Skip to main content

split_tensor

onnx_runtime_session::sequence

Function split_tensor 

Source 
pub fn split_tensor(
    input: &SeqTensor,
    axis: i64,
    split: SplitSpec<'_>,
    keepdims: bool,
) -> SequenceResult<SequenceValue>
Expand description

Split a shared tensor into metadata-only views over the same allocation.

No tensor bytes are copied. Every returned element owns an Arc clone of input’s storage and records its own shape, strides, and byte offset.