windows-cupy-nvrtc

CuPy NVRTC compilation errors on Windows occur when CUDA binaries aren't discoverable at runtime. This Skill dynamically adds CUDA directories to PATH before importing CuPy, resolving errors like NVRTC_ERROR_BUILTIN_OPERATION_FAILURE and missing nvrtc-builtins64 DLLs on Windows 10/11 with CUDA 12.x and CuPy 13.x.

This error occurs when Windows cannot locate NVRTC runtime libraries because CUDA installation directories aren't in the system PATH. The Skill detects nvrtc DLL presence and automatically configures PATH at runtime or per-session to make CUDA binaries discoverable before CuPy GPU operations execute.

Yes. This Skill provides Jupyter notebook integration to add GPU support with a single setup cell at notebook start, automatically detecting and configuring CUDA paths so CuPy GPU operations work without manual environment configuration.

This Skill is designed specifically for Windows 10/11 environments running CUDA 12.x and CuPy 13.x in Conda or pip setups. It handles runtime PATH modification and DLL detection for these versions to resolve NVRTC compilation failures.

The Skill supports Windows 10 and Windows 11 with CUDA 12.x and CuPy 13.x installed via Conda or pip. It modifies PATH at runtime or per-session and detects nvrtc DLL presence to ensure CUDA binaries are discoverable.

No. This Skill automatically configures your Windows environment by detecting CUDA directories and adding them to PATH at runtime before importing CuPy, eliminating the need for manual environment variable editing or system restarts.