cuequivariance-jax

To execute equivariant polynomials in JAX, you use cuequivariance primitives to perform representation-aware computations. This enables GPU-accelerated workflows for building neural networks that strictly respect symmetry groups.

Yes, you can build symmetry-aware models using Flax NNX layers. The integration provides specific modules like IrrepsLinear and SphericalHarmonics to construct neural networks that maintain equivariant properties.

Supported representations for equivariant polynomial evaluation include RepArray and ir_dict structures. These representations fit different workflows and ensure consistent, representation-aware computations across your symmetry groups.

Segmented polynomials handle GPU execution in JAX by utilizing naive and uniform_1d backends alongside per-irrep handling. This approach enables efficient GPU execution for large-scale equivariant neural network computations.

You need representation-aware computations for symmetry groups when designing neural networks that must remain geometrically consistent under transformations. Equivariant polynomials ensure network outputs transform predictably with input symmetries.