Map CPU-bound Python workloads to NVIDIA GPU libraries like CuPy and cuDF.
npx skills add https://github.com/CRAG666/dotfiles --skill optimize-for-gpu-crag666
Please help me install this Agent Skill. Skill: optimize-for-gpu Source: https://github.com/CRAG666/dotfiles/tree/main/skills/optimize-for-gpu Command: npx skills add https://github.com/CRAG666/dotfiles --skill optimize-for-gpu-crag666
This Skill helps you accelerate slow, CPU-bound Python workloads by mapping them onto NVIDIA GPU libraries to achieve large speedups for numerics, dataframes, ML, graphs, images, geospatial, IO, and vector search.
Use the optimize-for-gpu skill to rewrite your CPU-bound Python code into a RAPIDS/CUDA-accelerated version, keeping data on the GPU and selecting the best matching library for your bottleneck, for example focusing on CuPy for NumPy-like array operations or cuDF/cuML/cuGraph for dataframe/ML/graph workloads.
High-intent search queries and answers about installing and using this skill.
To accelerate CPU-bound Python, map your computation to NVIDIA GPU libraries like CuPy for NumPy arrays or cuDF for pandas dataframes, keeping data on-device to minimize host transfers and achieve large speedups.
The best way to run pandas and scikit-learn workloads on the GPU is using drop-in replacements like cuDF and cuML, which allow you to transform existing dataframe and machine learning code for GPU acceleration with minimal rewrites.
GPU acceleration minimizes host-device memory transfer overhead by keeping data resident on the device, batching workloads, and applying GPU-appropriate memory and dtype choices to reduce expensive host-to-GPU copies.
Yes, you can use Numba CUDA or Warp to write targeted rewrites for custom physics simulations, differentiable rendering, and mesh ray casting, mapping specialized geometry and particle system computations directly onto NVIDIA GPUs.
Yes, this GPU acceleration approach supports geospatial analysis and medical image processing by mapping these workloads to specialized NVIDIA libraries like cuSpatial for geospatial data and cuCIM for medical imaging tasks.
Use CuPy for NumPy-like array operations and sparse eigensolvers, while RAPIDS libraries like cuDF, cuML, and cuGraph are better suited for accelerating pandas dataframes, scikit-learn models, and NetworkX graph analytics respectively.
jasrajtulsiConvert CPU-bound Python workloads to NVIDIA GPU pipelines with RAPIDS libraries.
galeepAccelerate Python code execution on NVIDIA GPUs using CuPy, Numba, and cuDF.
mkurmanEnable GPU acceleration for Python workloads across CUDA-enabled libraries.
reverb256Manage power, cooling, and performance settings for multi-GPU and CPU systems.
