Maintainer execution helpers

These modules are documented for maintainers because they sit near dispatch and quantized execution helpers. They are not the preferred public import surface for user code.

Relation execution

mlx_lattice.ops._relation_exec.sparse_quantized_conv_features_from_relation(feats, weight, relation)[source]

Execute quantized sparse convolution over a prebuilt relation.

Return type:

array

Parameters:
mlx_lattice.ops._relation_exec.sparse_conv_features_from_relation(feats, weight, relation)[source]

Execute sparse convolution over a prebuilt relation.

Return type:

array

Parameters:
mlx_lattice.ops._relation_exec.sparse_conv_features_sorted_from_relation(feats, weight, relation, *, store_sorted=False)[source]

Execute the sorted implicit-GEMM convolution path explicitly.

Return type:

array

Parameters:
mlx_lattice.ops._relation_exec.sparse_conv_features_sorted_direct_reference_from_relation(feats, weight, relation, *, store_sorted=False)[source]

Execute the sorted direct reference path for diagnostics/tests.

Return type:

array

Parameters:
mlx_lattice.ops._relation_exec.sparse_pool_features_from_relation(feats, relation, *, input_exclusive, mode)[source]

Execute sparse pooling over a prebuilt kernel relation.

Return type:

array

Parameters:

Quantized execution

mlx_lattice.ops._quantized.quantized_matmul(feats, weight)[source]

Apply affine packed-weight matrix multiplication to feature rows.

Return type:

array

Parameters: