Lib.rs

The caffe2op-accum crate provides an operator for accumulating values in networked digital signal processing and deep learning computations. This crate defines the AccumulateOp which performs accumulation on the input tensors according to a specified axis. The resulting tensor is returned as the output. The Accumulated token is used to refer to the output tensor.

Note: This crate is currently being translated from C++ to Rust, and some function bodies may still be in the process of translation.

The operation performed by AccumulateOp can be represented mathematically as follows:

If the input tensor is of shape [a, b, c, ..., m, n] and the accumulation axis is k, then the output tensor will be of shape [a, b, c, ..., l, n] where l is the size of the tensor along the accumulation axis. The elements of the output tensor are calculated as follows:

output[i1, i2, ..., il, j] 

= sum(input[i1, i2, ..., i_{k-1}, l, i_{k+1}, ..., i_{m}, j])

where the sum is over all l along the accumulation axis.

The caffe2op-accum crate also provides tokens such as Accumulation, accumulates, accumulations, depends, fiddles, interim, and reshaped, which are used within the implementation of the accumulation operation.