Optimize PyTorch and CUDA training on NVIDIA GPUs with 8–24GB VRAM.
npx skills add https://github.com/Mathews-Tom/praxis-skills --skill gpu-optimizer
Please help me install this Agent Skill. Skill: gpu-optimizer Source: https://github.com/Mathews-Tom/praxis-skills/tree/main/skills/gpu-optimizer Command: npx skills add https://github.com/Mathews-Tom/praxis-skills --skill gpu-optimizer
Provides evidence-based optimization patterns to reduce out-of-memory errors, improve training throughput, and migrate CPU-bound workflows to NVIDIA GPUs for consumer-class cards (8–24GB VRAM).
Optimize your model training on an NVIDIA GPU with 8–24GB VRAM by enabling BF16 or FP16, turning on gradient checkpointing, and running the diagnostics to validate memory and performance.
High-intent search queries and answers about installing and using this skill.
Resolve PyTorch CUDA out-of-memory errors by enabling gradient checkpointing, activating BF16 or FP16 mixed precision, and implementing VRAM accumulation strategies to reduce peak memory usage on consumer NVIDIA cards.
Accelerate XGBoost training with CUDA by utilizing QuantileDMatrix, configuring the gpu_hist tree method, and applying device-aware training parameters to execute faster boosting operations directly on CUDA-enabled builds.
Migrate Pandas and NumPy workflows to cuDF and CuPy using established transition patterns that support zero-copy memory interoperability with PyTorch tensors for seamless GPU data pipeline execution.
torch.compile works with mixed precision for PyTorch throughput optimization by combining BF16/FP16 selection, GradScaler fallbacks, and fused optimizers to maximize training performance and increase GPU pipeline efficiency.
Choose between gradient checkpointing and mixed precision to prevent OOMs by diagnosing specific VRAM bottlenecks through peak memory monitoring and profiling, applying clear trade-offs and anti-patterns to preserve GPU pipeline efficiency.
galeepAccelerate Python code execution on NVIDIA GPUs using CuPy, Numba, and cuDF.
dongzhuoyaoOptimize PyTorch training speed and memory usage on CUDA GPUs.
mkurmanEnable GPU acceleration for Python workloads across CUDA-enabled libraries.
cxnaiveImplement fully sharded data-parallel training for PyTorch models across multi-GPU clusters.
