The Agent Skills Directory

[COMMAND_EXECUTION]: The file triangle/triton_utils.py uses subprocess.run to execute ptxas --version. This is a standard environment discovery mechanism used to identify the installed CUDA compiler version and ensure compatibility with the generated GPU kernels.
[DYNAMIC_EXECUTION]: The skill integrates with the Triton compiler to generate and execute GPU kernels at runtime. This is the primary intended functionality of the library, providing high-performance implementations for equivariant tensor products and attention mechanisms used in scientific computing and AI models.
[SAFE]: The skill uses environment variables such as CUEQ_TRIMUL_FALLBACK_THRESHOLD, CUE_PRINT_STATS, and CUEQUIVARIANCE_MAX_PATH_UNROLL for configuration and performance tuning, which is consistent with its role as a high-performance numerical library.

cuequivariance-jax