Search Results for "gpu as computation"
Sort By:
Showing 59 open source projects for "gpu as computation"
View related business solutions-
Veeam Data Platform v13.1 - Get Your Free TrialTry Veeam Data Platform today. Experience the unified platform that's secure by design, portable by default, and proven to recover clean, fast, and anywhere.
-
MongoDB Atlas runs apps anywhereMongoDB Atlas gives you the freedom to build and run modern applications anywhere—across AWS, Azure, and Google Cloud. With global availability in over 115 regions, Atlas lets you deploy close to your users, meet compliance needs, and scale with confidence across any geography.
-
1
GPU Puzzles
Solve puzzles. Learn CUDA
GPU Puzzles is an educational project designed to teach GPU programming concepts through interactive coding exercises and puzzles. Instead of presenting traditional lecture-style explanations, the project immerses learners directly in hands-on programming tasks that demonstrate how GPU computation works. The exercises are implemented using Python with the Numba CUDA interface, which allows Python code to compile into GPU kernels that run on CUDA-enabled hardware. ... -
2
DualPipe
A bidirectional pipeline parallelism algorithm
DualPipe is a bidirectional pipeline parallelism algorithm open-sourced by DeepSeek, introduced in their DeepSeek-V3 technical framework. The main goal of DualPipe is to maximize overlap between computation and communication phases during distributed training, thus reducing idle GPU time (i.e. “pipeline bubbles”) and improving cluster efficiency. Traditional pipeline parallelism methods (e.g. 1F1B or staggered pipelining) leave gaps because forward and backward phases can’t fully overlap with communication. DualPipe addresses that by scheduling micro-batches from both ends of the pipeline in a bidirectional fashion—i.e. some micro-batches flow forward while others flow backward—so that computation on one partition can coincide with communication for another. -
3
AirLLM
AirLLM 70B inference with single 4GB GPU
AirLLM is an open source Python library that enables extremely large language models to run on consumer hardware with very limited GPU memory. The project addresses one of the main barriers to local LLM experimentation by introducing a memory-efficient inference technique that loads model layers sequentially rather than storing the entire model in GPU memory. This layer-wise inference approach allows models with tens of billions of parameters to run on devices with only a few gigabytes of VRAM. ... -
4
Qulacs
Variational Quantum Circuit Simulator for Quantum Computation Research
Variational Quantum Circuit Simulator for Quantum Computation Research. Qulacs is a Python/C++ library for fast simulation of large, noisy, or parametric quantum circuits. Qulacs is developed at QunaSys, Osaka University, NTT, and Fujitsu. -
AI-generated apps that pass security reviewRetool lets you generate dashboards, admin panels, and workflows directly on your data. Type something like “Build me a revenue dashboard on my Stripe data” and get a working app with security, permissions, and compliance built in from day one. Whether on our cloud or self-hosted, create the internal software your team needs without compromising enterprise standards or control.
-
5
Halide
A language for fast, portable data-parallel computation
Halide is a programming language for fast, portable data-parallel computation. It was designed to make writing high-performance image and array processing code much easier on modern machines. It works on all major operating systems and with several CPU architectures (X86, ARM, MIPS, Hexagon, PowerPC) and GPU Compute APIs (CUDA, OpenCL, OpenGL, among others). It isn't a standalone programming language however; rather it is embedded in C++ which means that you write C++ code, building an in-memory representation of a Halide pipeline using Halide's C++ API. ... -
6
FlexLLMGen
Running large language models on a single GPU
...The architecture distributes computation and memory usage across the GPU, CPU, and disk in order to maximize the number of tokens processed during inference. This design allows organizations to deploy powerful language models for high-volume tasks without the infrastructure costs typically associated with large-scale AI systems. The project is particularly useful for workloads that prioritize throughput over latency, including benchmarking experiments and large corpus analysis. -
7
ImplicitGlobalGrid.jl
Distributed parallelization of stencil-based GPU and CPU applications
...Samuel Omlin) with Stanford University (Dr. Ludovic Räss) and the Swiss Geocomputing Centre (Prof. Yuri Podladchikov). It renders the distributed parallelization of stencil-based GPU and CPU applications on a regular staggered grid almost trivial and enables close to ideal weak scaling of real-world applications on thousands of GPUs [1, 2, 3]. ImplicitGlobalGrid relies on the Julia MPI wrapper (MPI.jl) to perform halo updates close to hardware limit and leverages CUDA-aware or ROCm-aware MPI for GPU-applications. ... -
8
Flash-MoE
Running a big model on a small laptop
Flash-MoE is a high-performance implementation of mixture-of-experts (MoE) architectures designed to optimize the efficiency and scalability of large AI models. It focuses on accelerating routing and computation by leveraging optimized kernels and memory management techniques, allowing models to dynamically select specialized sub-networks during inference. The project aims to reduce the computational cost typically associated with MoE systems while maintaining or improving performance. It likely includes support for GPU acceleration and parallel processing, enabling it to handle large-scale workloads effectively. ... -
9
SwissGL
SwissGL is a minimalistic wrapper on top of WebGL2 JS API
SwissGL is a compact JavaScript library that provides a streamlined abstraction layer over the WebGL2 API, designed to minimize boilerplate when building GPU-accelerated graphics, simulations, and procedural visualizations. Acting as a "Swiss Army knife" for WebGL2, it simplifies shader, texture, and framebuffer management into a single, expressive interface that enables developers to write complex GPU workflows in just a few lines of code. The library centers around one main function that... -
Earn up to 16% annual interest with Nexo.Generate interest, access liquidity without selling, and execute trades seamlessly. All in one platform. Geographic restrictions, eligibility, and terms apply.
-
10
PyTorch
Open source machine learning framework
PyTorch is a Python package that offers Tensor computation (like NumPy) with strong GPU acceleration and deep neural networks built on tape-based autograd system. This project allows for fast, flexible experimentation and efficient production. PyTorch consists of torch (Tensor library), torch.autograd (tape-based automatic differentiation library), torch.jit (a compilation stack [TorchScript]), torch.nn (neural networks library), torch.multiprocessing (Python multiprocessing), and torch.utils (DataLoader and other utility functions). ... -
11
CubeCL
Multi-platform high-performance compute language extension for Rust
CubeCL is a low-level compute language and compiler framework designed to simplify and optimize GPU programming for high-performance workloads, particularly in machine learning and numerical computing. It provides an abstraction layer that allows developers to write portable, hardware-efficient compute kernels without directly dealing with complex GPU APIs such as CUDA or OpenCL. CubeCL focuses on delivering predictable performance and composability by exposing explicit control over memory... -
12
TensorLy
Tensor Learning in Python
TensorLy is a Python library that aims at making tensor learning simple and accessible. It allows to easily perform tensor decomposition, tensor learning and tensor algebra. Its backend system allows to seamlessly perform computation with NumPy, PyTorch, JAX, TensorFlow, CuPy or Paddle, and run methods at scale on CPU or GPU. -
13
LLaMA-Factory
Unified Efficient Fine-Tuning of 100+ LLMs & VLMs (ACL 2024)
LLaMA-Factory is a fine-tuning and training framework for Meta's LLaMA language models. It enables researchers and developers to train and customize LLaMA models efficiently using advanced optimization techniques. -
14
Zoo Design Studio
The Zoo Design Studio app
...Users can interact with the system through a familiar point-and-click interface, but every action is translated into code in the underlying modeling language, ensuring consistency between visual and programmatic representations. The application is powered by a GPU-first geometry engine that streams rendered output as video frames, enabling high-performance modeling even when heavy computation is offloaded to remote infrastructure. It uses WebSockets for real-time communication between the client and the modeling engine, allowing immediate feedback and interactive design updates. -
15
Faster Whisper
Faster Whisper transcription with CTranslate2
Faster Whisper is an optimized implementation of the Whisper speech recognition model designed to deliver significantly faster inference while maintaining comparable accuracy. It leverages efficient inference engines and optimized computation strategies to reduce latency and resource consumption. The system is particularly useful for real-time or large-scale transcription tasks where performance is critical. It supports multiple model sizes, allowing users to balance speed and accuracy based... -
16
CTranslate2
Fast inference engine for Transformer models
...The project implements a custom runtime that applies many performance optimization techniques such as weights quantization, layers fusion, batch reordering, etc., to accelerate and reduce the memory usage of Transformer models on CPU and GPU. The execution is significantly faster and requires less resources than general-purpose deep learning frameworks on supported models and tasks thanks to many advanced optimizations: layer fusion, padding removal, batch reordering, in-place operations, caching mechanism, etc. The model serialization and computation support weights with reduced precision: 16-bit floating points (FP16), 16-bit integers (INT16), and 8-bit integers (INT8). ... -
17
Meridian
Meridian is an MMM framework
...The framework provides a robust foundation for constructing in-house MMM pipelines capable of handling both national and geo-level data, with built-in support for calibration using experimental data or prior knowledge. Meridian uses the No-U-Turn Sampler (NUTS) for Markov Chain Monte Carlo (MCMC) sampling to produce statistically rigorous results, and it includes GPU acceleration to significantly reduce computation time. -
18
clip-retrieval
Easily compute clip embeddings and build a clip retrieval system
clip-retrieval is an open-source toolkit designed to build large-scale semantic search systems for images and text by leveraging CLIP embeddings to enable multimodal retrieval. It allows developers to compute embeddings for both images and text efficiently and then index them for fast similarity search across massive datasets. The system is optimized for performance and scalability, capable of processing tens or even hundreds of millions of embeddings using GPU acceleration. It includes... -
19
Profile Data
Analyze computation-communication overlap in V3/R1
profile-data is a repository that publishes profiling traces and metrics from DeepSeek’s training and inference infrastructure (especially during DeepSeek-V3 / R1 experiments). The profiling data targets insights into computation-communication overlap, pipeline scheduling (e.g. DualPipe), and how MoE / EP / parallelism strategies interact in real systems. The repository contains JSON trace files like train.json, prefill.json, decode.json, and associated assets. Users can load them into tools like Chrome tracing to inspect GPU idle times, overlapping operations, and scheduling alignment. ... -
20
uzu
A high-performance inference engine for AI models
uzu is a high-performance inference engine designed to run artificial intelligence models efficiently on Apple Silicon hardware. Written primarily in Rust and leveraging Apple’s Metal framework, the project focuses on maximizing performance when executing large language models and other AI workloads on devices such as Mac computers with M-series chips. The engine implements a hybrid architecture in which model layers can be executed either as custom GPU kernels or through Apple’s MPSGraph... -
21
EvoTorch
Advanced evolutionary computation library built on top of PyTorch
EvoTorch is an evolutionary optimization framework built on top of PyTorch, developed by NNAISENSE. It is designed for large-scale optimization problems, particularly those that require evolutionary algorithms rather than gradient-based methods. -
22
SageAttention
NeurIPS2025 Spotlight] Quantized Attention
SageAttention is an open-source optimization library designed to accelerate the attention mechanism used in transformer-based neural networks. Since attention operations are often the most computationally expensive component of modern AI models, SageAttention introduces quantization techniques that significantly reduce computational overhead while preserving model accuracy. The system achieves this by using low-precision numerical formats such as INT4, FP8, or INT8 to represent key matrices... -
23
Bend
A massively parallel, high-level programming language
Bend is an interactive programming environment (REPL) built on top of the Kotlin language, designed to allow users to explore, experiment, and learn Kotlin in a live, feedback-driven manner. The tool lets you define variables, functions, or values at the prompt and iteratively refine them—immediately seeing output and types—while preserving state across commands. It emphasizes discoverability and experimentation: users can inspect functions, call them on sample inputs, and evolve logic... -
24
Covalent workflow
Pythonic tool for running machine-learning/high performance workflows
Covalent is a Pythonic workflow tool for computational scientists, AI/ML software engineers, and anyone who needs to run experiments on limited or expensive computing resources including quantum computers, HPC clusters, GPU arrays, and cloud services. Covalent enables a researcher to run computation tasks on an advanced hardware platform – such as a quantum computer or serverless HPC cluster – using a single line of code. Covalent overcomes computational and operational challenges inherent in AI/ML experimentation. -
25
Koila
Prevent PyTorch's `CUDA error: out of memory` in just 1 line of code
Koila is a lightweight Python library designed to help developers avoid memory errors when training deep learning models with PyTorch. The library introduces a lazy evaluation mechanism that delays computation until it is actually required, allowing the framework to better estimate the memory requirements of a model before execution. By building a computational graph first and executing operations only when necessary, koila reduces the risk of running out of GPU memory during the forward pass of neural network training. This approach enables developers to experiment with larger batch sizes and more complex architectures while maintaining stable training behavior. ...
