Search Results for "cuda benchmark"

...Its architecture combines large-scale data synthesis, a skill-augmented CUDA development environment, and long-horizon reinforcement learning to build intrinsic optimization capability rather than relying on simple post-hoc tuning. The system operates in a ReAct-style loop where the agent profiles baseline implementations, writes CUDA code, compiles it in a sandbox, and iteratively refines performance. CUDA-Agent has demonstrated strong benchmark results, achieving high pass rates and significant speedups compared with compiler baselines such as torch.compile.

High performance, open-source, cross-platform RandomX, KawPow, CryptoNight, and AstroBWT CPU/GPU miner, RandomX benchmark, and stratum proxy. XMRig is a high-performance, open-source, cross-platform RandomX, KawPow, CryptoNight, and AstroBWT unified CPU/GPU miner and RandomX benchmark. Official binaries are available for Windows, Linux, macOS, and FreeBSD. The preferred way to configure the miner is the JSON config file as it is more flexible and human-friendly. The command-line interface...

It consists of various methods for deep learning on graphs and other irregular structures, also known as geometric deep learning, from a variety of published papers. In addition, it consists of an easy-to-use mini-batch loader for many small and single giant graphs, a large number of common benchmark datasets (based on simple interfaces to create your own), and helpful transforms, both for learning on arbitrary graphs as well as on 3D meshes or point clouds. We have outsourced a lot of functionality of PyTorch Geometric to other packages, which needs to be additionally installed. These packages come with their own CPU and GPU kernel implementations based on C++/CUDA extensions. ...

DeepGEMM is a specialized CUDA library for efficient, high-performance general matrix multiplication (GEMM) operations, with particular focus on low-precision formats such as FP8 (and experimental support for BF16). The library is designed to work cleanly and simply, avoiding overly templated or heavily abstracted code, while still delivering performance that rivals expert-tuned libraries. It supports both standard and “grouped” GEMMs, which is useful for architectures like Mixture of...

...The key innovation in this version is DeepSeek Sparse Attention (DSA), a sparse attention mechanism that aims to optimize training and inference efficiency in long-context settings without degrading output quality. According to the authors, they aligned the training setup of V3.2-Exp with V3.1-Terminus so that benchmark results remain largely comparable, even though the internal attention mechanism changes. In public evaluations across a variety of reasoning, code, and question-answering benchmarks (e.g. MMLU, LiveCodeBench, AIME, Codeforces, etc.), V3.2-Exp shows performance very close to or in some cases matching that of V3.1-Terminus. The repository includes tools and kernels to support the new sparse architecture—for instance, CUDA kernels, logit indexers, and open-source modules like FlashMLA and DeepGEMM are invoked for performance.

...Accuracies from research papers have just begun to surpass human accuracies on some benchmarks. The accuracies of open source face recognition systems lag behind the state-of-the-art. See our accuracy comparisons on the famous LFW benchmark.

...It provides training/testing scripts, a demo (tiny_face_detector.m), model loading, evaluation on WIDER FACE, and supporting utilities (e.g. cnn_widerface_eval.m). The code depends on MatConvNet, which must be compiled (with GPU / CUDA / cuDNN support) for full performance. Pretrained model provided (ResNet101-based, plus alternatives). Demo and evaluation scripts for benchmark datasets. Use of “foveal descriptors” to incorporate context for low-resolution faces. Pretrained model provided (ResNet101-based, plus alternatives).

Library for continuous convex optimization in image analysis, together with a command line tool and Matlab interface. Implements several recent algorithms for inverse problems and image segmentation with total variation regularizers and vectorial multilabel transition costs. Also included is a suite for variational light field analysis, which ties into the HCI light field benchmark set and givens reference implementations for a number of our recently published algorithms. *** NOTE: ...

This application can train SVM using LibSVM and several CUDA implementations. Supported input file formats are LibSVM text file and Bottou's LaSVM binary file. Wanted implementation can be chosen using command line parameter. Training, input data loading and output data saving times are measured and reported. Output model is saved in LibSVM text format.

GPUBench2 is a cross platform suite to analyze the performance of GPU( Graphics Processing Unit). GPUBench2 will gather of state-of-the-art parameters from different interfaces ( OpenGL, Cg, CUDA) available.