Module pruning 

Structured sparsity patterns and pruning for TensorLogic.

Provides magnitude-based and structured pruning to reduce model size while maintaining hardware efficiency through regular sparsity patterns.

Structs

MagnitudePruner

Magnitude-based pruner: zero out elements with smallest absolute values.

PruningConfig

Configuration for the pruning process.

SparsityStats

Statistics about the sparsity of a pruned tensor.

Enums

PruningError

Errors from pruning operations.

SparsityPattern

Structured sparsity pattern types.

Functions

compute_sparsity

Compute sparsity statistics for an N-D tensor.

row_norms

Compute per-row L2 norms for a 2D matrix.