with average R² of 0.9 on out-of-sample 2024 data, outperforming linear regression, CNN, LSTM and Transformer models.
A neural network trained on simulations infers stripping times for Sagittarius stream stars from phase-space data, measuring a 0.3 dex/Gyr metallicity gradient and estimating ages for globular clusters such as Pal 12 and NGC 2419.
Events trigger on-the-fly LoRA module generation via hypernetworks over a shared team policy in MARL, paired with a Neural Manifold Diversity metric, enabling sequential role reassignment while preserving reward maximization.
Vibrational mode graphs from molecular dynamics enable sequence-free protein function prediction via graph neural networks, with entrainment improving signals for collective dynamics.
Dingo-Pop uses a transformer to perform amortized, end-to-end population inference from GW strain data in seconds, bypassing per-event Monte Carlo sampling.
Learning in low-rank RNNs reduces to an exact low-dimensional ODE system in overlap space, where loss-invisible overlaps encode training history without affecting function.
Dynamical magnetotropic susceptibility k(ω) acts as a probe of uniform spin and charge fluctuations, with its static scaling in α-RuCl3 arising specifically from dominant Kitaev interactions in the models examined.
Transformer networks sample up to 180x180 2D Ising systems and 64x64 Edwards-Anderson systems by generating spin groups with probability approximations, yielding ~20x higher effective sample size than prior neural samplers at criticality.
A single end-to-end Transformer model unifies stellar labels from heterogeneous spectroscopic surveys into a self-consistent scale without post-hoc recalibration.
Parametric neural networks learn likelihood ratios to infer top-philic scalar resonances from dip patterns caused by signal-background interference in hadron collider data.
A graph-conditioned meta-optimizer learns QAOA parameter trajectories from one problem class and transfers them to others, yielding better initializations than standard methods in an empirical study of 64 settings.
SMC forgets its initial condition geometrically in the jump chain and as 1/ℓ in continuous genetic distance, justifying independent-locus approximations.
Concept Graph Convolutions perform message passing on node concepts to increase interpretability of graph neural networks without losing task performance.
citing papers explorer
-
Efficient Training on Multiple Consumer GPUs with RoundPipe
RoundPipe achieves near-zero-bubble pipeline parallelism for LLM training on consumer GPUs by dynamically dispatching computation stages round-robin, yielding 1.48-2.16x speedups and enabling 235B model fine-tuning on 8x RTX 4090.
-
Stability and Generalization in Looped Transformers
Looped transformers with recall and outer normalization produce reachable, input-dependent fixed points with stable gradients, enabling generalization, while those without recall cannot; a new internal recall variant performs competitively or better.
-
LeLaR: The First In-Orbit Demonstration of an AI-Based Satellite Attitude Controller
First in-orbit demonstration of a DRL-trained AI satellite attitude controller that performs robust inertial pointing after sim-to-real transfer.
-
Automated discovery of heralded ballistic graph state generators for fusion-based photonic quantum computation
A two-pass optimization framework with polynomial-based simulation discovers heralded ballistic circuits for 3-5 qubit graph states achieving up to 7.5x higher success probabilities than fusion baselines, including first known circuits for some 5-qubit states.
-
Editing Models with Task Arithmetic
Task vectors from weight differences allow arithmetic operations to edit pre-trained models, improving multiple tasks simultaneously and enabling analogical inference on unseen tasks.
-
Traces of Helium Detected in Type Ic Supernova 2014L
Quantitative Bayesian inference using a deep-learning emulator detects 0.018-0.020 M_sun of helium in the Type Ic supernova 2014L.
-
ffortissimo: A Freeform Forward-Modeling Pipeline for High-Contrast Images of Circumstellar Disks Based on Automatic Differentiation
ffortissimo is a JAX-based freeform forward-modeling pipeline that fits complex dust distributions and infers scattering properties in KLIP-reduced images of circumstellar disks such as HR 4796A.
-
A matrix-free, differentiable PyTorch solver for phase-field fracture: Formulation, benchmarks, and inverse analysis
A matrix-free, GPU-compatible PyTorch implementation of phase-field fracture with explicit dynamics, custom differentiable implicit damage solve, benchmarks on dynamic and quasi-static cases, and inverse recovery of fracture energy G_c via L-BFGS.
-
Attention-based optimizer for symmetry finding
A Set-Transformer architecture with self-attention encodes Pauli-string correlations, optimizes via commutation objective, and finds symmetries with near-deterministic success on physical models like Ising and Toric code.
-
A Fast Method to Compute Scalar Induced Gravitational Waves on a Lattice with Primordial Non-Gaussianities
A new lattice method recasts SIGW integrals as FFT convolutions to compute fully non-Gaussian spectra in seconds with ~10% error on a radiation-dominated background.
-
ARBITER: Reasoning Trajectory Basins and Majority Vote Failures in Test-Time Sampling
ARBITER models reasoning trajectory basins in test-time sampling and uses model-internal signals to correct majority-vote failures, recovering part of the oracle gap on math benchmarks.
-
Forecasting megaelectron-volt electron flux in the Earth's outer radiation belt using supervised machine learning algorithms and a timeseries foundation model
Hybrid TimesFM plus ridge regression on covariates forecasts 1-MeV electron flux with average R² of 0.9 on out-of-sample 2024 data, outperforming linear regression, CNN, LSTM and Transformer models.
-
Reconstructing the Stripping History of the Sagittarius Stream with Neural Networks
A neural network trained on simulations infers stripping times for Sagittarius stream stars from phase-space data, measuring a 0.3 dex/Gyr metallicity gradient and estimating ages for globular clusters such as Pal 12 and NGC 2419.
-
Events as Triggers for Behavioral Diversity in Multi-Agent Reinforcement Learning
Events trigger on-the-fly LoRA module generation via hypernetworks over a shared team policy in MARL, paired with a Neural Manifold Diversity metric, enabling sequential role reassignment while preserving reward maximization.
-
Frequency-Space Mechanics: A Sequence and Coordinate-Free Representation for Protein Function Prediction
Vibrational mode graphs from molecular dynamics enable sequence-free protein function prediction via graph neural networks, with entrainment improving signals for collective dynamics.
-
End-to-End Population Inference from Gravitational-Wave Strain using Transformers
Dingo-Pop uses a transformer to perform amortized, end-to-end population inference from GW strain data in seconds, bypassing per-event Monte Carlo sampling.
-
Learning reveals invisible structure in low-rank RNNs
Learning in low-rank RNNs reduces to an exact low-dimensional ODE system in overlap space, where loss-invisible overlaps encode training history without affecting function.
-
Dynamical magnetotropic susceptibility as a new probe of Kitaev materials and beyond
Dynamical magnetotropic susceptibility k(ω) acts as a probe of uniform spin and charge fluctuations, with its static scaling in α-RuCl3 arising specifically from dominant Kitaev interactions in the models examined.
-
Sampling two-dimensional spin systems with transformers
Transformer networks sample up to 180x180 2D Ising systems and 64x64 Edwards-Anderson systems by generating spin groups with probability approximations, yielding ~20x higher effective sample size than prior neural samplers at criticality.
-
Homogeneous Stellar Parameters from Heterogeneous Spectra with Deep Learning
A single end-to-end Transformer model unifies stellar labels from heterogeneous spectroscopic surveys into a self-consistent scale without post-hoc recalibration.
-
Big Dipper, Help Me Find A Way -- Dip-hunting at hadron colliders
Parametric neural networks learn likelihood ratios to infer top-philic scalar resonances from dip patterns caused by signal-background interference in hadron collider data.
-
Graph-Conditioned Meta-Optimizer for QAOA Parameter Generation on Multiple Problem Classes
A graph-conditioned meta-optimizer learns QAOA parameter trajectories from one problem class and transfers them to others, yielding better initializations than standard methods in an empirical study of 64 settings.
-
Rates of forgetting for the sequentially Markov coalescent
SMC forgets its initial condition geometrically in the jump chain and as 1/ℓ in continuous genetic distance, justifying independent-locus approximations.
-
Concept Graph Convolutions: Message Passing in the Concept Space
Concept Graph Convolutions perform message passing on node concepts to increase interpretability of graph neural networks without losing task performance.
-
A Non-stationary, Amortized, Transfer Learning Approach for Modeling Italian Air Quality
A neural network learns non-stationary anisotropic correlations from gridded CTM outputs and transfers the structure via LatticeKrig basis functions to station data for refined fine-scale NO2 predictions with uncertainty.
-
Probing the 3D Structures of Supernovae through IR Signatures of CO and SiO
MOFAT applied to SN2024ggi shows CO triggering inner SiO formation with a receding edge, order-of-magnitude mass drop, clumping signatures, and no dust formation.
-
Partitioning Unstructured Sparse Tensor Algebra for Load-Balanced Parallel Execution
A new partitioning algorithm that provably load-balances arbitrary sparse tensor algebra expressions by generalizing parallel merging to multi-operand, multi-dimensional hierarchical structures, implemented in a compiler framework.
-
Tensor Memory Engine: On-the-fly Data Reorganization for Ideal Locality
The Tensor Memory Engine provides on-the-fly data reorganization to achieve ideal memory locality for CPU computations in edge systems.
-
Dual Triangle Attention: Effective Bidirectional Attention Without Positional Embeddings
Dual Triangle Attention achieves effective bidirectional attention with built-in positional inductive bias via dual triangular masks, outperforming standard bidirectional attention on position-sensitive tasks and showing strong masked language modeling results with or without positional embeddings.
-
How pore-scale disorder controls fluid stretching in porous media
Pore-scale disorder accelerates fluid stretching in porous media, producing quadratic time growth and faster mixing than the linear growth seen in ordered structures.
-
Vibe Coding XR: Accelerating AI + XR Prototyping with XR Blocks and Gemini
XR Blocks supplies an LLM-optimized Reality Model and Vibe Coding XR workflow that converts high-level prompts into working physics-aware XR applications with high one-shot success.
-
Polarized Target Nuclear Magnetic Resonance Measurements with Deep Neural Networks
Deep neural networks reduce fitting uncertainties in CW-NMR polarization measurements for dynamically polarized targets.
-
RLGT: A reinforcement learning framework for extremal graph theory
RLGT is a modular reinforcement learning framework for extremal graph theory that handles undirected, directed, looped, and multi-colored graphs to facilitate future research.
-
Reduced-Order Surrogates for Forced Flexible Mesh Coastal-Ocean Models
Koopman autoencoders with forcings and temporal unrolling deliver accurate year-long predictions for coastal-ocean models at 300-1400x speedup, outperforming POD in two of three cases.
-
Cobble: Compiling Block Encodings for Quantum Computational Linear Algebra
Cobble is a domain-specific language for quantum block encodings that compiles high-level matrix expressions to optimized circuits using analyses and quantum singular value transformation, achieving 2.6x-25.4x speedups over unoptimized baselines on benchmarks.
-
Atomistic Machine Learning with Irreducible Cartesian Natural Tensors
CarNet develops irreducible Cartesian natural tensors and an equivariant model that matches leading spherical-tensor performance for ML interatomic potentials and high-rank tensor predictions like elastic constants.
-
pop-cosmos: Star formation over 12 Gyr from generative modelling of a deep infrared-selected galaxy catalogue
A score-based diffusion generative model on deep infrared galaxy photometry yields a star formation rate density peaking at z=1.3 and shows distinct non-parametric star formation histories plus AGN activity peaking during the quenching transition of massive galaxies.
-
Meson spectroscopy of exotic symmetries of Ising criticality in Rydberg atom arrays
Rydberg arrays realize Ising criticality with E8 mass spectra in chains and first signatures of D8^(1)-organized bound states from interchain confinement in ladders.
-
GraphGDel: Constructing and Learning Graph Representations of Genome-Scale Metabolic Models for Growth-Coupled Gene Deletion Prediction
GraphGDel builds graph representations from constraint-based metabolic models and trains a deep learning framework integrating graph structure with gene and metabolite sequences to predict growth-coupled gene deletions, showing accuracy gains of 4-16% over baselines on three models.
-
KernelBench: Can LLMs Write Efficient GPU Kernels?
KernelBench shows that even the best current LLMs generate correct and faster-than-baseline GPU kernels in fewer than 20 percent of realistic ML workloads.
-
Clustering in pure-attention hardmax transformers and its role in sentiment analysis
Hardmax transformers converge to leader-determined clusters, enabling an interpretable model for sentiment analysis.
-
Diffusion Models Beat GANs on Image Synthesis
Diffusion models with architecture improvements and classifier guidance achieve superior FID scores to GANs on unconditional and conditional ImageNet image synthesis.
-
The Platonic Defense: Backdoor Defense for Self-Supervised Encoders in the Era of Large Scale Pre-training
Introduces an attack-agnostic black-box defense for SSL encoders that trains a conditional energy function via NCE and DSM to detect and purify representations, with an energy gap lower-bounded by mutual information.
-
Neural posterior estimation of Galactic Binary signals for the LISA mission
Conditional normalizing flows perform likelihood-free parameter estimation for single and overlapping LISA galactic binaries, generating thousands of posterior samples per second after training on simulations.
-
The Impact of Host Galaxy Properties on Supernova Classification with Hierarchical Labels
Host galaxy properties enable >90% pure Type Ia samples from photometry alone and improve classification accuracy when redshift is unavailable, via a new hierarchical cross-entropy objective.
-
21cmEMUv3: a hybrid diffusion-LSTM emulator of 21cmFAST summary observables
21cmEMUv3 emulates the cylindrical 21cm power spectrum via score-based diffusion and six other 21cmFAST observables via LSTM networks at sub-percent accuracy, then uses the emulator to infer a lower limit on soft-band X-ray luminosity from HERA data.
-
First steps towards gauge-independent vortex identification through machine learning
A neural network trained on 2D SU(2) lattices with inserted thin Z2 vortices, after random gauge transformations, noise, and cooling, can locate center vortices at moderate visibility levels and scales via tiling.
-
GenSBI: Generative Methods for Simulation-Based Inference in JAX
GenSBI delivers JAX-native implementations of generative SBI methods with transformer backbones and reports near-ideal calibration scores on standard benchmarks.
-
RT-Lynx: Putting the GEMM Sparsity In a Right Way for Diffusion Models
RT-Lynx shifts DiT sparsity from weights to activations, reports up to 1.55x linear-layer speedup while preserving generation quality across multiple diffusion models.
-
Bargmann Zeros as a Diagnostic of the Tunneling Transition in Double-Well Quantum Systems
Bargmann zeros of double-well eigenstates condense on the imaginary axis as a signature of the tunneling regime, obtained via variational wavefunctions projected to the Fock basis.