For the complete documentation index, see llms.txt. Markdown versions of all pages are available by appending .md to any URL (e.g. /max/get-started.md).

Mojo function

gather_reduce

def gather_reduce[dtype: DType, gather_axis: Int, reduce_axis: Int, simd_width: Int, reduce_fn: def[dtype: DType, width: Int](SIMD[dtype, width], SIMD[dtype, width]) -> SIMD[dtype, width]](output: TileTensor[dtype, address_space=output.address_space, linear_idx_type=output.linear_idx_type, element_size=output.element_size], input: TileTensor[dtype, address_space=input.address_space, linear_idx_type=input.linear_idx_type, element_size=input.element_size], indices: TileTensor[DType.int32, address_space=indices.address_space, linear_idx_type=indices.linear_idx_type, element_size=indices.element_size], reduce_init: Scalar[dtype], ctx: Optional[DeviceContext] = None)

Computes output[i, j, k] = input[indices[i, j], k] and simultaneously reduces the output across axis 1 to produce output[i, k].

The motivating use-case for this is multi-hot embeddings in recommender models. This provides similar functionality to Torch's EmbeddingBag layer. In that context, i is the batch dimension, j is the multi-hot dimension, and k is the embedding dimension.