The paper introduces a method to compute Gaussian kernel sums—central to many kernel-based machine learning techniques—using flash attention, a highly optimized softmax attention implementation. By applying two small input augmentations, the authors transform the normalized softmax reduction into an unnormalized Gauss sum, eliminating the need for custom GPU code. For feature dimensions greater than eight in fp16, this approach outperforms both compiled PyTorch code and PyKeOps kernels in speed, memory overhead, and accuracy, while maintaining linear memory scaling.
arXiv:2606. 10896v1 Announce Type: new Abstract: We present \textbf{Flash-GMM}, a fused Triton kernel for efficient computation of Gaussian Mixture Models (GMMs) over large-scale data in a single GPU pass.
FlashBoB introduces an I/O‑efficient algorithm for exact backward‑over‑backward (BoB) in softmax attention, enabling precise second‑order differentiation without large intermediate tensors. By exploiting a hierarchical affine structure, the method confines computation to on‑chip tiles and limits off‑chip memory traffic, achieving θ(N² d²/M) HBM usage. Experiments show FlashBoB scales to sequence lengths of 262K on a single A100 GPU, outperforming prior exact baselines and FlashBack by up to 6.3×.
arXiv:2606. 20005v1 Announce Type: cross Abstract: Attention distillation, which trains one attention distribution to match another by minimizing their Kullback-Leibler (KL) divergence, is widely used in knowledge distillation, model compression, continual learning, and sparse-attention LLM training.
The paper investigates using short polynomial approximations to accelerate special‑function operations in large language models on NVIDIA Blackwell GPUs. By replacing native sigmoid, tanh, and SiLU with degree‑3 or degree‑4 bfloat16 programs, the authors achieve up to 2.19× speed‑ups in isolated FP16 benchmarks and modest training‑step throughput gains (2.7–8.0%) across four integration tasks. The study also evaluates model behavior, finding negligible training‑loss differences within 100 billion tokens.
arXiv:2609.15810v1 Announce Type: new Abstract: Diffusion Transformers deliver state-of-the-art video generation, but their long spatiotemporal sequences make attention the dominant deployment cost,...
arXiv:2606. 17781v1 Announce Type: cross Abstract: The rapid growth of Large Language Models (LLMs) has intensified the need for specialized hardware accelerators that can satisfy stringent inference latency and power constraints.
arXiv:2606. 13894v1 Announce Type: cross Abstract: AdamW is a default optimizer for modern deep learning, but its first and second moment states add roughly two parameter-sized buffers to training memory.
arXiv:2609.13612v1 Announce Type: new Abstract: Modern AI systems are built on the Transformer architecture, whose core operation, attention, accounts for the majority of computation and memory cost....
arXiv:2605. 01910v2 Announce Type: replace-cross Abstract: Autoregressive decoding becomes bandwidth-limited at long contexts, as generating each token requires reading all $n_k$ key and value vectors from KV cache.
arXiv:2508. 18224v3 Announce Type: replace-cross Abstract: Recent advances in sparse attention mechanisms have demonstrated strong potential for reducing the computational cost of long-context training and inference in large language models (LLMs).
arXiv:2607. 16241v1 Announce Type: cross Abstract: Recent large language models (LLMs) can generate custom CUDA kernels that appear to outperform PyTorch on benchmarks such as KernelBench.
arXiv:2607. 24762v1 Announce Type: new Abstract: Machine learning models are increasingly embedded in everyday software, and most of their runtime is spent in a small set of compute kernels such as matrix multiplication, convolution, and normalization.